University of Southern California Dissertations and Theses

Electric vehicle integration into the distribution grid: impact, control and forecast

(USC Thesis Other)

Electric vehicle integration into the distribution grid: impact, control and forecast

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content USC Viterbi School of Engineering PhD Thesis Electric Vehicle Integration into the Distribution Grid: Impact, Control and Forecast Zeroing Jiang A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) August 2017 USC Viterbi School of Engineering PhD Thesis Table of Content TABLE OF CONTENT ..................................................................................................... I LIST OF FIGURES ........................................................................................................ VI LIST OF TABLES ......................................................................................................... XII ABBREVIATIONS ........................................................................................................ XV ABSTRACT ............................................................................................................... XVIII ACKNOWLEDGEMENTS ........................................................................................ XIX 1. CHAPTER 1: INTRODUCTION ............................................................................. 1 1.1 THESIS BACKGROlJND ............................................................................................ 1 1. 1.1 Smart Grid Reginal Demonstration Project Original Scope Statement.. .......... 1 1. 1. 2 Overview of EV and Charging Levels ............................................................... 2 1. 1. 3 Impact of EV Battery Charging on the Grid ...................................................... 4 1.1. 4 Battery Aggregation and Backfill, Vehicle-to-Grid Technology and Renewable Energy .......................................................................................................................... 5 1. 1. 5 Estimate EV Charging Demand ..................................................................... ... 6 1.2 THESIS OBJECTIVES ............................................................................................... 6 1.3 CONTRIBUTIONS TOKNOWLEDGE .......................................................................... 8 1.4 PUBLICATIONS ARISING FROM THIS THESIS ................ ............................................ 9 1.5 THESIS STRUCTURE .............................................................................................. 10 2. CHAPTER 2: IMPACT OF EV INFRASTRUCTURE CHARGING ON THE GRID WITH REGARD TO SYSTEM LOAD FLOW, LOAD FACTOR, AND STAB IL I TY ..................................................................................................................... 12 2.1 RESEARCH OVERVIEW ......................................................................................... 12 2.2 SYSTEMS MODELING AND REPRESENTATION UTILIZED ........................................ 13 2. 2.1 Technologies and Systems Used ...................................................................... 13 2.2.2 Configurations of the Smart Grid Systems, Subsystems and Components ...... 14 2.3 DESCRIPTION OF THE METHODOLOGIES AND ALGORITHMS UTILIZED .................. 19 2. 3.1 Distribution System Load Level Consideration ............................................... 20 2. 3. 2 EV Load Integration Scenarios ....................................................................... 21 2.3.3 Load Flow, Load Factor and Stability Limits ................................................. 23 2.4 SUMMARIES OF THE RESULTS OF THE PERFORMANCE OF THE SMART GRID SYSTEMS AND TECHNOLOGIES DERIVED FROM LAB TESTS, FIELD T ESTS, OR GRID- CONNECTED APPLICA TIONS ............................................................................................ 24 2. 4.1 IEEE 34 Node Distribution Test System .......................................................... 24 2. 4. 2 USC Micro-grid ............................................................................................... 38 2. 4. 3 Chatsworth Distribution System ...................................................................... 45 USC Viterbi School of Engineering PhD Thesis 2.5 SUMMARIES OF THE RESULTS OF THE ANALYSIS AND BENEFITS ........................... 58 2. 5.1 IEEE 34 Node Distribution Test System .......................................................... 58 2.5.2 USC Micro-grid ............................................................................................... 60 2.5.3 Chatsworth Distribution System ...................................................................... 61 3. CHAPTER 3: DISTRIBUTION EFFECT OF BATTERY AGGREGATION AND BACKFILL COUPLED WITH RENEWABLE ENERGY ............................... 65 3 .1 RESEARCH OVERVIEW ......................................................................................... 65 3.2 OVERVIEW OF THE SYSTEMS UTILIZED ................................................................ 67 3. 2.1 Technologies and Systems Used ...................................................................... 67 3.2.2 Configurations of the Smart Grid Systems, Subsystems and Components ...... 68 3.3 DESCRIPTION OF THE METHODOLOGIES AND ALGORITHMS UTILIZED .................. 69 3. 3.1 Demand Load Level ......................................................................................... 69 3. 3. 2 EV Charging/Discharging Strategies .............................................................. 69 3. 3. 3 Battery Aggregation and Backfill Strategies in Microsoft Excel ..................... 70 3. 3. 4 Battery Aggregation and Backfill Strategies in MATLAB Modeling. .............. 77 3.3.5 EV Coupled with Storage in PSCAD ............................................................... 79 3 .4 SUMMARIES OF THE RESULTS OF THE PERFORMANCE OF THE SMART GRID SYSTEMS AND TECHNOLOGIES DERIVED FROM LAB TESTS, FIELD TESTS, OR GRID-CONNECTED APPLICATIONS ................................................................................................................. 83 3. 4.1 EV Charging/Discharging Strategy ................................................................. 83 3. 4. 2 Battery Aggregation and Backfill Strategies in Microsoft Excel ..................... 84 3. 4. 3 Battery Aggregation and Backfill Strategies in MATLAB Modeling. .............. 87 3.4.4 EV Coupled with Storage in PSCAD ............................................................... 90 3.5 SUMMARIES OF THE RESULTS OF THE ANALYSIS AND BENEFITS ........................... 92 3. 5.1 EV Charging/Discharging Strategies .............................................................. 9 2 3.5.2 EV Coupled with Storage in PSCAD ............................................................... 93 3.5.3 Conclusion ....................................................................................................... 93 4. CHAPTER 4: EV CHARGING PATTERNS AND POWER USAGE OF CHARGING STATIONS ............................................................................................... 95 4.1 RESEARCH OVERVIEW ......................................................................................... 95 4.2 DATA COLLECTION ANDTESTFLEET .................................................................... 96 4. 2.1 Large Corporate Fleet and Charger Information ........................................... 96 4.2.2 LAD WP Local Residential Fleet and Charger Information .......................... 101 4. 2. 3 LAD WP Ride Share Fleet and Charger Information .................................... 103 4.3 EV CHARGING PATTERN .................................................................................... 105 4. 3.1 Charging Pattern of Large Corporate Fleet ................................................. 105 4.3.2 Charging Pattern of LADWP Ride Share EVs .............................................. 108 4.4 MONITOR OF LARGE CORPORATE CHARGING STATION ....................................... 110 4.4.1 Station Usage in LADWP ............................................................................... 110 ii USC Viterbi School of Engineering PhD Thesis 4. 4. 2 Station Usage in UCLA .................................................................................. 115 4.5 MONITOR OF LAD WP LOCAL RESIDENTIAL CHARGERS ..................................... 119 4. 5.1 Hourly Charging Distribution Analysis ......................................................... 119 4. 5. 2 Weekday vs Weekend Energy Distribution ...................................................... 119 4. 5. 3 Energy Distribution in Different Month ........................................................ 120 4. 5. 4 Three Month Cumulative Energy Distribution .............................................. 121 4.6 MONITOR OF LAD WP RIDE SHARE CHARGING STATION ................................... 122 4. 6.1 Methodology .................................................................................................. 122 4. 6. 2 8760-Hour Charging Distribution Analysis .................................................. 122 4. 6. 3 Weekday vs Weekend Energy Distribution. .................................................... 123 4. 6. 4 Energy Distribution in Different Seasons ...................................................... 123 4. 6. 5 One Year Cumulative Energy Distribution .................................................... 124 4.7 LOAD CURVE PREDICTION IN CALIFORNIA WITH DIFFERENT EV PENETRATION RATES ........................................................................................................................... 125 4. 7.1 Original Load Profile in California .............................................................. 125 4. 7. 2 California EV Information ............................................................................ 125 4. 7. 3 California Load Curve Prediction ................................................................. 126 4.8 CONCLUSION ...................................................................................................... 128 4. 8.1 Conclusion for Large Corporate Fleet and Associated Charging Stations .. 128 4.8.2 Conclusion for Residential EVandAssociated Charging Stations ............... 129 4.8.3 Conclusion for LAD WP Ride Share EV and Associated Charging Stations. 130 5. CHAPTER 5: ANALYSIS ON EV USER BEHAVIORS ................................... 131 5 .1 RESEARCH OVERVIEW ....................................................................................... 131 5.2 LADWPEVUSERHABIT ANALYSIS BASED ON ONE YEAR SURVEY ................. 132 5. 2.1 Raw Data and Overall Analysis .................................................................... 13 2 5.2.2 Comparison of all EV Users .......................................................................... 135 5.3 LADWPEVUSERHABIT ANALYSIS BASED ON FLEETCARMA ......................... 137 5.3.1 Methodology .................................................................................................. 137 5. 3. 2 Overview of all EV Usage and Charging ...................................................... 138 5. 3. 3 Overview of LAD WP Assigned EVs and Carpool EVs .................................. 145 5.3.4 Individual EV Users Habit Analysis ............................................................. 158 5. 3. 5 Comparison of EV and non-EV. .................................................................... 158 5.3.6 Comparison of all EV Users .......................................................................... 159 5.4 THE STUDY OF EV USER'S HABIT FOR LARGE CORPORATE FLEET- UCLA USERS .. ........................................................................................................................... 161 5.4.1 Methodology .................................................................................................. 162 5. 4. 2 UCLA Users Behavior Analysis ................................................................... 162 5.4.3 Comparison ................................................................................................... 162 5.5 THE STUDYOFEVUSER' S HABIT FOR TYPICAL OWNERS IN THE CITY .............. 163 5. 5.1 Methodology .................................................................................................. 164 iii USC Viterbi School of Engineering PhD Thesis 5. 5. 2 Typical Owners Behavior Analysis ............................................................... 164 5. 5. 3 Comparison ................................................................................................... 164 5.6 CONCLUSION ...................................... ................................................................ 167 5.6.1 LAD WP EV User Habit ................................................................................. 167 5.6.2 UCLA EV User Habit .................................................................................... 169 5.6.3 Typical Owner Habit ..................................................................................... 169 6. CHAPTER 6: EV CHARGING DEMAND ESTIMATION AND FORECAST .... .................................................................................................................................. 172 6.1 RESEARCH OVERVIEW ....................................................................................... 172 6.2 INDUSTRIAL EV CHARGING DEMAND ESTIMATION AND FORECAST ................... 173 6.2.1 Methodology .................................................................................................. 175 6.2.2 Industrial Compact EV Demand Forecast Model ......................................... 184 6.2.3 Industrial Midsize EV Demand Forecast Model ........................................... 194 6.2.4 Industrial Electric SUV Demand Forecast Model ........................................ 195 6.2.5 Complete Industrial EV Demand Forecast Model ........................................ 195 6.3 COMMERCIAL EV CHARGING DEMAND ESTIMATION AND FORECAST ................ 199 6.3.1 Methodology .................................................................................................. 200 6.3.2 Commercial Compact EV Demand Forecast Model ..................................... 201 6.3.3 Commercial Electric SUV Demand Forecast Model... .................................. 201 6.3.4 Complete Commercial EV Demand Forecast Model .................................... 201 6.4 RESIDENTIAL EV CHARGING DEMAND ESTIMATION AND FORECAST ................. 205 6.4.1 Assumptions ................................................................................................... 205 6.4.2 Methodology .................................................................................................. 205 6. 4. 3 Monte Carlo Simulation Result ..................................................................... 206 6.5 CONCLUSION ...................................................................................................... 210 7. CHAPTER 7: CONCLUSIONS AND FUTURE WORK ................................... 211 7.1 THESIS CONTRIBUTION ....................................................................................... 211 7 .2 IMPACT OF EV INFRASTRUCTURE CHARGING ON THE GRID ............................... 212 7.2.1 Summary of Research Work ........................................................................... 212 7.2.2 Limitations and Suggestions for Future Work ............................................... 213 7.3 DISTRIBUTION EFFECT OF BATTERY AGGREGATION AND BACKFILL COUPLED WITH RENEWABLE ENERGY ................................................................................................... 213 7.3.1 Summary of Research Work ........................................................................... 213 7.3.2 Limitations and Suggestions for Future Work ............................................... 214 7.4 EV CHARGING PATTERNS AND POWER USAGE OF CHARGING S TATIONS ............ 215 7.4.1 SummaryofResearch Work ........................................................................... 215 7. 4. 2 Limitations and Suggestions for Future Work ............................................... 215 7.5 ANALYSIS ON EVUSERBEHAVIORS ................................................................... 216 7. 5.1 Summary of Research Work ........................................................................... 216 iv USC Viterbi School of Engineering PhD Thesis 7.5.2 Limitations and Suggestions for Future Work ............................................... 216 7.6 EV CHARGING DEMAND ESTIMATION AND FORECAST ....................................... 216 7. 6.1 Summary of Research Work ........................................................................... 216 7.6.2 Limitations and Suggestions for Future Work ............................................... 217 REFERENCE ................................................................................................................ 218 APPENDIX .................................................................................................................... 226 A. INDUSTRIAL MIDSIZE EV DEMAND FORECAST MODEL ......................................... 226 B. INDUSTRIAL ELECTRIC SUV DEMAND FORECAST MODEL .................................... 236 C. COMMERCIAL COMPACT EV DEMAND FORECAST MODEL .................................... 245 D. COMMERCIAL ELECTRIC SUV DEMAND FORECAST MODEL ................................. 256 E. MATLAB CODE FOR EV DEMAND FORECAST MODEL .......................................... 266 v USC Viterbi School of Engineering PhD Thesis List of Figures FIGURE 1.1 RESPONSIBILITIES OF EACH PARTNER IN SGRDP .............................................. 2 FIGURE 1.2 V2G CONCEPT [84] ........................................................................................... 5 FIGURE 2.1 ONE-LINE DIAGRAM OF IEEE 34-BUS TEST FEEDER ....................................... 15 FIGURE 2.2 ONE-LINE DIAGRAM OF USC MICRO-GRID ..................................................... 16 FIGURE 2.3 ONE-LINE DIAGRAM OF THE CHATSWORTH DISTRIBUTION SYSTEM ................ 18 FIGURE 2.4 CAISO's SUMMER LOAD CURVE AS AREPRESENTATIVE FOR DEVELOPING LOAD PERIODS ............................................................................................................ 20 FIGURE 2.5 LOAD PERIODS DEFINITION ............................................................................. 21 FIGURE 2. 6 RESULT FOR CASE 1.1 .................. .................................................................. 25 FIGURE 2.7 RESULT FOR CASE 1.2 .................................................................................... 25 FIGURE 2. 8 RESULT FOR CASE 1. 3 .................. .................................................................. 25 FIGURE 2. 9 RESULT FOR CASE 2.1 .................................................................................... 26 FIGURE 2.10 RESULT FOR CASE 2.2 .................................................................................. 26 FIGURE 2.11 RESULT FOR CASE 2.3 .................................................................................. 26 FIGURE 2.12 RESULT FOR CASE 3.1 .................................................................................. 27 FIGURE 2.13 RESULT FOR CASE 3 .2 ........................................ .......................................... 27 FIGURE 2.14 RESULT FOR CASE 3.3 .................................................................................. 27 FIGURE 2.15 IEEE34 BUS SYSTEM VOLTAGE PROFILE Vs DISTANCE FROM DISTRIBUTION SUBSTATION ................................................................................................................ 28 FIGURE 2.16 CASE 1 RESULTS IN 0PENDSS (A) ................................................................. 30 FIGURE 2.17 CASE 1 RESULTS IN OPENDSS (B) ..................... ········· ................................... 30 FIGURE 2.18 CASE 1 RESULTS IN 0PENDSS ( c) ................................................................. 31 FIGURE 2.19 CASE 2 R ESULTS IN 0PENDSS (A) ................................................................. 31 FIGURE 2.20 CASE 2 RESULTS IN 0PENDSS (B) ................................................................. 32 FIGURE 2.21CASE3 RESULTS IN 0PENDSS (A) ................................................................. 32 FIGURE 2.22 CASE 3 RESULTS IN OPENDSS (B) ................................................................. 33 FIGURE 2.23 CASE 3 RESULTS IN 0PENDSS (c) ................................................................. 33 FIGURE 2.24 CASE 4 RESULTS IN 0PENDSS (A) ................................................................. 33 FIGURE 2.25 CASE 4 RESULTS IN 0PENDSS (B) ................................................................. 34 FIGURE 2.26 CASE 5 RESULTS IN 0PENDSS (A) ................................................................. 35 FIGURE 2.27 CASE 5 RESULTS IN OPENDSS (B) ....................... ........................ .................. 35 FIGURE 2.28 CASE 6 RESULTS IN 0PENDSS (A) ................................................................. 36 FIGURE 2.29 CASE 6 R ESULTS IN 0PENDSS (B) ................................................................. 37 FIGURE 2.30 CASE 6 RESULTS IN 0PENDSS (c) ................................................................. 37 FIGURE 2.31CASE6 RESULTS IN 0PENDSS (D) ................................................................. 37 FIGURE 2.32 RESULT FOR CASE 1 - LIMIT 1 ...................................................................... 38 FIGURE 2.33 RESULT FOR CASE 1-LIMIT2 ...................................................................... 38 FIGURE 2 .34 R ESULT FOR CASE 1 - LIMIT 3 ... ... .................... .......................... .................. 39 vi USC Viterbi School of Engineering PhD Thesis FIGURE 2.35 RESULT FOR CASE 1 - C UMULATIVE INCREASE ............................................ 39 FIGURE 2.36 RESULT FOR CASE l IN 0PENDSS ................................................................ 40 FIGURE 2.37 PROJECTED TOTAL USC LOAD FOR FOUR DAYS ............................................. 41 FIGURE 2.38 CASE 2 RESULTS IN 0PENDSS ....................................................................... 41 FIGURE 2.39 CASE 2 RESULTS IN OPENDSS ....................................................................... 42 FIGURE 2.40 CASE 3 RESULTS IN 0PENDSS (A) ................................................................. 43 FIGURE 2.41CASE3 RESULTS IN 0PENDSS (B) ................................................................. 44 FIGURE 2.42 CASE 3 RESULTS IN 0PENDSS (c) ................................................................. 45 FIGURE 2.43 RESULT FOR 88-22 CASE 1 ........................................................................... 46 FIGURE 2.44 RESULT FOR 88-22 CASE 2 ......... .................................................................. 46 FIGURE 2.45 RESULT FOR 88-22 CASE 3 ........................................................................... 46 FIGURE 2.46 R ESULT FOR 88-23 CASE 1 ........................................................................... 52 FIGURE 2.47 RESULT FOR 88-23 CASE 2 ........................................................................... 52 FIGURE 2.48 RESULT FOR 88-23 CASE 3 ........................................................................... 53 FIGURE 3.1 DAILY PV OUTPUT CURVE: ACTUAL (UPPER) AND SIMPLIFIED (LOWER) ......... 73 FIGURE 3. 2 LOCATION FOR BATTERY STORAGE ATTACHMENT ........................................... 79 FIGURE 3.3 BATTERYMODEL ............................................................................................. 80 FIGURE 3.4 BATTERY STORAGE MODEL PARAMETERS ....................................................... 80 FIGURE 3.5 PI CONTROLLER .............................................................................................. 81 FIGURE 3.6 SPWM TECHNIQUE TO GENERATE SWITCHING SIGNAL .................................. 82 FIGURE 3.7 THREE PHASE IGBTBRIDGE ........................................................................... 83 FIGURE 3.8 BESTV2G CHARGING STRATEGY PROPOSED FOR LADWP ............................ 84 FIGURE 3. 9 SMART CHARGING IMPACT ON DAILY LOAD CURVE ........................................ 85 FIGURE 3 .10 24 HOUR CHARGING LOAD CURVE WITH AND WITHOUT V2G FOR IEEE 34 .. 86 FIGURE 3 .11 HOUR CHARGING LOAD C URVE WITH AND WITHOUT V2G FOR CW88-22 ..... 86 FIGURE 3.12 24 HOUR CHARGING LOAD CURVE WITH AND WITHOUT V2G FOR CW88-23 86 FIGURE 3 .13 24 HOUR CHARGING LOAD CURVE WITH V2G AND PV FOR CW88-22 ......... 87 FIGURE 3 .14 RESULTING PLOT IN MATLAB CASE 1 ......................................................... 87 FIGURE 3 .15 RESULTING PLOT IN MATLAB CASE 2 ......................................................... 88 FIGURE 3.16 RESULTING PLOT IN MATLAB CASE 3 ......................................................... 88 FIGURE 3.17 ECONOMIC SAVINGS VERSUS THE N UMBER OF ELECTRIC VEHICLES ............. 89 FIGURE 3.18 ECONOMIC SAVINGS V ERSUS THE CAPACITY OF THE POWER SYSTEM ........... 90 FIGURE 4.1 LAD WP JFB PARKING LEVEL 3 CHARGING STATION ARRANGEMENT ............ 97 FIGURE 4.2 LOCATION OF UCLA CHARGING STATIONS ..................................................... 99 FIGURE 4.3 SCHEMATIC OF UCLA ................................................................................... 100 FIGURE 4.4 LAD WP JFB PARKING LEVEL 3 CHARGING STATION ARRANGEMENT .......... 104 FIGURE 4. 5 ENERGY DELIVERY PER CHARGING EVENT OF LAD WP ASSIGNED EVs FROM MAY2015TOMAY2016 ........................................................................................... 105 FIGURE 4.6 HISTOGRAM OF CHARGING PLUG-IN TIME OF LAD WP ASSIGNED EVs FROM MAY 2015 -MAY 2016 ...... ........................................................................................ 106 vii USC Viterbi School of Engineering PhD Thesis FIGURE 4. 7 ENERGY DELIVERY PER CHARGING EVENT IN UCLA FROM JUNE 2014 TO AUGUST2015 ........................................................................................................... 107 FIGURE 4.8 HISTOGRAM OF CHARGING PLUG-IN TIME IN UCLA FROM JUNE 2014 TO AUGUST2015 ........................................................................................................... 108 FIGURE 4. 9 ENERGY DELIVERY PER CHARGING EVENT OF LAD WP RIDE SHARE FROM MAY 2015 TO MAY 2016 ................................................................................................... 108 FIGURE 4.10 HISTOGRAM OF CHARGING PLUG-IN TIME OF LAD WP RIDE SHARE FROM MAY2015TOMAY2016 ........................................................................................... 109 FIGURE 4.11 EXAMPLE OF 24-HR CHARGING ENERGY DISTRIBUTION FOR ONE EV .......... 110 FIGURE 4.12 EXAMPLE OF ORIGINAL CHARGING RECORD FOR ONE EV ............................ 111 FIGURE 4.13 EXAMPLE OF 24-HR CHARGING ENERGY DISTRIBUTION PLOT FOR ONE EV USING D ESIGNED ALGORITHM ................................................................................... 112 FIGURE 4.14 HOURLY CHARGING DISTRIBUTION OF LAD WP 36 ASSIGNED EVs, MAY 2015 -MAY 2016 ................................................................................................................ 112 FIGURE 4.15 COMPARISON OF AVERAGED WEEKDAY AND WEEKEND 24 HOUR ENERGY DISTRIBUTION OF LADWP 36 AsSIGNEDEVs FROMMAY2015 TO MAY 2016 ......... 113 FIGURE 4.16 S EASONAL COMPARISON OF 24 HOUR CUMULA TIVE CHARGING ENERGY DISTRIBUTION OF LAD WP 36 ASSIGNED EVs FROM MAY 2015 TO MAY 2016 ......... 114 FIGURE 4.17 ONE YEAR CUMULATIVE 24 HOUR ENERGY DISTRIBUTION OF LAD WP 36 ASSIGNED EVs FROM MAY 2015 TO MAY 2016 ......................................................... 114 FIGURE 4.18 8760 HOURLY CHARGING DISTRIBUTION IN UCLA FROM JULY 2014 TO JUNE 2015 ........................................................................................................................... 116 FIGURE 4.19 COMPARISON OF AVERAGED WEEKDAY AND WEEKEND 24 HOUR ENERGY DISTRIBUTION IN UCLA FROM JULY 2014 TO JUNE 2015 ........................................... 117 FIGURE 4.20 COMPARISON OF WINTER 2014 AND SUMMER 2015's CUMULATIVE CHARGING E NERGY DISTRIBUTION IN UCLA .............................................................................. 118 FIGURE 4.21 ONE YEAR CUMULATIVE 24 HOUR E NERGY DISTRIBUTION IN UCLA FROM JULY 2014 TO JUNE 2015 ............................................................................................ 118 FIGURE 4.22 HOURLY CHARGING DISTRIBUTION IN LAD WP RESIDENTIAL AREA, FEB - APRIL2016 ................................................................................................................ 119 FIGURE 4.23 COMPARISON OF NORMALIZED WEEKDAY AND WEEKEND 24 HOUR ENERGY DISTRIBUTION IN LADWPRESIDENTIALAREA, F EBRUARY TOAPRIL 2016 ............. 120 FIGURE 4.24 MONTHLY COMPARISON OF 24 HOUR CUMULATIVE CHARGING ENERGY DISTRIBUTION IN LADWPRESIDENTIALAREA, FEBRUARYTOAPRIL2016 ............. 121 FIGURE 4.25 CUMULATIVE 24 HOUR ENERGY DISTRIBUTION IN LAD WP RESIDENTIAL AREA, FEBRUARYTOAPRIL2016 ............................................................................. 121 FIGURE 4.26 8808 HOURLY CHARGING DISTRIBUTION OF LAD WP RIDE SHARE (30 EVs) MAY 2015 TO MAY 2016 ........................................................................................... 122 FIGURE 4.27 AVERAGED WEEKDAY 24 HOUR ENERGY DISTRIBUTION OF LAD WP RIDE SHARE (30 EVs) MAY 2015 TO MAY 2016 ................................................................ 123 viii USC Viterbi School of Engineering PhD Thesis FIGURE 4.28 SEASONAL COMPARISON OF 24 HOUR CUMULATIVE CHARGING ENERGY DISTRIBUTION OF LADWPRIDE SHARE (30 EVs) MAY 2015 TOMAY2016 ............ 124 FIGURE 4.29 CUMULATIVE 24 HOUR ENERGY DISTRIBUTION OF LAD WP RIDE SHARE (30 EVs) M AY 2015 TO MAY 2016 .................................................................................. 124 FIGURE 4.30 DAILY LOAD VARIATION IN CALIFORNIA BY CAISO ................................... 125 FIGURE 4.31 PLUG-IN ELECTRIC VEHICLE REGISTRATIONS PER THOUSAND PEOPLE BY STATE, 2014 .............................................................................................................. 126 FIGURE 4.32 DAILY LOAD VARIATION PREDICTION BASED ON LARGE CORPORATE FLEET DATA ......................................................................................................................... 127 FIGURE 4.33 DAILY LOAD VARIATION PREDICTION BASED ON TYPICALEVDATA ............ 127 FIGURE 5.1 24-HOUR DISTRIBUTION OF THE TRIP START TIME ........................................ 132 FIGURE 5.2 DISTRIBUTION OF TRIP DURATION ................................................................. 133 FIGURE 5.3 DISTRIBUTION OF TRIP DURATION ................................................................. 134 FIGURE 5. 4 PIE CHART OF TRIP NUMBER IN 24 HOURS .................................................... 134 FIGURE 5.5 TRIP MILEAGE DISTRIBUTION ........................................................................ 135 FIGURE 5. 6 NUMBER OF TRIPS FOR EACH USER ............................................................... 136 FIGURE 5.7 AVERAGE MILES PER TRIP FOR EACH USER ................................................... 137 FIGURE 5.8 AVERAGE MILES/VEHICLE IN THE WEEKS DURING MAY 2015 -MAY 2016 ..... 145 FIGURE 5. 9 24-HOUR DISTRIBUTION OF THE TRIP START TIME OF ASSIGNED BEV .......... 14 7 FIGURE 5.10 THE DISTRIBUTION OF TRIP MILEAGES OF ASSIGNED BEV ......................... 147 FIGURE 5 .11 THE PIE CHART OF TRIP MILEAGES OF ASSIGNED BEV ............................... 148 FIGURE 5.12 DISTRIBUTION OF ELECTRICITY CONSUMED (KWH) OF ASSIGNED BEVs .... 148 FIGURE 5.13 DISTRIBUTION OF START SOC OF ASSIGNED BEVs ..................................... 149 FIGURE 5.14 DISTRIBUTION OF END SOC OF ASSIGNED BEVs ........................................ 149 FIGURE 5.15 24-HOUR DISTRIBUTION OF THE TRIP START TIME OF ASSIGNED PHEVs ... 150 FIGURE 5.16 THE DISTRIBUTION OF TRIP MILEAGES OF ASSIGNED PHEV. ...................... 151 FIGURE 5 .17 THE PIE CHART OF TRIP MILEAGES OF ASSIGNED PHEV ............................ 151 FIGURE 5.18 THE DISTRIBUTION OF ELECTRICITY CONSUMED (KWH) OF ASSIGNED PHEVs .................................................................................................................................. 152 FIGURE 5.19 THE DISTRIBUTION OF START SOC OF ASSIGNED PHEVs ........................... 153 FIGURE 5.20 THE DISTRIBUTION OF END SOC OF ASSIGNED PHEVs .............................. 153 FIGURE 5.21 T HE DISTRIBUTION OF MILEAGE/KWH FOR BEV AND PHEV ..................... 154 FIGURE 5.22 24-HOUR DISTRIBUTION OF THE TRIP START TIME OF CARPOOL USERS ...... 155 FIGURE 5.23 THE DISTRIBUTION OF TRIP MILEAGES OF CARPOOLBEVs ........................ 156 FIGURE 5.24 THE PIE CHART OF TRIP MILEAGES OF CARPOOLBEVs .............................. 156 FIGURE 5.25 THE DISTRIBUTION OF ELECTRICITY CONSUMED (KWH) OF CARPOOLBEVs .................................................................................................................................. 157 FIGURE 5.26 THE DISTRIBUTION OF START SOC OF CARPOOLBEVs .............................. 157 FIGURE 5.27 THE DISTRIBUTION OF END SOC OF CARPOOL BEVs ................................. 158 FIGURE 5.28 AVERAGE CHARGING ENERGY PER DAY OF 10 USERS IN UCLA FROM 07/2014 ix USC Viterbi School of Engineering PhD Thesis - 08/2015 .................................................................................................................. 162 FIGURE 5.29 AVERAGE PLUG-IN TIME OF 10 USERS IN UCLA FROM 07/2014 -08/2015 .. 163 FIGURE 5.30 THE DISTRIBUTION MAP OF CHARGING METER .......................................... 164 FIGURE 5.31 NUMBER OF CHARGING COMPARISON OF TOP 10 TYPICAL OWNERS IN THE CITY .......................................................................................................................... 165 FIGURE 5.32 CHARGED ENERGY (KWH) COMPARISON OF TOP 10 TYPICAL OWNERS IN THE CITY .......................................................................................................................... 166 FIGURE 5.33 AVERAGE CHARGING TIME COMPARISON OF TOP 10 TYPICAL OWNERS IN THE CITY .......................................................................................................................... 166 FIGURE 5.34 AVERAGE CHARGED ENERGY COMPARISON OF TOP 10 TYPICAL OWNERS IN THE CITY ................................................................................................................... 166 FIGURE 6.1 HISTORICAL EV CHARGING D EMAND COMPARISON OF LAD WP 36 ASSIGNED EVs, LEVEL 2 CHARGING, 05/20/15 - 02/04/17 ......................................................... 174 FIGURE 6.2 GRAPH THAT ILLUSTRATE BOOTSTRAP TECHNIQUE ....................................... 180 FIGURE 6. 3 COMPLETE ALGORITHM OF EV CHARGING DEMAND ESTIMATION AND FORECAST ................................................................................................................. 181 FIGURE 6.4 A LGORITHM TO CHECK IF THE N EWLY G ENERATED CHARGING EVENTS HAS OVERLAP WITH THE PREVIOUS ONE WHEN CE>= 2 ................................................. 183 FIGURE 6.5 ALGORITHM TO CHECK IF THE NEWLY GENERATED CHARGING EVENTS HAS OVERLAP WITH THE PREVIOUS ONE WHEN CE > = 3 ................................................. 183 FIGURE 6.6 ALGORITHM TO CHECK IF THE NEWLY GENERATED CHARGING EVENTS HAS OVERLAP WITH THE PREVIOUS ONE WHEN CE= 4 .................................................... 184 FIGURE 6. 7 CHARGING EVENT DISTRIBUTION OF INDUSTRIAL COMPACT EV S, 05/20/15 - 02/04/17 ................................................................................................................... 186 FIGURE 6.8 CHARGING START TIME DISTRIBUTION OF INDUSTRIAL COMPACT EVs, 05/20/15 - 02/04/17 .................................................................................................. 187 FIGURE 6. 9 CST CURVE FITTING ON TRAINING S ET FOR INDUSTRIAL C OMPACT EVs ...... 187 FIGURE 6.10 CST VALIDATION ON TESTING SET FOR INDUSTRIAL COMPACT EVs ........... 188 FIGURE 6.11 PLUG-IN DURA TION DISTRIBUTION OF INDUSTRIAL COMPACT EVs, 05/20/15 - 02/04/17 ................................................................................................................... 188 FIGURE 6.12 PLUG-IN DURATION C URVE FITTING ON TRAINING SET FOR INDUSTRIAL CoMPACTEVs .......................................................................................................... 189 FIGURE 6.13 PLUG-IN DURATION VALIDATION ON TESTING SET FOR INDUSTRIAL COMPACT EVs .......................................................................................................................... 189 FIGURE 6.14 ssoc DISTRIBUTION OF INDUSTRIAL CoMPACTEVs, 05/20/15 -02/04/17 190 FIGURE 6.15 SSOC CURVE FITTING ON TRAINING SET FOR INDUSTRIAL COMPACT EVs. 190 FIGURE 6.16 SSOC VALIDATION ON TESTING SET FOR INDUSTRIAL COMPACTEVs ........ 191 FIGURE 6.17 GENERATED CST, PLUG-IN DURATION, CHARGING DURATION AND SSOC RANDOM VARIABLES FOR INDUSTRIAL COMPACT EV ............................................... 192 FIGURE 6.18 MONTE CARLO SIMULATION RESULT FOR INDUSTRIAL COMPACT EV DEMAND x USC Viterbi School of Engineering PhD Thesis MODEL ...................................................................................................................... 192 FIGURE 6.19 MONTE CARLO SIMULATION R ESULT FOR INDUSTRIAL COMPACT EV D EMAND MODEL - SURF PLOT IN MATLAB ............................................................................ 193 FIGURE 6.20 MONTE C ARLO SIMULATION RESULT FOR INDUSTRIAL COMPACT EV DEMAND MODEL - SURF WITH SPLINE PLOT IN MATLAB ....................................................... 193 FIGURE 6.21 G ENERA TED CST, P LUG-IN D URATION, CHARGING D URATION AND SSQC RANDOM VARIABLES FOR 10,000 INDUSTRIALEVs .................................................. 195 FIGURE 6.22 MONTE C ARLO SIMULATION RESUL T FOR INDUSTRIAL EV DEMAND MODEL (10,000 EVs) ............................................................................................................ 196 FIGURE 6.23 MONTE CARLO SIMULATION RESULT FOR INDUSTRIAL EV DEMAND MODEL- SURF PLOT IN MATLAB (10,000 EVs) .................................................................... 196 F IGURE 6.24 MONTE CARLO SIMULATION R ESULT FOR INDUSTRIAL EV D EMAND M ODEL - SURF WITH SPLINE PLOT IN MATLAB (10,000 EVs) ............................................... 197 FIGURE 6.25 HISTORICAL EV CHARGING D EMAND COMPARISON OF LAD WP 30 C ARPOOL EVs, LEVEL 2 CHARGING, 05/20/15 - 02/04/17 ........................................................ 199 FIGURE 6.26 GENERATED CST, PLUG-IN DURATION, CHARGING DURATION AND SSQC RANDOM VARIABLES FOR 10,000 C OMMERCIALEVs ............................................... 202 FIGURE 6.27 MONTE CARLO SIMULATION RESULT FOR COMMERCIAL EV DEMAND MODEL (10,000 EVs) ............................................................................................................ 202 FIGURE 6.28 MONTE CARLO SIMULATION RESULT FOR INDUSTRIAL EV DEMAND MODEL- SURF PLOT IN MATLAB (10,000 EVs) .................................................................... 203 FIGURE 6.29 MONTE CARLO SIMULATION RESULT FOR INDUSTRIAL EV DEMAND MODEL- SURF WITH SPLINE PLOT IN MATLAB (10,000 EVs) ............................................... 203 FIGURE 6.30 G ENERATED CST, C HARGING D URATION AND SSQC RANDOM VARIABLES FOR 10,000 RESIDENTIALEVs ......................................................................................... 206 FIGURE 6.31 MONTE C ARLO SIMULATION RESULT FOR RESIDENTIAL EV DEMAND MODEL (10,000 EVs ) ............................................................................................................ 207 FIGURE 6. 3 2 MONTE CARLO SIMULATION RESULT FOR RESIDENTIAL EV DEMAND MODEL - SURF PLOT IN MATLAB (10,000 EVs) .................................................................... 207 FIGURE 6.33 MONTE CARLO SIMULATION RESULT FOR RESIDENTIAL EV DEMAND MODEL- SURF WITH SPLINE PLOT IN MATLAB (10,000 EVs) ............................................... 208 xi USC Viterbi School of Engineering PhD Thesis List of Tables TABLE 1.1 EV PROJECTION IN UK IN [ 4] .......... .................................................................... 3 TABLE 1.2 CHARGING POWER LEVEL. .................................................................................. 3 TABLE 1.3 ANNUALEVDEMANDINUKIN [4] .................................................................... 4 TABLE 2.1 EDD POWER FLOW RESULTS OF IEEE 34-BUS TEST FEEDER ........................... 15 TABLE 2.2 POWER FLOW RESULTS OF USC MICRO-GRID ................................................... 17 TABLE 2.3 POWER FLOW RESULTS OF CHATSWORTH (DS 88) DISTRIBUTION SYSTEM ....... 18 TABLE 2.4 POWER FLOW RESULTS OF CHATSWORTH CIRCUIT# 22 DISTRIBUTION SYSTEM 19 TABLE 2.5 POWER FLOW RESULTS OF CHATSWORTH CIRCUIT# 23 DISTRIBUTION SYSTEM 19 TABLE 2.6 MONITORED BUSES VOLTAGES ....... .................................................................. 28 TABLE 2.7 CASE 1 RESULT IN 0PENDSS ............................................................................ 29 TABLE 2.8 CASE 2 RESULTIN OPENDSS .......... .................................................................. 31 TABLE 2.9 CASE 3 RESULT IN 0PENDSS ............................................................................ 32 TABLE 2.10 CASE 4 RESULT IN 0PENDSS .......................................................................... 34 TABLE 2.11CASE5 RESULT IN 0PENDSS .......................................................................... 35 TABLE 2.12 CASE 6 R ESULT IN 0PENDSS .......................................................................... 36 TABLE 2.13 SOUTHERN CALIFORNIA EDISON ELECTRICITY RATE ...................................... 42 TABLE 2.14 UTILIZATION FACTORS FOR CHATSWORTH 88-22 UNDER DIFFERENT LOADING CONDITIONS ................................................................................................................ 47 TABLE 2.15 CAPACITY FACTORS FOR CHATSWORTH 88-22 UNDER PEAK LOADING CONDITION 100% ....................................................................................................... 48 TABLE 2.16 CAPACITY FACTORS FOR CHATSWORTH 88-22 UNDER INTERMEDIATE LOADING CONDITION 55% ......................................................................................................... 49 TABLE 2.17 CAPACITY FACTORS FOR CHATSWORTH 88-22 UNDER OFF-PEAK LOADING CONDITION 30% ......................................................................................................... 50 TABLE 2.18 POWER FACTORS FOR CHATSWORTH 88-22 UNDER PEAK LOADING CONDITION (100%) ........................................................................................................................ 51 TABLE 2.19 POWER FACTORS FOR CHATSWORTH 88-22 UNDER INTERMEDIATE LOADING CONDITION (55o/o) ....................................................................................................... 51 TABLE 2.20 POWER FACTORS FOR CHATSWORTH 88-22 UNDER OFF-PEAK LOADING CONDITION (30%) ....................................................................................................... 51 TABLE 2.21 POWER QUALITY FOR CHATSWORTH 88-22 UNDER DIFFERENT LOADING CONDITIONS ................................................................................................................ 51 TABLE 2.22 UTILIZATION FACTORS FOR CHATSWORTH 88-23 UNDER DIFFERENT LOADING CONDITION ................................................................................................................. 53 TABLE 2.23 CAPACITY FACTORS FOR CHATSWORTH 88-23 UNDER PEAK LOADING CONDITION (100%) ..................................................................................................... 54 TABLE 2.24 CAPACITY FACTORS FOR CHATSWORTH 88-23 UNDER INTERMEDIATE LOADING CONDITION (55%) ......... .. ... ................... ... ... .................... ... ....................... .................. 55 xii USC Viterbi School of Engineering PhD Thesis TABLE 2.25 CAPACITY FACTORS FOR CHATSWORTH 88-23 UNDER OFF-PEAK LOADING CONDITION (30o/o) ....................................................................................................... 56 TABLE 2.26 POWER FACTORS FOR CHATSWORTH 88-23 UNDER PEAK LOADING CONDITION (100%) ........................................................................................................................ 57 TABLE 2.27 POWER FACTORS FOR CHATSWORTH 88-23 UNDER INTERMEDIATE LOADING CONDITION (55%) ....................................................................................................... 57 TABLE 2.28 POWER FACTORS FOR CHATSWORTH 88-23 UNDER OFF-PEAK LOADING CONDITION (30%) ....................................................................................................... 57 TABLE 2.29 POWER QUALITY FOR CHATSWORTH 88-23 UNDER DIFFERENT LOADING CONDITIONS ................................................................................................................ 57 TABLE 2.30 S UMMARY OF THE DIFFERENT CASES CHARGING AND DISCHARGING COSTS .. 60 TABLE 3 .1 PV POWER ADDED AT EACH LOAD (PERIOD 3) ................................................. 7 4 TABLE 3.2 PVPOWERADDED AT EACH LOAD (PERIOD 2) ................................................. 75 TABLE 3 .3 PV POWER ADDED AT EACH LOAD (PERIOD 1) ................................................. 76 TABLE 3.4 VARIABLES IN MATLAB MODELING ................................................................ 77 TABLE 3 .5 ECONOMIC SAVINGS FOR FIXED SYSTEM PEAK POWER IN MATLAB ............... 89 TABLE 3 .6 ECONOMIC SAVINGS FOR FIXED N UMBER OF EVs IN MATLAB ....................... 90 TABLE 3.7 CHANGES OF P & Q AT SELECTED BUS WHEN EVs ARE CONNECTED ................. 91 TABLE 3.8 VARIATION TREND IN MATLAB ....................................................................... 91 TABLE 3.9 CHARGING STARTING TIME MAXIMUM HARMONICS AT SELECTED BUSES ........ 91 TABLE 3 .10 CHARGING STEADY STATE AVERAGE HARMONICS AT SELECTED BUSES .......... 92 TABLE 4.1 EVMODELAND BATTERY SIZE IN LADWP ...................................................... 96 TABLE 4.2 EV MODEL AND BATTERY SIZE IN UCLA ......................................................... 98 TABLE 4.3 EVREBATE METER INFORMATION .................................................................. 101 TABLE 4.4 EVMODELAND BATTERY SIZE IN LADWP .................................................... 103 TABLE 5 .1 TRIP NUMBER AND THE AVERAGE MILEAGE FOR ONE TRIP OF E ACH EV USER 136 TABLE 5.2 EV JOURNEYS PER V EHICLE ............................................................................ 138 TABLE 5.3 EV JOURNEYS BY BRAND ............................................................................... 140 TABLE 5.4 EV JOURNEYS BY TYPE ................................................................................... 141 TABLE 5.5 EVUSAGE ON WEEKDAYS .............. ................................................................ 141 TABLE 5.6 EVUSAGE ON WEEKENDS .............................................................................. 143 TABLE 5.7 AVERAGE EVUSAGE ON W EEKDAYS & W EEKENDS ....................................... 144 TABLE 5.8 DETAIL OF ASSIGNED BEV ............................................................................. 146 TABLE 5.9 DET AIL OF ASSIGNEDPHEV ........................................................................... 150 TABLE 5.10 DETAILS OF LADWP CARPOOL BEV ........................................................... 155 TABLE 5.11 THE COMPARISON OF ICE ANDBEVUSERS .................................................. 159 TABLE 5.12 THE COMPARISON BETWEEN ASSIGNED BEV AND ASSIGNED PHEV ............ 160 TABLE 5.13 THE COMPARISON BETWEEN ASSIGNED BEV AND CARPOOLBEV ............... 161 TABLE 5.14 TOP 10 H EAVYUSERBEHAVIOR COMPARISON OF TYPICAL OWNERS IN THE CITY ............ ....................... ............................................... ........................................ 165 xiii USC Viterbi School of Engineering PhD Thesis TABLE 5 .15 INFORMATION OF TYPICAL OWNERS IN TIIE CITY .......................................... 170 TABLE 6.1 SUMMARY OF EV CHARGING DATAAND ANALYSIS OF CHAPTER 4 AND 5 ....... 172 TABLE 6.2 SUMMARY OF MODELS AND TECHNIQUES USED FOR DEMAND ESTIMATION AND FORECAST ................................................................................................................. 173 TABLE 6.3 DETAILED INFORMATION ON 36 LAD WP ASSIGNED EVs ............................... 175 TABLE 6.4 SUMMARY OF LAD WP ASSIGNED EVs UNDER DIFFERENT BODY CLASS ....... 175 TABLE 6.5 INDIVIDUAL EV CHARGING EVENT OF INDUSTRIAL COMPACT EVs, 05/20/15 - 02/04/17 ................................................................................................................... 185 TABLE 6.6 INDIVIDUAL EV CHARGING EVENT OF INDUSTRIAL COMPACT EVs, 05/20/15 - 02/04/17 ................................................................................................................... 186 TABLE 6.7 GOODNESS OF FIT INDICES SUMMARY OF INDUSTRIAL COMPACT EV MODEL 191 TABLE 6.8 10,000 INDUSTRIAL COMPACT EV CHARGING D EMAND ESTIMA TION AND FORECAST RESULT .................................................................................................... 194 TABLE 6.9 10,000 INDUSTRIALEVs CHARGING DEMAND ESTIMATION AND FORECAST RESULT ..................................................................................................................... 198 TABLE 6.10 DETAILED INFORMATION ON 30 LADWP CARPOOLEVs .............................. 200 TABLE 6.11 SUMMARY OF LAD WP CARPOOL EVs UNDER DIFFERENT BODY C LASS ...... 200 TABLE 6.1210,000 COMMERCIALEVs CHARGING DEMAND ESTIMATION AND FORECAST RESULT ..................................................................................................................... 204 TABLE 6.13 10,000 RESIDENTIAL EVs CHARGING DEMAND ESTIMATION AND FORECAST RESULT ..................................................................................................................... 209 xiv USC Viterbi School of Engineering ABBREVIATIONS AC AMI ARRA B2G B2V BEV BMS BS CAISO CE CES CST D DC DEW DG DOE DSM EDD EPRI EREV EV Alternating Current Advanced Metering Infrastructure American Recovery and Reinvestment Act Battery to Grid Battery to Vehicle Battery Electric Vehicle Battery Management System Battery Storage California Independent System Operator Charging Event Community Energy Storage Charging Start Time Duration I Plug-in Duration Direct Current Distribution Engineering Workstation Distributed Generation Department of Energy Demand Side Management Electrical Distribution Design Electric Power Research Institute Extended-Range Electric Vehicle Electric Vehicle xv PhD Thesis USC Viterbi School of Engineering PhD Thesis G2B Grid to Battery G2B Grid to Vehicle HEV Hybrid Electric Vehicle ICE Internal Combustion Engine IGBT Insulated Gate Bipolar Transistor JFB Jolm Ferraro Building JPL Jet Propulsion Lab LAD WP Los Angeles Department of Water and Power MAPE Mean Absolute Percentage Error MPG Miles per Gallon MPGeq Miles per Gallon Equivalent NEV Neighborhood Electric Vehicle PEV Plug-in Electric Vehicle PHEV Plug-in Hybrid Electric Vehicle PI Phase Inverter PDF Probability Density Function PS CAD Power System Computer Aided Design PV Photovoltaic Coefficient of determination RMSE Root Mean Square Error SG Smart Grid SGRDP Smart Grid Regional Demonstration Project soc State of Charge xvi USC Viterbi ' School of Engineering PhD Thesis ssoc Start State of Charge ST Start Time SUV Sport Utility Vehicle UCLA University of California at Los Angeles USC University of Southern California V2G Vehicle to Grid xvii USC Viterbi School of Engineering PhD Thesis ABSTRACT With the growing penetration of the Electric Vehicles (EV) to our daily transportation needs owing to their economic and environmental benefits, there will be both opportunities and challenges to the electric power utilities when adopting plug-in electric vehicles to the distribution network. The purpose of this thesis is to investigate the impact of electric vehicle battery charging on grid demand and steady state parameters of distribution network in the Los Angeles Department of Water and Power (LADWP) seivice territory. The research work was conducted under American Recovery and Reinvestment Act (ARRA) Smart Grid Regional Demonstration Project (Federal Grant Number DE-OE0000192) and funded by the US Department of Energy and LAD WP. In order to evaluate the impact of EV infrastructure charging on the distribution grid, three real-world distribution systems have been selected and tested via diverse EV penetration levels and system loading conditions. The designed cases have been tested and verified by two specialized power system analysis software: EDD and OpenDSS. Moreover, various positive and negative effects on the distribution network have been analyzed. The following part of this thesis examines the possibility of using EV s to provide support to the grid, instead of negative impact discussed in the first part. Specifically, distribution effects of battery aggregation and backfill coupled with renewable energy has been studied. Various EV charging/discharging models, based on the concept of vehicle-to-grid (V2G), are designed and tested on two distribution networks. An integrated algorithm that incorporate EV, PV and storage system is developed and verified to achieve the goal of shaving the peak and filling the valley. In the next two parts of this thesis, EV charging patterns, power usage of charging stations, and user behaviors have been analyzed based on one year real-world data. The unique data sources and analysis are important for LADWP and other similar utility companies. The results could be used to provide valuable information and practical insights for the utility companies. In the last part of this thesis, an EV charging demand estimation and forecast model based on Monte Carlo simulation technique has been developed. The model is based on historical charging record and the analysis mentioned above, and can be used to forecast future EV demand for the three customer classes: Industrial, Commercial and Residential. xviii USC Viterbi School of Engineering PhD Thesis ACKNOWLEDGEMENTS The research presented in this thesis is supported from the ARRA Smart Grid Regional Demonstration Project (Federal Grant Number DE-OE0000192) and funded by the US Department of Energy and LADWP. The development of the above mentioned project and this PhD thesis over the past few years has been one of the most challenging but unique and rewarding experience in my lifetime. I would like to thank all the colleagues from University of Southern California and express my sincere appreciation to all the persons that related to this research work: First and foremost, I would like to thank my PhD academic supervisor, Professor Mohammed J. Beshir, for his trust, patience, consistent help and support. I will always remember the valuable moments of our discussion and be thankful for the supervision under his guidance. I would like to use this opportunity to express my deepest gratitude and special thanks to Professor Edward Maby, for his patience and caring, kind words and sharing from the first stage of my research and academic work. I am very grateful to Professor Edmond Jonckheere, Professor Najmedin Meshkati and Professor Paul Bogdan for being my PhD Qualifying exam and Dissertation Committee members. Thank you for the help, support and valuable suggestions for the improvement of my thesis. Also, I would like to express my deepest thanks to Ali Mazloomzadeh, Daniel Kim, Ruslan Sibagatullin and other people I met through the Smart Grid Demonstration Project for their help in this research work. Last but most importantly, my best regards and deepest sense of gratitude to my friends and family for their understanding, patience and encouragement. It won't be possible without you guys. xix USC Viterbi School of Engineering PhD Thesis 1. CHAPTER 1: Introduction 1.1 Thesis Background This thesis is based on an actual project called Smart Grid Regional Demonstration Project (SGRDP). Detailed information of SGRDP and thesis background is given in this section. 1.1.1 Smart Grid Reginal Demonstration Project Original Scope Statement The Los Angeles SGRDP is a demonstration project intended to support the goal of the Department of Energy (DOE) Smart Grid (SG) Demonstration Funding Opportunity Announcement. The goal is to demonstrate SG and Energy Storage technologies in regions across the United States that embody essential and salient characteristics of each region and present a suite of use cases for national implementation and replication. From these use cases, the objective is to collect and provide the information necessary for customers, distributors, and generators to change their behavior in a way that reduces system demand and costs, increases energy efficiency and reliability, and optimally, allocates and matches demand and resources. The social benefits of a SG and energy storage technologies are to reduce emissions, lower costs, increase reliability, improve security, and enhance system flexibility so that new energy technologies, including renewable and distributed resources can be accommodated. To support these objectives, Los Angeles Department of Water and Power (LAD WP) and its research partners the University of Southern California (USC), the University of California at Los Angeles (UCLA) and Jet Propulsion Lab (JPL) will demonstrate innovations in key areas of Smart Grid (SG) technologies. The LADWP SGRDP consists of the four separate projects: (1) Demand Response/Advanced Metering Infrastructure (AMI), (2) Electric Vehicle (EV), (3) Customer Behavior Studies, and ( 4) Next Generation Cyber Security. The complete overview for each partner's responsibilities in SGRDP can be found in Figure 1.1. USC research team is mainly responsible for the EV Project, which is the background for this thesis. The EV Project demonstration work is divided into six categories: 1. Smart charging 2. Battery aggregation and backfill 1 USC Viterbi School of Engineering PhD Thesis 3. A fully operational Micro-grid with Substation and Distribution Automation 4. Renewables and EV battery integration 5. A car/ride share program at LADWP 6. Grid impact and Power Study -~----------------<:. UCLA Lab • Hp.Jj/ • Sman Cllareme • BattO<Y Aareeauon • Gnd Impact • Ronewob~ • Mocroend • Resldential · Commeroal • lnstrtuuonal · CB Studies • OR.vents • B2G (Bide. to Grid) USC Lab • Car/Ride Share • Sman Cllar&me • BattO<Y Aareiatoon • Gnd lmpact • Ronewob~ • Mocroend • Resldential • Commeroal • lnstlt\Jtlonal • CBStudaos • OR events • cs Systems - Deter Test • B2G (Bids. to Gnd) Field T ests • Chatsworth • Demonstration Sit~ • Rideshar~EV LAD WP Power L ab • Sman Chareone • Banery Aeereeatoon • Grod Impact LAD WP Computer Systems Figure 1.1 Responsibilities of Each Partner in SGRDP 1.1.2 Overview of EV and Charging Levels Today EVs bring important system solutions for the growing dependence from fossil fuels. EVs allow a considerable reduction of air pollution and energy issues [1]. Connection to the electric power grid provides opportunities such as ancillary services, reactive power support, tracking the output of renewable energy sources, and load balance. In the United States, an official domestic goal of putting one million EVs on the road by 2015 has been established [2], and public policies to encourage electrification have been implemented by governments at all levels. In UK, British government published a document [ 4] that estimate the required annual energy demand to cover EV charging under different penetration levels. EV uptake level in [ 4] is summarized in Table 1.1. 2 USC Viterbi School of Engineering PhD Thesis Table 1.1 EV Projection in UK in [4] Year 2020 2030 Scenario\ EV Type BEV (wiits) PHEV (Wlits) BEV (Wlits) PHEV (wiits) Extreme Range 2,600,000 500,000 5,800,000 14,800,000 High Range 1,200,000 350,000 3,300,000 7,900,000 Middle Range 600,000 200,000 1,600,000 2,500,000 Business as lJ sual 70,000 200,000 500,000 2,500,000 There are four types of electric cars that are generally considered in the EV discussions. They are: •Hybrid Electric Vehicle (HEV): Typical range: 500 miles. •Plug-in Hybrid Electric Vehicle (PHEV): Typical Range: 500-600 miles. •Battery Electric Vehicle (BEV): Typical Range: 80-120 miles. •Neighborhood Electric Vehicle (NEV): Typical Range: 20-40 miles •Extended-Range Electric Vehicle (EREV): Typical Range: 40-440 miles There are three standard charger types available in the US market today [57,58]. See Table 1.2. Though all of the charging types are available for application in the power grid, most of the studies conducted in this report assume Level 2 chargers as specified in the SGRDP design document [50]. When chargers other than Level 2 are used, it would be clearly identified in the report. Table 1.2 Charging Power Level Voltage Current Power Freq Phase Standard (VAC) (Amps) (kVA) (Hz) Outlet 120 12 1.44 208/240 32 6.7/7.7 480 400 192 60 60 60 single Single three NEMA5-15R SAE J1772/3 NIA Depending on the size of the vehicle battery, time for full-charge under Level 1 ranges 8- 14 hours, under Level 2 is 4-6 hours, and under Level 3 is 30 minutes-I hour. The chargers are either on or off the vehicle. Levels 1 and 2 are typically on-board charger, while Level 3 is typically off-board. Another charging consideration is conductive vs inductive charging. Conductive chargers use metal-to-metal contact as in most appliances and electronic devices. Inductive charging of EVs is based on magnetic contactless power transfer. 3 USC Viterbi School of Engineering PhD Thesis 1.1.3 Impact of EV Battery Charging on the Grid The EV market has grown rapidly over the past few years and a deeper penetration level for daily transportation can be seen in the near future. The increasing penetration rate of EVs in the U.S. is expected to bring great contribution to reduce greenhouse gases (GHG) and the need to use traditional fuels. It has been studied that the integration of EV charging on the grid may have significant potential impact on the grid [3]. From [4], the estimated EV charging demand for each level in UK can be found in Table 1.3. The authors of [ 4] concluded that the generating capacity in UK will be sufficient to meet the estimated EV charging demand, assume there will be ways to control EVs not to charge in the peak hours. However, this issue was not further discussed. Table 1.3 Annual EV Demand in UK in [4] Generating Capacity IOOGW 120GW Projected Annual UK Demand 360TWh 390 TWh (GWh) (GWh) 7,400 31,000 High Range 3,500 17,000 Middle Range 1,800 6,700 Business as Usual 400 4,200 In [5], the author used 6 U.S. regions' load demand data to find out the possibility of filling the valley by charging EV batteries. The results illustrated that EV charging could increase minimum load by 18% to 40%, depending on the region. Charging a large amount ofEVs simultaneously is equivalent to a type of nonlinear and large capacity load attached to the grid and may increase system losses, peak demand value, as well as deteriorating the power quality [ 6-8]. Authors in [9] presented a stochastic method to estimate the daily impact of EV charging on the distribution networks, and an evaluation of single phase EV charging impact on the residential grid was done in [10]. Deterministic and probabilistic approaches have been discussed in [ 11]. 4 USC Viterbi School of Engineering PhD Thesis 1.1.4 Battery Aggregation and Backfill, Vehicle-to-Grid Technology and Renewable Energy If EV scan be effectively integrated, they will also play a crucial role to reduce other system impacts and become great resources for smart grid infrastructure [27 ,28]. V2G technology, defined in [12], is a system in which there is capability of controllable, bi-directional electrical energy flow between a vehicle and the electrical grid. Traditionally, electricity flows from the grid to charge the vehicle in one direction. V2G allows the electricity flows in the reverse direction to provide power and feed the grid. The concept of V2G can also be viewed in Figure 1.2. V2G Unit Grid Electric Vehicle ~ A ~c __ .,... Bi-directional W ..,.._,,,. ;~t;r ia-- ZisghVoltage ! • ttery ~ 11111111111 111111·~·~ -o-i ............ 1111 .. Control U nit Control Unit Figure 1.2 V2G Concept [84] Study in [13] indicates the vehicles are not in use for transportation up to 95% of the time, and these time periods can be utilized for EVs to service the grid without compromising its primary transportation function [ 12]. PHEV' s charging/discharging characteristics have been discussed in [ 14], and the formulas of their statistics were derived. Various researches have been conducted to evaluate the distribution effect of battery aggregation and backfill with V2G concept. Large scale application of the V2G concept will bring potential impact on the grid. The authors in [ 15, 16] developed and tested different EV charging strategies and evaluated their impact on the local residential distribution grid. Then, they designed a battery charging/discharging schedule to achieve peak shaving, load variability reduction and an economic analysis. [17-23] treated PHEVs as responsible loads that are appropriate to be charged when the electricity demand is low, and also discussed the opportunity of participating into the Demand Side Management (DSM) to improve the load profile. 5 USC Viterbi School of Engineering PhD Thesis Photovoltaic (PV) is one important intermittent renewable resource, and it can be coupled with V2G technology to provide additional support and benefits to the generation system. A model was constructed and evaluated the environmental and economic impact of a combined PV and PHEV ecosystem [24]. The authors in [25] built a residential PV system connected to the grid to support PHEV charging. Another residential PV system that can support PHEV charging was considered in [26]. 1.1.5 Estimate EV Charging Demand Predicting the correct EV charging demand is important to understand the impact of EV infrastructure charging on the grid. [29-31] proposed several methods for EV load prediction. However, these methods rely heavily on the driving patterns. Real-world user habits were used to develop the driving patterns. Driving behaviors from Finland travel survey was used for the EV demand model [32]. The authors in [33,34] used Danish National Travel Smvey results to forecast EV charging demand. Driving patterns were the main component in [33], while average plug-in duration were studied and used in [34]. Studies in [35,36] used mathematical methodology to forecast EV demand, while researches in [37-43] chose to use other methods like Support Vector Machines and Monte Carlo technique for the demand estimation. EV spatial temporal distribution incmporating both transportation behaviors and EV battery's uncertainty characteristics was developed by using Monte Carlo simulation [42]. The authors in [43] proposed different prediction methods based on historical charging data. However, the sources of such data tend to violet personal privacy, and the data processing time is longer than expected due to the large data volume. Variables such as vehicle speed, the grade of road and acceleration were defined to estimate EV charging demand [ 44]. Other variables related to drive patterns were considered when constructing the model, they were travel distance, battery SOC and start charging time. The research work in [ 45] was based on the driving behaviors from national travel survey, and studies in [46-48] used real-world driving patterns. A utility grid used kernel density estimation method to define user travel patterns, and then constructed a model that can simulate temporal availability of charging EVs [49]. 1.2 Thesis Objectives Despite the growing interest in the benefits of electric vehicles and replacing them as the alternative of the traditional gas-powered vehicles in our daily life, not enough attentions have been paid yet to seriously examine the impact of electric vehicle's integration to the power grid. Previous studies suggested that there might be severe damage to the power system when large scales of electric vehicles have been deployed and charged at the same time during grid peak demand. 6 USC Viterbi School of Engineering PhD Thesis To better understand what impacts would actually bring to the grid by the integration of electric vehicles, research is needed on the latest electric vehicle technologies, customer behaviors and also from the utility's perspective. Therefore, the author propose to do an in-depth study to: • Evaluate the impact of EV integration to the grid; • Develop control strategy to reduce or mitigate such impacts by applying the vehicle to grid technology and coupled with renewables; • Analyze real-world EV charging patterns, stations usage and user behavior; and • Develop models for estimating and forecasting the charging-demand ofEVs Answering such questions requires the thorough analysis on how real-world distribution networks will behave and react when different EV penetration levels add on these systems; the development of an integrated solution that combines EV, renewable and other latest technology to mitigate the impact; statistical analysis on actual EV usage, charging patterns, station usage and also focus on the user's behavior. Finally, a model need to be developed that can be used to estimate and forecast future EV charging demand to help utility operators to plan the generation and operation profiles accordingly and benefit the utilities. The scope of the research is to demonstrate Smart Grid, Electric Vehicle and Energy Storage technologies in the LADWP service territory, evaluate the impacts of Electric Vehicle integration and develop EV charging demand forecast model that applicable not only to the LADWP area, but also transferrable for national implementation and replication. From the research results, the goal is to collect and provide the information necessary for customers, distributors, and generators to change their behavior in a way that reduces system demands and costs, increases energy efficiency, optimally allocates and matches demand and resources to meet that demand, and increases the reliability of the grid. In order to solve the thesis concerns, the following activities have been taken: 1. Literature review on current and historical EV charging and their impacts on the distribution grid to find out research gap 2. Select multiple distribution networks and develop various EV loading case scenarios to test how different EV penetration rate will influence the grid, and the worst impact of EV charging 3. Design different EV charging/discharging models and test on the distribution systems 7 USC Viterbi School of Engineering PhD Thesis 4. Develop an integrated algorithm that incorporate EV, PV and storage system to achieve optimum daily load profile 5. Collect and process real-world EV usage data, charging records and power consumption of charging stations 6. Analyze EV charging patterns and power usage for different type of charging stations 7. Categorize EV users into different group and study their behavior 8. Build a model to estimate and forecast future EV charging demand 1.3 Contributions to Knowledge In this thesis, several important contributions have been made to engineering field in regards to the development of the advanced techniques and knowledge: 1. Three real-world distribution systems have been selected and tested via diverse EV penetration levels and system loading conditions. The designed cases have been tested and verified by two power system analysis software: EDD and OpenDSS and the results turned out to be very similar to each other. The study results reveal the fact that the existing distribution infrastructure has sufficient capacity to carry a reasonable level of EV penetration. The results also suggest that distribution transformers are one of the key equipment that may need to be replaced to achieve higher EV penetration level. In addition, predicting the correct loading profile is important in understanding the infrastructure system limits. This contribution can be found in Chapter 2. 2. Various EV charging/discharging models, based on the concept of vehicle-to-grid (V2G), are designed and tested on two distribution networks. An integrated algorithm that incorporate EV, PV and storage system is developed and verified to achieve the goal of obtaining an optimum daily load profile, or, in other words, shave the peak and fill the valley. These charging/discharging models are presented in Chapter 3. 3. The actual usage of EV and charging stations, as well as the charging patterns have been recorded and analyzed for a time period of one year. This unique database and analysis is important for the utility companies. The results could be used to provide information and practical insights for LAD WP and other similar utility companies for resource planning and future installations of wired and wireless charging infrastructures for EV. The EV charging analysis results suggested the charging energy demand pattern could be much different. Detailed study can be found in Chapter4. 8 USC Viterbi School of Engineering PhD Thesis 4. It gives a detailed analysis of EV user behaviors based on one year real-world data from various EV users, while traditional similar analyses mostly used survey data as the only source. These users are categorized into three groups: large corporate EV users, EV users employed at a university, and typical EV owners in the city. From the data, charging habit and key trip information, such as trip start time, duration, mileage, and energy consumed have been studied. These information reveal different patterns of EV user behavior, and the results suggest the behavior varies in different groups. Details of the analysis is presented in Chapter 5. 5. An EV charging demand estimation and forecast model has been developed using Monte Carlo simulation technique. Convention models do not have EV user behavior information and their driving behaviors, and those models were mainly rely on traffic patterns, EV battery characteristics and charging characteristics, without using historical real-world charging record. The designed model can be used to estimate future EV demand for three customer classes: Industrial, Commercial and Residential. The model in this thesis defines four random variables based on historical 20-months' real-world data. The presented model could be used to evaluate the EV charging on the distribution grid, and may benefit LADWP or other similar utility operators to plan the generation and operation profiles accordingly in the future. The Monte Carlo model is presented in detail in Chapter 6. 1.4 Publications Arising from this Thesis Through the development of this thesis, the author has contributed to the following journal articles: Zeming Jiang, Laith Shalalfeh, Mohammed J. Beshir, "Impact of Electric Vehicles on the IEEE 34 Node Distribution Infrastructure", International Journal of Smart Grid and Clean Energy, vol. 3, no. 4, October 2014: pp. 417-424 The author has also contributed to the following conference papers as main author: Zeming Jiang, Hao Tian, M. J. Beshir, R. Sibagatullin and A. Mazloomzadeh, "Statistical Analysis of Electric Vehicles Charging, Station Usage and Impact on the Grid," 2016 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT), Minneapolis, MN, 2016, pp. 1-5. Z. Jiang, H. Tian, M. J. Beshir, S. Vohra and A. Mazloomzadeh, "Analysis of Electric Vehicle Charging Impact on the Electric Power Grid: Based on Smart Grid Regional Demonstration Project - Los Angeles," 2016 IEEE PES Transmission & Distribution Conference and Exposition-Latin America (PES T&D-LA), Morelia, 2016, pp. 1-5. 9 USC Viterbi School of Engineering PhD Thesis Z. Jiang, L. Shalalfeh and M. J. Beshir, "Impact of Electric Vehicle Infrastructure on the City of Chatsworth Distribution System," 2014 IEEE International Electric Vehicle Conference (IEVC), Florence, 2014, pp. 1-5. Z. Jiang, L. Shalalfeh and M. J. Beshir, "Impact of Electric Vehicle Infrastructure on the University of Southern California Micro-grid: Based on Smart Grid Regional Demonstration Project - Los Angeles," 2014 International Conference on Connected Vehicles and Expo (ICCVE), Vienna, 2014, pp. 7-11. 1.5 Thesis Structure This thesis is structured in the following way: Chapter 2 analyzes the impact of EV infrastructure charging on the grid with regards to system load flow, load factor and stability. Three distribution networks have been selected: IEEE 34 node distribution test system, USC micro-grid, and Chatsworth distribution system. These three systems are tested under diverse system conditions: various case scenarios and three load periods. Moreover, various positive and negative effects on the distribution network and micro-grid are analyzed, such as: power quality (over-voltage, voltage sages, switching surges, harmonics and noise), impact on distribution infrastructure (transformers, wires, cables, and capacitors), capacity factor, utilization factor, load factor, system reliability. Chapter 3 describes distribution effects of battery aggregation and backfill coupled with renewable energy. The concept of vehicle-to-grid (V2G) is first introduced and then applied into EV charging/discharging strategies to evaluate the distribution effects of battery aggregation and backfill, and the possibility of "shaving the peak and filling the valley". In addition, the designed V2G strategies are coupled with renewable energy and storage systems and become an EV, battery and PV integrated solution to achieve optimum load profiles. Chapter 4 presents EV charging patterns and power usage of three types of charging stations: large corporate, local residential and ride share program. EV charging data was collected either through meters or through a third party (FleetCarma). Analyses are made based on historical one year EV charging and usage records. Hourly, weekday vs weekend, seasonally and yearly energy distribution cmves are generated and compared to examine whether they have similar patterns. The last part of this chapter estimates the effect of EV loading on California network under different EV penetration rates. Chapter 5 provides an analysis on EV user behaviors. Customer profiles are analyzed and divided into three categories: large corporate EV users, EV users employed at a university, 10 USC Viterbi School of Engineering PhD Thesis and typical EV owners in the city. For each user group, top 10 heavy users are selected and their daily trips including mileage, trip start time and battery energy distribution are analyzed. Chapter 6 presents an EV charging demand estimation and forecast model. The historical charging records on 36 LADWP assigned EVs, and 30 LADWP carpool EVs have been chosen to construct the forecast model and represent for the Industrial and Commercial area. LAD WP residential EV rebate meter data incorporate with additional assumptions is used to estimate EV demand for the Residential area. Monte Carlo simulation, together with probability theory, statistics stochastic elements and Bootstrap confidence interval have been used and applied. The developed model could be used to evaluate the impact of EV charging on the distribution grid. The proposed model may benefit LADWP or other similar utility operators to plan the generation and operation profiles accordingly in the future. Chapter 7 concludes and summarizes the thesis. Learnings from each chapter are brought together. The above mentioned contribution to knowledge are justified and key thesis questions are properly answered. Suggestions and future works on the subject are given. 11 USC Viterbi School of Engineering PhD Thesis 2. CHAPTER 2: Impact of EV Infrastructure Charging on the Grid with regard to System Load Flow, Load Factor, and Stability 2.1 Research Overview This chapter will demonstrate the interaction of EV(s), battery storage, and grid during the charging I discharging process. Several methodologies will be looked at to manipulate the daily load profile on the grid to take full advantage of the EV(s) and battery storage. The fully functional Micro-grid that will be studied in this chapter will include the installed and operating EV chargers and battery storage systems connected to the electrical distribution system in Chatsworth. This Micro-grid or electrical distribution system will be comprised of building loads, lighting loads, co-generation (Thermal storage, Photovoltaic, or Gas Tmbine ), battery storage, and EV chargers. The study will analyze the various positive and negative effects on the Micro-grid such as: power quality issues (over-voltage, voltage sags, switching surges, harmonics and noise), impact on distribution infrastructure (transformers, wires, cables, and capacitors), capacity factor, utilization factor, load factor, system reliability, asset management, infrastructure optimization, cost effectiveness, and minimizing electric utility costs. The SGRDP design document [50] has anticipated Micro-grid system studies as well as a real-word Micro-grid application using the LADWP's Chatsworth system network. Due to the lack of system infrastructure at Chatsworth that is required for the application, data collection and analysis of a real-world Micro-grid system, this chapter addresses only the associated system studies conducted to address the above outlined Micro-grid issues at Chatsworth. However, various levels of Micro-grid modeling considerations are additionally tested in the studies to compensate for the lack of real-word Micro-grid application at Chatsworth. Micro-grid application using a standard IEEE test system, as well as, the USC university campus park network models are used in the studies. Distribution system analysis will focus on the following: 1. Utilization factor - the ratio of peak power demand to the maximum power that the system can accommodate. 2. Capacity factor - the ratio of peak power handling for a system component such as a transformer to the maximum power that the component can accommodate. 12 USC Viterbi School of Engineering PhD Thesis 3. Load factor - the ratio of actual load to peak load. 4. Power factor - the ratio of real useable power that can perform work to the magnitude of apparent power (current times voltage). 5. Power quality - an overall rating of voltage and current wave shapes, frequency regulation, and noise or harmonic distortion levels. These indices provide an assessment of overall distribution system reliability, asset management, infrastructure optimization, and cost effectiveness. The results will guide future systems development at LAD WP and other utilities. 2.2 Systems Modeling and Representation Utilized 2.2.1 Technologies and Systems Used To meet the general system modeling and representation requirements outline in the SGRDP design document [50], the following activities are under taken by the author: 1. Three sets of distribution network models are selected to enable analysis of EV distribution effects as described in the SGRDP Preliminary Design Document [50]. Furthermore, two software programs are selected for the development of these models. These software programs are: a. Distribution system modeling software Distribution Engineering Workstation (DEW) by Electrical Distribution Design, Inc. (EDD) is to be used to model the distribution network b. Power system distribution simulator software OpenDSS by Electric Power Research Institute (EPRI) is to be used to model the distribution networks 2. Three distribution network models are selected. They are: a. An IEEE 34 node test feeder reflecting an industry standard distribution test model [51,52] b. A USC distribution network or Micro-grid reflecting a large urban complex in the LADWP system c. A Chatsworth Distribution System or Micro-grid representing a residential/industrial area in the LAD WP system with a special focus to the circuit 22 and 23 3. The EDD models on the three system representations (IEEE 34 bus, USC, and Chatsworth) and OpenDSS models on the two system representations (IEEE 34 bus 13 USC Viterbi School of Engineering PhD Thesis and USC) allow the attachment of multiple EV-charger loads of vanous classifications 4. The EDD models on the three system representations (IEEE 34 bus, USC, and Chatsworth) and OpenDSS models on the two system representations (IEEE 34 bus and USC) are designed to reveal the aggregate effect of the multiple EV-charger loads at substation connections to the transmission grid 2.2.2 Configurations of the Smart Grid Systems, Subsystems and Components The network configurations of the test systems and the system modeling considerations are discussed in this subsection. The actual system studies test results are provided in Section 2.4. The three test system and their configurations are discussed below: 2.2.2.1 IEEE 34 Node Distribution Test System IEEE 34 node test feeder is a representation of an actual feeder located in Arizona. The system configuration for the IEEE 34 node test feeder is shown in Figure 2.1. Some of the key system parameters given in the standard IEEE model and they are compared with the EDD and OpenDSS representations are given in Table 2.1. The feeder's nominal voltage is 24.9 kV. It is characterized by: • Three phase 4-wire and single phase, 2-wire overhead lines arranged in different configurations. • Very long and lightly loaded. • Two in-line regulators required to maintain a good voltage profile. • An in-line transformer reducing the voltage to 4.16 kV for a short section of the feeder. • Unbalanced loading with both "spot" and "distributed" loads. Distributed loads are assumed to be connected at the center of the line segment. Loads are comprise of three-phase (balanced or unbalanced) and single phase systems. Three-phase loads are connected in Y or D while single-phase loads are connected line-to ground or line-to-line. All loads can be modeled as constant kW and kV AR (PQ), constant impedance (Z) or constant current (I) depending based on the actual system load. 14 USC Viterbi School of Engineering PhD Thesis 802 806 808 812 814 800 810 Figure 2.1 One-line Diagram of IEEE 34-bus Test Feeder Table 2.1 EDD Power Flow Results of IEEE 34-bus Test Feeder EDD Result OpenDSS Result Standard kW 2,043.17 2,032.47 2,042.872 Total Feeder kVAR 292.14 290.258 282.52 Flow kVA 2,063.95 2,052.01 2,063.389 kW 273.41 270.49 273.049 Total Losses kVAR 36.96 34.20 34.999 kVA 275.90 272.64 275.283 Ph-A 54.15 51.51 51.58 Bus Current Ph-B 46.81 44.20 44.57 Flow Ph-C 42.98 40.59 40.93 2.2.2.2 USC Micro-grid USC distribution system-park campus is the main campus of the university that is, located in Los Angeles, California. The USC distribution Micro-grid structure and its EDD model structure are shown in Figure 2.2. Some of the key system parameters in the EDD and OpenDSS representations are given in Table 2.2. As can be seen from Table 2.2, the EDD and OpenDSS representations are very similar and are producing similar results. 15 USC Viterbi School of Engineering PhD Thesis The feeder's nominal voltage is 4.8 kV. Loads are comprised of three-phase (balanced or unbalanced) and single phase systems. Three-phase loads are connected in wye or delta while single-phase loads are connected line-to-ground. All loads are modeled as constant kW and kV AR (PQ), constant impedance (Z) or constant current (I) depending the close approximation of the actual system load. Other system data include: F • 2 substations connect to the LADWP electric grid • 4 Feeder stations • 19 Feeder circuits • 5 parking structures: PSA, PSB, PSD L, PSD R, PSX TO ow iililf>WP #2 PSB SCI - ---oi:I ASB M.AalCABFPF L HER LTSCFH CFX RRI ORB ST light> HARMHP Figure 2.2 One-line Diagram of USC Micro-grid 16 PSX JKP LAW HOH USC Viterbi School of Engineering PhD Thesis Table 2.2 Power Flow Results of USC Micro-grid EDD Result OpenDSS Result kW 26,413.71 26,943.70 Total Feeder kVAR 14,633.21 15,991.60 Flow kVA 30,317.07 32,332.00 kW 89.02 97.83 kVAR 272.28 807.74 Total Losses kVA 292.22 813.64 2.2.2.3 Chatsworth Distribution System Chatsworth is a district of Los Angeles, California, United States; in the northwestern San Fernando Valley. The system configuration for the Chatsworth distribution system that is feed by Distribution System (DS) 88 is shown in Figure 2.3. Some of the key system parameters in the EDD modeling are given in Table 2.3. (No OpenDSS modeling is done for the Chatsworth system.) There are 26 feeder circuits in the Chatsworth distribution system and the nominal voltage is 4.8 kV. Loads are comprised of three-phase (balanced or unbalanced) and single phase. Three-phase loads are connected in wye or delta while single-phase loads are connected line-to-ground or line-to-line. AU loads can be modeled as constant kW and kV AR (PQ), constant impedance (Z) or constant current (I) depending based on the close approximation of actual system load. The system is feed by five lines coming from Northridge-Chatsworth at 34.5 kV. Two specific feeders, Circuits #22 and #23, are specified in the SGRDP design document [50] as the key circuits where many of the research work are to occur. The data associated with these circuits is shown in Table 2.3, Table 2.4 and Table 2.5 respectively. 17 USC Viterbi School of Engineering PhD Thesis Figure 2.3 One-line Diagram of the Chatsworth Distribution System Table 2.3 Power Flow Results of Chatsworth (DS 88) Distribution System EDD Result MW 67.75 Total Feeder MVAR -6.175 Flow MVA 69.411 MW 0.889 Total Losses MVAR 2.562 MVA 2.792 18 USC Viterbi School of Engineering PhD Thesis Table 2.4 Power Flow Results of Chatsworth Circuit# 22 Distribution System EDD Result kW 578.59 Total Feeder kVAR 87.51 Flow kVA 585.17 kW 6.77 Total Losses kVAR 5.56 kVA 8.76 Ph-A 66.91 Bus Current Ph-B 71.57 Flow Ph-C 70.23 Table 2.5 Power Flow Results of Chatsworth Circuit# 23 Distribution System EDD Result kW 865.83 Total Feeder kVAR 133 .26 Flow kVA 876.03 kW 11.49 Total Losses kVAR 10.59 kVA 15.63 Ph-A 107.44 Bus Current Ph-B 105.07 Flow Ph-C 99.92 2.3 Description of the Methodologies and Algorithms Utilized As described in Section 2.2, three system representations are provided with the EDD modeling. The methodologies used and the scenarios selected are to meet some of the key EV project objectives. These include: 1. The EDD models and the three system representations (IEEE 34 bus, USC, and Chatsworth) allow the attachment of multiple EV-charger loads of various classifications 19 USC Viterbi School of Engineering PhD Thesis 2. The EDD models and the three system representations (IEEE 34 bus, USC, and Chatsworth) are designed to reveal the aggregate effect of the multiple EV-charger loads at substation connections to the transmission grid 3. A sensitivity analysis are setup on the test systems with the goal of revealing the substations within the LADWP territory that have the greatest impact on the transmission grid when they are similarly subjected to EV loads. 2.3.1 Distribution System Load Level Consideration In order to define the load level to study, a consideration need to be made on the varying nature of the power system load. Due to the daily and season changes of the power system loads, certain simplifying assumptions need to be considered when analysis system issues that depend as well as impact the system loading and capabilities. A typical summer season load curve for the CAISO is shown Figure 2.4. In developing a simplifying consideration, the load curve is represented with three load levels or periods: Peak, Intermediate, and off-peak as shown in Figure 2.5. These load periods are defined as follows: • Period 1: Peak Load (100%), • Period 2: Intermediate Load (55%), and • Period 3: Off-peak Load (30%) The estimate in the percentage of the peak load for the Periods 2 and 3, consideration is made the impact of the seasonal changes to the summer peak load condition that is the basis for Period 1. G Cal ifor .!:)fil .L~.Q Cailomio Independent System Operator Ccwpcwation CAISO Load Curve July 24, 2006 5 ),085 IM"" / ...._ / r-.. ./ t' enr -i 1 ....... ~ / I\. I \ / \ II _/ ~ . . \ , ~ )' <: : 1 IUL L. I viii ''-'" I \ / ' - ~ j "- I vii\.. IU .:J / ~ 8,42 ~Ml v ......___ __.,.../ 30.CIDD == O:CO 1::D Z..C 4:0> s::x> «Al aco 9::?0 1Q40 t2:0) 1.32) 1.C40 1SCO t7";;Z 18;o:K) 32CO Zt::X'I z::AO Q:CO - A.c:cu;:ill..a.11d-1"2~- 1in10~ Figure 2.4 CAISO's Summer Load Curve as a Representative for Developing Load Periods 20 USC Viterbi School of Engineering 8 Ca l iforLJ~ .L~.Q PhD Thesis CalOOmia Independent System Opef'ator Corporation CAISO Load Curve July 24, 2006 II IL~ I I I I )Y.r;ad 2 .... ~ I I I / • Period 3 ---~;= 0 l 2 3 4 !i 6 I ~A.c:a..3t~.-1l 1~1St .li-1~10~t 18 19 20 21 22 23 24 Hours Figure 2.5 Load Periods Definition 2.3.2 EV Load Integration Scenarios Multiple EV load increase scenarios (Cases) are been analyzed using EDD and OpenDSS. The detail of all scenarios for IEEE 34, USC Micro-grid and Chatsworth distribution system can be found in this section. In order to keep this thesis short and enjoyable to be read, detailed analysis for each individual EV user will not be presented here, interested readers can contact DOE or LAD WP for the full deliverables. 2.3.2.1 Scenarios for IEEE 34 Node Distribution Test System Six EV load increase scenarios (Cases) are been analyzed using EDD and OpenDSS. These cases are defined as follows: • Case 1: Random EV Load Increase: o Increase each load (spot or distributed) until bus or system limit (transformer loading, voltage, and line limit) is reached • Case 2: 10% System Incremental increase Per Spot Bus o Increase 10% (or higher) circuit or system incremental starting from the far end of the circuit • Case 3: Distributed Incremental Increases o Increase loads throughout the system in percentage proportion to the load at the bus (spot load only) • Case 4: Random EV load Charging and Discharging Increase 21 USC Viterbi School of Engineering PhD Thesis o In addition to the charging at the spot bus, this case will have V2G loads at the spot bus to see the maximum kW that can be discharged to the system. • Case5: 10% Charging and Discharging Increment of system total load to each Spot Bus o Increase 10% charging and discharging of the system load starting from the far end of the circuit. 10% Load Continuous Increment in Order (840-860- 848-844-890-830). • Case 6: Distributed 10% incremental increase of the charging and discharging loads: 2.3.2.2 o Increase charging and discharging loads throughout the system m percentage proportion to the load at the bus (spot load only). Scenarios for USC Micro-grid Three EV load increase scenarios (Cases) are been analyzed using EDD and OpenDSS. These cases are defined as follows: • Case 1: Random EV Load Increase: o Increase each load (spot or distributed) until bus or system limit (transformer loading, voltage, and line limit) is reached • Case 2: 24-hour Charging Analysis: o Projecting all spot loads in USC Micro-grid over a 24-hour period • Case 3: 24-hour Charging/Discharging Analysis: 2.3.2.3 o At each hour, the maximum EV charging load will be assumed, and then the maximum possible discharging kW will be found Scenarios for Chatsworth Distribution System Three EV load increase scenarios (Cases) are been analyzed using EDD. These cases are defined as follows: • Case 1: Random EV Load Increase: o Increase each load (spot or distributed) until bus or system limit (transformer loading, voltage, and line limit) is reached • Case 2: 10% System Incremental increase Per Spot Bus o Increase 10% (or higher) circuit or system incremental starting from the far end of the circuit 22 USC Viterbi School of Engineering PhD Thesis • Case 3: Distributed Incremental Increases o Increase loads throughout the system in percentage proportion to the load at the bus (spot load only) 2.3.3 Load Flow, Load Factor and Stability Limits Load flow analysis using EDD program was conducted for each of the cases by incrementally increasing EV or other storage loads until system limits are reached. The limits for consideration typically are the equipment limits and other system based limits, such as voltages, load factors, and system stability. For distribution systems of the nature that are being studied, system stability issues are not a consideration. Thus, the only remaining issues are power and current flow, and voltage level consideration. Specifically, these are the limiting conditions that are monitored for the system studies. • Transformer Loading Limit: current rating of 100% and 225% • Voltage limit: 114-126V (normal); 110-127V (emergency) • Line Loading Limit: 100% The above criteria are used in determining the level of incremental EV or other storage devices that can be connected to the test systems. In addition, distribution system analysis will focus on the following: 1. Utilization factor - the ratio of peak power demand to the maximum power that the system can accommodate. 2. Capacity factor - the ratio of peak power handling for a system component such as a transformer to the maximum power that the component can accommodate. 3. Load factor - the ratio of actual load to peak load. 4. Power factor - the ratio of real useable power that can perform work to the magnitude of apparent power (current times voltage). 5. Power quality - an overall rating of voltage and current wave shapes, frequency regulation, and noise or harmonic distortion levels. 23 USC Viterbi School of Engineering PhD Thesis 2.4 Summaries of the Results of the Performance of the Smart Grid Systems and Technologies Derived from lab Tests, Field Tests, or Grid-connected Applications The study results related to the system studies are presented in this section. 2.4.1 IEEE 34 Node Distribution Test System 2.4.1.1 EDD Results for IEEE 34 In total, nine cases were run using EDD based on the three case scenarios and three load levels (periods). The study run cases are shown below: • CASE 1.1: Random Charging at period 1 ( 100% Load Condition) • CASE 1.2: Random Charging at period 2 (55% Load Condition) • CASE 1.3: Random Charging at period 3 (30% Load Condition) • CASE 2.1: 10% Load Continuous Increment in Order at period 1 • CASE 2.2: 10% Load Continuous Increment in Order at period 2 • CASE 2.3: 10% Load Continuous Increment in Order at period 3 • CASE 3 .1: 10% Increment Each Time of all the Load at period 1 • CASE 3.2: 15% Increment Each Time of all the Load at period 2 • CASE 3.3: 20% Increment Each Time of all the Load at period 3 These study results of these cases are summarized below, Figure 2.6 - Figure 2.14. 24 500 ~ 400 ~ .. 300 ~ ~ 200 100 0 USC Viterbi School of Engineering PhD Thesis Random Charging (Peak 100%) 840 860 848 844 89 0 830 Spot Load Bu.s Figure 2.6 Result for CASE 1.1 Random Charging (Intermediate 55%) 2000 .....-~~~~~~~~~~~~~~~~~~~~- 1800 -+-~~~~~~~~~~~~~~~~~~- 1600 +---===---- 1400 ! 1200 G:i 1000 ~ 800 ~ 600 400 200 0 ~ 2000 ! .. 1 500 ~ ~ 1000 500 0 840 860 84 8 844 890 Spot Load Bus Figure 2. 7 Result for CASE 1.2 Random Charging {Off-Peak 30%) 840 860 8 48 844 890 Spot Load Bus Figure 2.8 Result for CASE 1.3 25 830 830 USC Viterbi School of Engineering 10% incremental Charging (Peak 100%) 2 3 Ho. of 10% incA!merts Figure 2.9 Result for CASE 2.1 10% Incremental Charging (Intermediate 55%) Ho. of 1~ incmn.n:s Figure 2.10 Result for CASE 2.2 10% Incremental Charging (Off-Peak 30%) _2~+-----------------~ ;;: " °;:' 200() +-----------------, g 0 ~ 1~+---------- ~ 0 8 9 Figure 2.11 Result for CASE 2.3 26 10 PhD Thesis 11 12 USC Viterbi School of Engineering PhD Thesis Distributed Incremental Charging (Peak 100% ) lncreu~ PtrcerUgt Figure 2.12 Result for CASE 3.1 Distributed Incremental Charging (Intermediate 55% ) ~2000+------------ "" I 1SOO +------ 1000 0 15% 45% 75% 105% 12014 ll5% 140% lncrHH Perc.rUgt Figure 2.13 Result for CASE 3.2 Distributed Incremental Charging (Off-Peak 30%) l500 3000 2500 ~2000 ~ .,, .§ 1500 1000 500 0 15% 30% 45% 75% 90% 105% 12014 ll5% 150% 1e5% lncrHH PtrCl'r'Ugt Figure 2.14 Result for CASE 3.3 27 USC Viterbi School of Engineering PhD Thesis 2.4.1.2 OpenDSS Results for IEEE 34 The most sensitive parameter is the line voltages since the lines current in the range of 50A which represents around 25-30 % of the lines capacity. OpenDSS has a unique feature to plot the lines voltage profile throughout the length of the lines away from the distribution substation as shown in below Figure 2.15. Figure 2.15 IEEE34 bus System Voltage Profile Vs Distance from Distribution Substation The most sensitive buses voltages are summarized in Table 2.6. Table 2.6 Monitored Buses Voltages Bus Base kV Phase 1 pu Phase 2 pu Phase3 pu 814 24.9 0.94683 0.99543 0.98993 852 24.9 0.96451 0.96954 0.9644 890 4.16 0.92336 0.92513 0.91833 The buses above have the lowest voltage profile and the system was tested and EVs added to the network, these buses should be monitored closely. In the original case without any additional load, it can be noticed that bus 890 have low voltage profile and need voltage support through shunt capacitors or voltage regulator. When the system was tested for different scenarios, it will be stopped at the next low bus voltage level to see the amount of additional EV that can be added to the system. In total, six cases were run using EDD based on the three case scenarios and three load levels (periods). The study run cases are shown below: 28 USC Viterbi School of Engineering PhD Thesis • Case 1: Random EV Charging Load Increase (G2V) • Case 2: 10% System Charging Incremental increase Per Spot Bus (G2V) • Case 3: Distributed Charging Incremental Increases (G2V) • Case 4: Random EV Charging Load Increase (G2V+V2G) • Case 5: 10% Charging and Discharging increment of system total load to each Spot Bus (G2V+V2G) • Case 6: Distributed 10% incremental increase of the charging and discharging loads (G2V+V2G) These study results of these cases are summarized below Table 2.7 -Table 2.12 and Figure 2.16 -Figure 2.31. Table 2. 7 Case 1 Result in OpenDSS Spot Load Additional Bus 890 Voltage Bus 852 Voltage Bus 814 Voltage Bus load pu pu pu Original Case below 0.92 NO Violation NO Violation Bus 840 250kW below 0.91 NO Violation NO Violation 500kW below 0.90 0.92 0.92 Original Case below 0.92 NO Violation NO Violation Bus 860 250kW below 0.90 NO Violation NO Violation 500kW below 0.89 0.92 0.92 Original Case below 0.92 NO Violation NO Violation Bus 848 250kW below 0.90 NO Violation NO Violation 500kW below 0.89 0.92 0.92 Original Case below 0.92 NO Violation NO Violation Bus 844 250kW below 0.90 NO Violation NO Violation 500kW below 0.89 0.92 0.92 Original Case 0.92 NO Violation NO Violation lOkW 0.91 NO Violation NO Violation Bus 890 250kW 0.83 NO Violation NO Violation 400kW 0.79 0.92 0.92 500kW 0.77 0.91 0.92 Original Case 0.92 NO Violation NO Violation 250kW 0.91 NO Violation NO Violation Bus 830 500kW 0.9 NO Violation NO Violation 600kW 0.89 0.92 0.92 29 1.15 1.1 1.05 0.95 OS 0.85 2000 1500 1000 500 800 600 400 200 0 USC Viterbi School of Engineering PhD Thesis Load Bus 852 Voltage Profile 1 2 3 4 5 6 7 8 9 w u u ll M u H D IB a w ll n D H - va -Vb - vc - va ( +250kW) - Vb(+250kW) - vc(+250kWJ - Va(+SOOkWJ - Vb(+SOOkW)- vc (+J:XJkWJ Load Bus 852 1 2 3 4 5 6 7 8 9 w u u ll M u H D IB a w ll n D H - va(+SOOkWJ - va(+500to+1750J 1.12 1.08 1.1)4 1 0.96 OS2 0.88 Oz.I 0.8 Load Bus 814 Voltage Profile 1 2 3 4 5 6 7 8 9 w u u ll M u H D IB a w ll n D H 1.15 I.I 1.05 0.95 OS 0.85 0.8 - va -Vb - vc - va( +250<WJ - Vb(+250kW) - Vc(+250kWJ- Va(+SOOkWJ-Vb(+500kW)- Vc (+J:XlkWJ Load Bus 814 I 2 3 4 5 6 7 8 9 w u u ll M u H D IB a w ll n D H - va(+SOOkWJ - va(+500to+l750J Figure 2.16 Case 1 Results in OpenDSS (a) Bus 840,860,848 and 844 Bus 830 2500 111111111111111111111111 2000 111111111111111111111111 1500 1000 500 0 123456 789WllUUMeUUIBH20llllDM 123456789WllUUMeUUIBH20llllDM I Additional EV kW U dditional EV kW Bus 840,860,848 and 844 Bus 830 1200 JJJJJJJJJJJJJJJJJJJJJJJJ 1000 JJJJJJJJJJJJJJJJJJJJJJJJ 800 600 400 200 0 123456 789WllUUMeUUIBH20llllDM 1 23456789WllUU Me UUIBH20llllDM 1 Number o f Evs Level 2Cl1arging 1 Number of Evs Level 1 Charging I Number of Evs Level 2 Charging I Number of Evs Level 1 Charging Figure 2.17 Case 1 Results in OpenDSS (b) 30 USC Viterbi School of Engineering PhD Thesis EVs Charging Cost on Bus 840,860,848 and 844 EVs Charging Cost on Bus 830 500 600 400 500 400 300 I 1111111 200 I 100 I 300 1111111 I I 200 100 1234S6789WUUDMeunIBnwnnna 1234S6789WUUDMeunIBnwnnna Bus 890 852 814 Vl 700 > UJ 600 ...... 0 500 .... Q) 400 ..c E 300 :::l 200 c 3' 100 j;'. 0 • EVS Charging Cost • EVs Charging Cost summer on peak summer off peak winter on peak winter off peak 12:00 PM to 9:00 PM all other time 12:00 PM to 9:00 PM all other time season, time and date June, 1 to October,1 June, 1 to October,1 October,1 to June, 1 October,1 to June, 1 total ($/kwh) 0.48964 0.17177 0.35203 0.1667 Figure 2.18 Case 1 Results in OpenDSS (c) Table 2.8 Case 2 Result in OpenDSS Additional Additional Additional 5% Additional Base case 10% (200kW) 10% (200kW) (lOOkW) on 10% (200kW) on bus 840 on bus 860 bus 848 on bus 848 below0.92 below 0.91 below 0.9 below 0.9 below0.89 No No Violation No Violation No Violation Below 0.92 Violation No Violation No Violation No Violation No Violation Below0.92 Bus 840 and 860 Bus 848 1 2 3 4 s 6 1 s 9 wn u u Me un IB u mn n n M 12 3456789WllllUMeUUIBUIDllnDM 1 Additional EV kw 1 Number of Evs Level 2 Charging 1 Number of Evs Level 1 C harging I Addrtional EV kW I Number of Evs Level 2 Charging I Number of Evs Level 1 Charging Figure 2.19 Case 2 Results in OpenDSS (a) 31 200 180 160 140 120 100 80 60 40 20 USC Viterbi School of Engineering PhD Thesis EVs Charging Cost on Bus 840 and 860 ($) 111111 I I 100 90 80 70 60 so 40 30 20 10 0 EVs Charging Cost on bus 848 ($) 111111 11 I 123456789WUllUMBUUIBemnnnM 123456789WllllUMBUUIBemnnnM • EVI Charging Cost • E\/s Charging Cost Figure 2.20 Case 2 Results in OpenDSS (b) Table 2.9 Case 3 Result in OpenDSS Bus 10% 20% 30% 40% 50% 890 0.91 Below 0.91 Below 0.90 Below 0.89 Below 0.88 852 No Violation No Violation No Violation No Violation Below 0.92 814 No Violation No Violation No Violation No Violation Below 0.92 Bus 860 Bus 840 I/) 100 I/) 40 > 35 UJ ! ~ ~~~~~m~~~LLLLLLL~~~~~~ ~ 123456789WllllUMBUUIBemnnnM I Additional EV kW • Numberof EvsLevel 2 Charging •Number of Evs Level 1 Charging Bus 844 1 Addrtional EV kW • Number of Evs Level 2 Charging •Number of Evs L e vel 1 Charging ..... 30 ~ 25 ~ 20 E 15 ~ 10 100 I/) 1 2 3 4 5 6 7 8 9 rou u u Me UUIBBWll n DM I Additional EV kW • Number of Evs Level 2 Charging •Number of Evs Level 1 Charging Bus 848 i ~ ~~~~~~~~~~~~LLLLL~~~~~~~ ~ 12 34 56789WllllUMeUUIBemnnnM ~ 1 Addrtional EV kW 1 Number of Evs Level 2 Charging •Number of Evs L evel 1 Charging Figure 2.21 Case 3 Results in OpenDSS (a) 32 700 ~ 600 UJ .... 500 0 :;:; 400 -" E 300 ::i c 200 s' .>< 100 180 160 140 120 100 80 60 40 20 USC Viterbi School of Engineering PhD Thesis Bus 890 1 2 3 4 s 6 1 s 9 w u u u M e u n IB n 20 n n n M 1 Additional EV kW 1 Number of Evs Level 2 Charging 1 Number of E vs Level 1 Charging 70 ~ 60 UJ .... 50 0 :;:; 40 -" E 30 ::i c 20 s ' .>< 10 Bus 830 1 2 3 4 s 6 1 s 9 w u u u M e u n IB n 20 n n n M 1 Additional EV kW 1 Number of Evs Level 2 Char~ng 1 Number of Evs Level 1 Char~ng Figure 2.22 Case 3 Results in OpenDSS (b) EV Charging Cost on Buses 844 and 890 EV Charging Cost on Buses 860,840,848 and 830 25 20 :~ I 1 1 l 1 ll1 li1~ 1~m ~ l1 l 1 l I ~ I 1 I ~I I 1 1 2 3 4 s 6 1 s 9 w u u u M e u n IB n 20 n n n M 1 11 1 1 1 2 3 4 5 6 7 8 9 w ll u u M e u " IB n 20 H n ll M I EVI Charging Cort 8us860 I El's Charging Cort 8us840 • EVI Charging Cort Bus 848 • El's Charging Cort Bus 830 1800 1600 1400 1200 1000 800 600 400 200 0 1 E l's Charging Cort Bus 844 1 El's Charging Cort Bus 890 Figure 2.23 Case 3 Results in OpenDSS (c) Bus 840,860,848 and 844 Bus 830 2500 2000 1500 1000 500 1 2 3 c s 6 1 s 9 w u u u M e u n IB n 20 n n n M 1 2 3 4 5 6 7 8 9 W 11 12 13 14 15 16 17 ~ H 20 21 22 23 M • Additional EV kW • Number of EVs level 2 Charging • Additional EV kW • Number of EVs Level 2 Charging •Number of EVs L evel 1 Charging • Dischargi ng kW •Numberof EVslevel !Charging • ChchargingkW • Number of EVs L evel 2 Discharging • Number of EVs le.-.iel 1 Discharging • Number of Evs Level 2 D.scharging • Number of Evs Levell llscha-ging Figure 2.24 Case 4 Results in OpenDSS (a) 33 USC Viterbi 500 400 EV charging and Discharging Cost on bus 840,860,848 and 844 ~ 1111111111'111111 ii I : 1 2 34 567S9wnu11111wewnnnM -300 I EV! Charging Cost I EV! Disdlarging Cost School of Engineering PhD Thesis EVs Charging and Discharging Cost on Bus 830 600 400 2 : lllll111111lllllll 1111 ~ i 234 567S9wnu11111wewnnnM -400 1 EV! Charging Cost 1 EV! Disdlargingcost Figure 2.25 Case 4 Results in OpenDSS (b) Table 2.10 Case 4 Result in OpenDSS Spot EVs Charging EVs Violation Load Bus kW Discharging kW 500 250 No Violation Bus 840 500 495 No Violation 500 500 (Above 1.08pu) Violation on Buses 832, 834,842,840,844,846,848,858,860,862, 864 500 250 No Violation Bus 860 500 495 No Violation 500 500 (Above 1.08pu) Violation on Buses 832, 834,842,840,844,846,848,858,860,862, 864 500 250 No Violation Bus 848 500 495 No Violation 500 500 (Above 1.08pu) Violation on Buses 832, 834,842,840,844,846,848,858,860,862, 864 500 250 No Violation Bus 844 500 495 No Violation 500 500 (Above 1.08pu) Violation on Buses 832, 834,842,840,844,846,848,858,860,862, 864 600 250 No Violation 600 500 No Violation Bus 830 600 750 No Violation 600 810 No Violation 600 815 (Above 1.08pu) Violation on Buses 832, 834,842,840,844,846,848,858,860,862, 864 34 USC Viterbi School of Engineering PhD Thesis Table 2.11Case5 Result in OpenDSS Spot Load Bus EVs Charging EVs Discharging System Violations Bus 840 additional 10% Bus 860 additional 10% Bus 848 additional 5% Bus 848 additional 10% 700 600 500 400 300 200 100 Bus 840 and 860 (kW) 200 200 100 200 1 2 3 4 s 6 1 s 9 w n u u M e u n IB e w n n n M U ddrtional EV kW • Number of EVs Ll'llel 2 Charging •Nu mber of EVs Ll'llel I Charging • Dschar~ng kW • Nu mber of EVs Level 2 Dischar ging I Number of EVs Level 1 CXscharging (kW) 200 No Violation 200 No Violation 100 No Violation (Above 1.08pu) Violation on 200 Buses 832,834,842,840,844, 846,848,858, 860,862 864 Bus 848 350 300 I 23456789WUUDMUUVIBUWDUDM • Additional EV kW • Number of EVs Ll'llel 2 Charging •Number of EVs Ll'llel I Charging • Dschar~ng kW I Number of Evs Level 2 OischNging I Number of Evs Level 1 llscharging Figure 2.26 Case 5 Results in OpenDSS (a) EV Charging and Discharging on bus 840 and 860 200 150 1111111 11111 ill _ 50 I 2 3 4 5 6 7 8 9 10 11 1211111 11 11 18 19 20 21 22 23 24 -100 100 so -150 I EVS Charging Cost I E VS Discllarging Cost 100 80 60 40 20 EV C harging and D ischarging on bus 848 111111 II I : '''.;. '. '""''1'1' 1'11""""""" 1 EVs Charging Cos t 1 EVS DiscllargingCost Figure 2.27 Case 5 Results in OpenDSS (b) 35 USC Viterbi School of Engineering PhD Thesis Table 2.12 Case 6 Result in OpenDSS Distributed Increase on Spot Buses 10% Charging and 10% Discharging 20% Charging and 20% Discharging 30% Charging and 30% Discharging 40% Charging and 40% Discharging 50% Charging and 50% Discharging 40% Charging and 60% Discharging 40% Charging and 70% Discharging 40% Charging and 80% Discharging 40% Charging and 90% 90 80 70 60 50 40 30 20 10 Discharging Bus 860 EVs Charging Violation No Violation No Violation No Violation No Violation Violation on bus 814 No Violation No Violation No Violation No Violation 123456789WllU13Mea!7~U20HllDM • Charging kW • Number of EVs Level 2 Charging •Number of EVs Ll'llel 1 Charging • Discharging kW • Number of EVs Level 2 Discharging • Number of EVs Level 1 Discharging EVs Discharging Violation No Violation No Violation No Violation No Violation No Violation No Violation No Violation No Violation (Above 1.08pu) Violation on Buses 832, 834,842,840,844,846,848,858,860,862 864 40 35 30 25 20 15 10 Bus 840 1 2 3 4 5 6 7 8 9 10 ll 12 13 14 15 16 17 18 19 20 21 22 23 24 • Charging kW • Number of Evs Level 2 Charging •Number of EvSLevel 1 Charging I Discharging kW • Number of Evs Level 2 Discharging • Number of Evs Level 1 Discharging Figure 2.28 Case 6 Results in OpenDSS (a) 36 600 500 400 300 200 100 700 600 500 200 USC Viterbi Bus 844 123456789WllUUMeUDIBDWllUEM • C harging kW • Number of Evs Level 2 C harging •Number of Evs level 1 Charging • !lscharging kW • N umber of Evs level 2 !lsch"'ging• Number of Evs level 1 !lsch,.ging School of Engineering PhD Thesis 90 80 70 Bus 848 f: ~~~~~~l~~~~~l~l~ 0 ------·~-- 1 2 3 4 s 6 7 8 9 w nu u Me u D IBDWll u n M • C harging kW • N umber of Evs l evel 2 Charging •Number of Evs level 1 Charging • !lsch,.ging kW • Number of Evs l e vel 2 !lsch"'gi ng • N umber of Evs l evel 1 !lsch"'gi ng Figure 2.29 Case 6 Results in OpenDSS (b) Bus 890 Bus 830 70 60 so 40 30 20 10 1 2 3 4 s 6 1 a 9 w nu u Me u D IB nw nun M • Charging kW • Number of Evs Level 2 Charging • Charging kW • Number of Evs Level 2 Charging •Number of Evs Level 1 Charging • Oscharging kW 1 Number of Evs L evel 1 Charging • Discha-gingkW • Number of E vs Level 2 llscharging I Number of E vs Level 1 Discharging • Number of Evs Level 2 Discharging • Number of Evs Level 1 Oscharging Figure 2.30 Case 6 Results in OpenDSS (c) E V Charging and Discharging on bus 844 and 890 30 E V Charging and Discharging on bus 860,840, 848 and 830 :~ II II I I I I Ii I I I I II II 1 I 1 1 I I II ii I I I I I I II I I 11 1 11 11 1 : l1 ll l1 ll I I I l 1 ll l1 ll l 1 li l 1 li l 1 ll l,lr f ,h 1 .11 l1 ll l 1 ll l1 ll l 1 ll l1 ll l 1 ll l 1 li l1 ll l1 ll 1 ,1 dr ll l 1 li l1 ll : . . . . . . . . . .. .. .. 11 .. . . .. .. .. . : .............. ~rn~rn ~ 11 ......... .. -200 1 EVs Charging COst on bus844 1 EVs llsdlargingCoston bus 844 IEVs Charging cost on bos890 1 EVs DsdHr~ngCOSton bos 8\IJ 1 EV5 Charging Cost on bus860 1 EVs [)sdlr1gmgCoston bus sro 1 EV5 Charging Cost on bus 840 • fV5 Dsdlcw-gingCoston bus Ml 1 E\15 Charging Cost on bus848 • EV5 Dsdlcw-gingCoston bus 848 • EV5 Charging Cost on bus830 1 EVs Dsdlcw-gingCoston bus SD Figure 2.31 Case 6 Results in OpenDSS (d) 37 USC Viterbi School of Engineering PhD Thesis 2.4.2 USC Micro-grid 2.4.2.1 EDD Results for USC Micro-grid Three cases were run using EDD based on the first case scenario and three load levels (periods). The reason is because the total capacity for USC Micro-grid is too big compared with the summation of the five spot loads (five parking structures on campus). Therefore, it is extremely difficult for the system to reach the limit when running at the other scenarios. To study the USC Micro-grid, Scenario 1 is conducted. The study run cases are described in Section 2.3.2.2. The study results of these cases are summarized below, Figure 2.32 to Figure 2.35. 700 600 500 400 300 200 100 0 1600 1400 1200 1000 800 600 400 200 0 USC System Case 1 - Limit 1 on Parking Structure 595 PSA. PSB PSDL PSDR PSX Figure 2.32 Result for CASE 1 - Limit 1 USC System Case 1 - Limit 2 on Parking Structure 1380 1024 1 38 PSA. PSB PSDL PSDR PSX Figure 2.33 Result for CASE 1 - Limit 2 38 • Original • Case 1.1 Case 1.2 111 Case1 .3 • Original 111Case 1.1 Case 1.2 lil CaSe 1.3 2.4.2.2 6000 5000 4000 3000 2000 1000 0 6000 5000 4000 3000 2000 1000 0 USC Viterbi School of Engineering PhD Thesis USC System Case 1 - Limit 3 on Parking Structure 3250 PSA PSS PSD L 1010 PSDR PSX • Original 11Case 1.1 11Case 1.2 lil CaSe 1.3 Figure 2.34 Result for CASE 1 - Limit 3 Cumulative Increase on Parking Structure PSA PSS PSD L PSD R PSX 100% 55% 30% 100% 55% 30% 100% 55% 30% 100% 55% 30% 100% 55% 30% lillirrit 3-V@XFM- 109.9V lillirrit 2-l@XFM- 225% • Linit 1-l@XFM- 100% Figure 2.35 Result for CASE 1 - Cumulative Increase OpenDSS Results for USC Micro-grid Three cases were tested using OpenDSS based on the case scenarios described in Section 2.3.2.2. The study results of these cases are summarized below, Figure 2.36. 39 2500 - s 2000 :=.::: "'C ra 1500 0 ....I ra c: 1000 0 ;t: 59 "'C "'C 500 <( 0 PSA 2054 USC Viterbi School of Engineering 1424 1036 50 90200 PSB 50 74 PSDl Parking Structure PSD2 Figure 2.36 Result for CASE 1 in OpenDSS PhD Thesis • No violation • 1@100% • 1@225% • V@0.9pu 1528 536 446 PSX The available loads in USC standard represent only the 100% loading case. However, it is important to analyze the system over a period of time with different loadings. Therefore, data from USC Facility Management Seivices system for a random day in the spring semester where the load reaches its peak is analyzed. As this system does not cover all the loads in USC micro-grid, the load factor for each hour will be calculated and then multiplied by the total load in the 100% loading case. The below calculation for hour 12:00 a.m. to illustrate this calculation: 100% loading total load= 26,943.7 kW April 2nd, 2015 Total load from USC Micro-grid at 12:00 a.m. = 4,192.06 kW April 2nd, 2015 maximum load from USC Micro-grid at= 5,509.45 kW (at 11:00 a.m.) Projected total load for USC micro-grid on April 2n d at 12:00 a.m. = 4 • 192 · 06 x 26,943.7 = 5,509.45 20,501 kW This method was applied to four days from each season with the April maximum being the overall maximum of the loads and the obtained data is shown on the below chart, Figure 2.37. 40 USC Viterbi School of Engineering PhD Thesis Total Load (kW} 30,000 25,000 20,000 1 15,000 10,000 ~ ~ --------- 5,000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9. 9. 9. 0 9. 9. 9. 9. 9. 0 9. 9. 9. 0 9. 9. 9. 9. 9. 0 9. 9. 9. 0 ...... N M tj' Ll'l \.D ,.._ 00 Ctl 0 ...... N M tj' Ll'l \.D ,.._ 00 Ctl 0 ...... N M ...... ...... ...... ...... ...... ...... ...... ...... ...... ...... N N N N - 21-Jan - 2-Apr - 7-Jul - 20-0ct Figure 2.37 Projected Total USC load for four days Due to time limitation, cases 2 and 3 will only be applied on the maximum load case on April 2nd 2015. The results can be found in Table 2.13 and Figure 2.38 to Figure 2.42. Summary of Case 2: #of EV Charging Max/Min 140 120 100 80 60 1 1 40 I 20 1 1 1 • II 0 • • •• I 11 Ll L2 L1 L2 Ll L2 Ll L2 Ll L2 PSA PSB PSDl PSD2 PSX • 6:00 AM • 11:00 AM Figure 2.38 Case 2 Results in OpenDSS 41 USC Viterbi 50 40 Cost of EV Charging in$ (PSA) 30 :: 111111111111 11 0:00 2:00 4:00 6:00 8:0010:0012:0014:0016:0018:0020:0022:00 30 25 Cost of EV Charging in$ (PSDl) :: 11111111111111111111 School of Engineering PhD Thesis Cost of EV Charging in$ (PSB) 25 20 15 ·: 111111111111 11 0:00 2 :00 4:00 6:00 8:()()10: 00 2: 00 4: 006: 008: 00!0 CXl2:00 Cost of EV Charging in$ (PSD2) 20 15 0 I I I I I I I I I I I I I 10 11 0:00 2:00 4:00 6:00 8:0010:0012:0014:0016:0018:0020:0022:00 0:00 2:00 4:00 6:008:0010:0012:0014:0016:0018:0020:0022:00 Cost of EV Charging in$ (PSX) 40 '.~ 111111111111 II I II 0:00 2:00 4:00 6:00 8:0010:0012:0014:0016:0018:0020:0022:00 Figure 2.39 Case 2 Results in OpenDSS Table 2.13 Southern California Edison Electricity Rate Summer on Summer off Winter on Winter off peak peak peak peak Season, Date June 1st to June istto October 1st to October pt to and Time October 1st October 1st June 1st June 1st 12:00 PM to all other time 12:00 PM to all other time 9:00 PM 9:00PM Total ($/kWh) 0.32991 0.12973 0.22267 0.12466 42 800 600 400 200 0 600 soo 400 300 200 100 0 USC Viterbi Maximum Charge/Discharge in kW (PSA) --Discharging -- chargil'l:: Maximum Charge/Discharge in kW (PSD1) --Discharging -- Char,1ing School of Engineering PhD Thesis 350 300 250 200 150 100 so 0 200 150 100 so 0 Maximum Charge/Discharge in kW (PSB) --Discharging --charging Maximum Charge/Discharge in kW (PSD2) --Discharging --Chargire Maximum Charge/Discharge in kW (PSX) 600 soo 400 300 200 100 0 --Discharging --charging Figure 2.40 Case 3 Results in OpenDSS (a) 43 40'.J 300 200 300 USC Viterbi Maximum # of E V Charging/Discharging (PSA) 000000000000000000000000 000000000000000000000000 o~Nm~ ~ w~oomo~Nm~~w~oomo~Nm .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 N N N N • Discha rging L evel 1 • Discharging L evel 2 I C harging Level 1 I Charging Level 2 Maximum # of E V Charging/Discharging (PSDl) 000000000000000000000000 000000000000000000000000 O~Nm~~w~oomo~Nm~~w~oomo~Nm ..-t.-IMM.-i.-i.-i.-1..-t.-iNNNN • Discharging L evel 1 • Discharging L evel 2 I C harging Level 1 I C harging L evel 2 School of Engineering PhD Thesis 200 100 Maximum # of EV Charging/Discharging (PSB) 000000000000000000000000 000000000000000000000000 o~N~~ ~ w~oomo~Nm~~w~oomo~Nm .-f .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 N N N N • Discharging Level 1 • Discharging Level 2 I C harging L evel 1 I Charging L evel 2 Maximum# of E V Charging/Discharging (PSD2) ~ u. ~ .. 1 . 1.LliJ . L.1 .. u~. 1.L 000000000000000000000000 000000000000000000000000 O~Nm~~w~oomo~Nm~~w~oomo~Nm .-I .-i .-i .-i .-I .-I ..-I ...-1 M .-i N N N N • Discharging Level 1 • Discharging Level 2 I C harging L evel 1 I Charging L evel 2 Maximum # of EV Charging/Discharging (PSX) 300 000000000000000000000000 000000000000000000000000 o~Nm~~~~oomo~Nm~~w~oomo~Nm .-+ .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 .-1 N N N N • Discharging Level 1 • Discharging Level 2 I C harging L evel 1 I C harging L evel 2 Figure 2.41 Case 3 Results in OpenDSS (b) 44 S160.0J S140.0J suo.oo Sl OO.OJ sao.oo $60.00 S40.oo s20.oo S· suo.oo S lOOOJ sao.oo USC Viterbi Cost of EV Dicharging in $ (PSA) 11111111 11 0:00 2:00 4:00 6:00 8:0010:0J12:CXJ14:CXJ16:CXJ18:CXJ20:0J22:CXJ Cost of EV Dicharging in$ (PSDl ) School of Engineering PhD Thesis sao.oo s10.oo S60.oo sso.oo $40.00 $30.00 s20.oo s1 0.oo S· Cost of EV Dicha rgi ng in$ (PSB) I I I I I I I I I I 0:00 2:00 4:00 6:00 8:00 10:0J12:CXJ14:CXJ16:CXJ18:0J20:0J22:CXJ Cost of EV Dicharging in$ (PSD2) ::~ 11111111 I 11 S40.oo S35.oo $30.00 $ 25.00 s20.oo SlS.00 s1 0.oo ss.oo S· I I I I I I I I I I 0:00 2:00 4:00 6:00 8:0010:0J12:0J14:0J16:0J18:0J20:0J22:0J 0:00 2:00 4:00 6:00 8:00 10:0J12:0J14:0J16:0J18:0J20:0J22:0J Cost of EV Dicharging in$ (PSX) su o.oo SlOO.OJ sao.oo :; 11111111 11 0:00 2:00 4:00 6:00 8:00 10:0J12:CXJ14:0J16:CXJ18:CXJ20:0J22:CXJ Figure 2.42 Case 3 Results in OpenDSS (c) 2.4.3 Chatsworth Distribution System Several cases have been tested using the three case scenarios and three load levels (periods). The study run cases are the similar as those shown in Section 2.3.2.3. For this study, there are two separate studies for the circuit 22 and 23. These study results of these cases have been summarized below. 2.4.3.1 Results of Chatsworth 88-22 45 USC Viterbi School of Engineering Chatsworth 88-22 Case 1 1 2 4 9 14 18 19 20 21 29 JO 31 45 40 47 49 ~ 57 58 e1 ea eg 10 11 n so Figure 2.43 Result for 88-22 CASE 1 Chatsworth 88-22 Case 2 4000 3000 II II I - ..... 2000 • 1 000 I , r I . •• iii if p ~ I 0 aa n 11 10 eg ea e1 58 57 58 49 47 40 45 31 30 29 21 20 19 18 14 9 4 2 1 aa 12 Figure 2.44 Result for 88-22 CASE 2 Chatsworth 88-22 Case 3 Figure 2.45 Result for 88-22 CASE 3 46 PhD Thesis • Case 1.1 • Case 1.2 Case 1.3 • Case2.1- 10% • Case2.2-15% Case 2.3- 20% • Case 3.1 • Case3.2 Case3.3 USC Viterbi School of Engineering PhD Thesis 1. Utilization factor - the ratio of peak power demand to the maximum power that the system can accommodate. The utilization factor for Chatsworth circuit 88-22 can be calculated by the following equation. The results can be found in Table 2.14. Peak Power Demand Utilization Factor (UF) = ------ Maximum Power Table 2.14 Utilization Factors for Chatsworth 88-22 under Different Loading Conditions Benchmark Limit 1 Limit 2 CASE 1 UF (reg) UF (dis) UF (reg) UF (dis) UF (reg) UF (dis) 0.551 0.134 1.97 0.478 3.75 0.911 Benchmark Limit 1 Limit 2 CASE2 UF (reg) UF (dis) UF (reg) UF (dis) UF (reg) UF (dis) 0.302 0.073 1.99 0.485 4.38 1.07 Benchmark Limit 1 Limit 2 CASE3 UF (reg) UF (dis) UF (reg) UF (dis) UF (reg) UF (dis) 0.164 0.04 1.85 0.45 4.85 1.18 2. Capacity factor - the ratio of peak power handling for a system component such as a transformer to the maximum power that the component can accommodate. The transformer capacity factor is calculated by the following equation. The results of capacity factor under different loading conditions can be found below, Table 2.15 - 2.17. Peak Power of Transformer Capacity Factor (CF) = ----------- Maximum Power of Transformer 47 USC Viterbi School of Engineering PhD Thesis Table 2.15 Capacity Factors for Chatsworth 88-22 under Peak Loading Condition 100% Transformer kW CF kW CF Name Total CF Name in kVA Limit Limit Limit Limit in Map kW Bench DEW 1 1 2 2 147701 67 25 5.0 0.20 17.5 0.70 35.0 1.40 147702 68 25 5.0 0.20 17.5 0.70 35.0 1.40 147704 71 15 3.0 0.20 10.5 0.70 21.0 1.40 147709 70 75 15.0 0.20 64.8 0.86 134.4 1.79 147714 90,91 38 12.6 0.34 25.4 0.68 43.2 1.15 147718 72,73 38 9.8 0.26 34.3 0.91 68.6 1.83 147719 76,77 50 29.7 0.59 57.7 1.15 96.9 1.94 147720 78,79,80 25 19.9 0.80 55.2 2.21 104.5 4.18 147721 74,75 50 16.8 0.34 54.3 1.09 106.8 2.14 147729 81,82,83 25 15.0 0.60 89.3 3.57 193.2 7.73 147730 84,85,86 50 29.7 0.59 71.7 1.43 130.5 2.61 147731 87,88,89 100 59.4 0.59 96.9 0.97 149.4 1.49 147745 93,94 15 5.1 0.34 26.1 1.74 55.5 3.70 147746 95,96 25 8.4 0.34 82.7 3.31 186.6 7.46 147747 100,101 25 11.2 0.45 53.2 2.13 112.0 4.48 147749 97,98,99 75 44.7 0.60 76.2 1.02 120.3 1.60 147756 102,103,104 50 29.7 0.59 104.0 2.08 207.9 4.16 147757 105,106 100 19.7 0.20 94.0 0.94 197.9 1.98 147758 107,108 100 33.9 0.34 97.7 0.98 186.9 1.87 147767 109,110 50 29.7 0.59 178.2 3.56 386.1 7.72 147768 113,114 75 14.1 0.19 98.9 1.32 217.5 2.90 147769 115,116 75 25.5 0.34 129.0 1.72 273.9 3.65 147770 117,118 25 14.1 0.56 162.6 6.50 370.5 14.82 147771 119,120 50 16.8 0.34 66.1 1.32 135.0 2.70 147772 111,112 167 41.4 0.25 76.7 0.46 126.0 0.75 147786 121,122,123 100 59.4 0.59 171.2 1.71 327.6 3.28 48 USC Viterbi School of Engineering PhD Thesis Table 2.16 Capacity Factors for Chatsworth 88-22 under Intermediate Loading Condition 55% Transformer kW CF kW CF Name Total CF Name in kVA Limit Limit Limit Limit in Map kW Bench DEW 1 1 2 2 147701 67 25 5 0.20 20 0.80 40 1.60 147702 68 25 5 0.20 20 0.80 40 1.60 147704 71 15 3 0.20 12 0.80 24 1.60 147709 70 75 15 0.20 74.7 1.00 154.3 2.06 147714 90,91 37.5 12.6 0.34 27.9 0.74 48.3 1.29 147718 72,73 37.5 9.8 0.26 39.2 1.05 78.4 2.09 147719 76,77 50 29.7 0.59 63.3 1.27 108.1 2.16 147720 78,79,80 25 19.9 0.80 62.2 2.49 118.6 4.74 147721 74,75 50 16.8 0.34 61.8 1.24 121.8 2.44 147729 81,82,83 25 15 0.60 104.1 4.16 222.9 8.92 147730 84,85,86 50 29.7 0.59 80.1 1.60 147.3 2.95 147731 87,88,89 100 59.4 0.59 104.4 1.04 164.4 1.64 147745 93,94 15 5.1 0.34 30.3 2.02 63.9 4.26 147,746 95,96 25 8 0.34 98 3.90 216 8.65 147747 100,101 25 11.2 0.45 61.6 2.46 128.8 5.15 147,749 97,98,99 75 45 0.60 83 1.10 133 1.77 147,756 102,103,104 50 30 0.59 119 2.38 238 4.75 147,757 105,106 100 20 0.20 109 1.09 228 2.28 147,758 107,108 100 34 0.34 110 1.10 212 2.12 147,767 109,110 50 30 0.59 208 4.16 446 8.91 147,768 113,114 75 14 0.19 116 1.54 251 3.35 147,769 115,116 75 26 0.34 150 2.00 315 4.20 147,770 117,118 25 14 0.56 192 7.69 430 17.20 147,771 119,120 50 17 0.34 76 1.52 155 3.09 147,772 111,112 167 41 0.25 84 0.50 140 0.84 147786 121,122,123 100 59.4 0.59 193.5 1.94 372.3 3.72 49 USC Viterbi School of Engineering PhD Thesis Table 2.17 Capacity Factors for Chatsworth 88-22 under Off-peak Loading Condition 30% Transformer kW CF kW CF Name Total CF Name in kVA Limit Limit Limit Limit in Map kW Bench DEW 1 1 2 2 147701 67 25 5 0.20 20 0.80 45 1.80 147702 68 25 5 0.20 20 0.80 45 1.80 147704 71 15 3 0.20 12 0.80 27 1.80 147709 70 75 15 0.20 74.7 1.00 174.2 2.32 147714 90,91 37.5 12.6 0.34 27.9 0.74 53.4 1.42 147718 72,73 37.5 9.8 0.26 39.2 1.05 88.2 2.35 147719 76,77 50 29.7 0.59 63.3 1.27 119.3 2.39 147720 78,79,80 25 19.9 0.80 62.2 2.49 132.7 5.31 147721 74,75 50 16.8 0.34 61.8 1.24 136.8 2.74 147729 81,82,83 25 15 0.60 104.1 4.16 252.6 10.10 147730 84,85,86 50 29.7 0.59 80.1 1.60 164.1 3.28 147731 87,88,89 100 59.4 0.59 104.4 1.04 179.4 1.79 147745 93,94 15 5.1 0.34 30.3 2.02 72.3 4.82 147,746 95,96 25 8 0.34 98 3.90 246 9.84 147747 100,101 25 11.2 0.45 61.6 2.46 145.6 5.82 147,749 97,98,99 75 45 0.60 83 1.10 146 1.94 147,756 102,103,104 50 30 0.59 119 2.38 267 5.35 147,757 105,106 100 20 0.20 109 1.09 257 2.57 147,758 107,108 100 34 0.34 110 1.10 238 2.38 147,767 109,110 50 30 0.59 208 4.16 505 10.10 147,768 113,114 75 14 0.19 116 1.54 285 3.80 147,769 115,116 75 26 0.34 150 2.00 357 4.76 147,770 117,118 25 14 0.56 192 7.69 489 19.57 147,771 119,120 50 17 0.34 76 1.52 174 3.49 147,772 111,112 167 41 0.25 84 0.50 154 0.92 147786 121,122,123 100 59.4 0.59 193.5 1.94 417 4.17 50 USC Viterbi School of Engineering PhD Thesis 3. Load factor - the ratio of actual load to peak load. The load factors are 1, 0.55, and 0.3 for the three different loading conditions, respectively. 4. Power factor - the ratio of real useable power that can perform work to the magnitude of apparent power (current times voltage). Table 2.18 Power Factors for Chatsworth 88-22 under Peak Loading Condition (100%) Benchmark Limit 1 Limit 2 P(kW) 578.59 2063.96 3937 Q (kVAR) 87.51 298.3 677.18 PF 0.989 0.990 0.986 Table 2.19 Power Factors for Chatsworth 88-22 under Intermediate Loading Condition (55%) Benchmark Limit 1 Limit 2 P(kW) 317.13 2093.27 4603.87 Q(kVAR) 46.59 291.75 815.03 PF 0.989 0.990 0.985 Table 2.20 Power Factors for Chatsworth 88-22 under Off-peak Loading Condition (30%) Benchmark Limit 1 Limit 2 P(kW) 172.68 1941.35 5093.18 Q(kVAR) 24.94 259.81 932.32 PF 0.990 0.991 0.984 5. Power Quality - an overall rating of voltage and current wave shapes, frequency regulation, and noise or harmonic distortion levels. Table 2.21 Power Quality for Chatsworth 88-22 under Different Loading Conditions Peak Intermediate Off-peak P(kW) 3937 4603.87 5093.18 Q(kVAR) 677.18 815.03 932.32 Worst 109.6 109.7 109.2 Violation (V) A Voltage (V) 10.4 10.3 10.8 51 USC Viterbi School of Engineering 2.4.3.2 Results of Chatsworth 88-23 Chatsworth 88-23 Case 1 8018158 Ul81782082e828838 839840 841842 849 &5-4 81!88C9 8708n 873879887 890891894902913923 924 925929 Figure 2.46 Result for 88-23 CASE 1 Chatsworth 88-23 Case 2 1 000 I r t ~ ~ .~11 1 1 I " Figure 2.47 Result for 88-23 CASE 2 52 PhD Thesis • Case 1.1 • Case 1.2 Case 1.3 • Case 2 .1-10% • Case 2 .2-15% . Case 2 .3-20% USC Viterbi School of Engineering PhD Thesis Chatsworth 88-23 Case 3 ~~!---------------------------~ g ~ 50001------------------==----~· ...... -11 1-----~ ~ • Case 3.1 ~ 40001-------------=-~ • Case 3.2 Case3.3 000 0 Figure 2.48 Result for 88-23 CASE 3 1. Utilization factor - the ratio of peak power demand to the maximum power that the system can accommodate. The utilization factor for Chatsworth circuit 88-23 can also be calculated by the following equation. The results can be found in Table 2.22. Peak Power Demand Utilization Factor (UF) = ------ Maximum Power Table 2.22 Utilization Factors for Chatsworth 88-23 under Different Loading Condition Peak Benclunark Limit 1 Limit 2 Limit 3 Loading Condition '. - (dis) (dis) (dis) (reg) (dis) ( l 00%) 0.838 0.204 1.68 0.408 2.55 0.62 4.66 1.13 Intennediate Benchmark Limit 1 Limit 2 Limit 3 Loading UF UF UF UF UF UF UF U F Condition (reg) (dis) (reg) (dis) (reg) (dis) (reg) (dis) (55%) 0.452 0.11 1.67 0.406 NIA NIA 5.53 1.34 Off-peak Benchmark Limit 1 Limit 2 Limit 3 Loading UF UF UF UF UF UF UF UF Condition '. - (dis) (re (dis) (re (dis) (re (dis) (30%) 0.06 1.87 0.454 NIA NIA 6.25 1.52 2. Capacity factor - the ratio of peak power handling for a system component such as a transformer to the maximum power that the component can accommodate. 53 USC Viterbi School of Engineering PhD Thesis The transformer capacity factor is calculated by the following equation. The results of capacity factor under different loading conditions can be found below, Table 2.23, Table 2.24 and Table 2.25. Peak Power of Transformer Capacity Factor (CF) = ---------- Maximum Power of Transformer Table 2.23 Capacity Factors for Chatsworth 88-23 under Peak Loading Condition (100%) Load Bus Name CF kW CF kW CF Name in DEW in Map kVA Total kW Bench Limit 1 Limit 1 Limit 2 Limit 2 147801 65 300 59.4 0.20 100.98 0.34 308.88 1.03 147815 04,05 75 16.9 0.23 28.73 0.38 87.88 1.17 147816 67,68,69 100 59.4 0.59 100.98 1.01 308.88 3.09 147817 07,08 75 14.1 0.19 23.97 0.32 73.32 0.98 147820 11,12 75 16.9 0.23 28.73 0.38 87.88 1.17 147826 3 300 59.4 0.20 100.98 0.34 308.88 1.03 147828 10 300 59.4 0.20 100.98 0.34 308.88 1.03 147838 74,75 50 11.2 0.22 19.04 0.38 58.24 1.16 147839 15,16 50 11.2 0.22 19.04 0.38 58.24 1.16 147840 70,71 100 28.3 0.28 48.11 0.48 147.16 1.47 147841 17,18,19 50 29.7 0.59 50.49 1.01 154.44 3.09 147842 20,21 100 59.4 0.59 100.98 1.01 308.88 3.09 147849 38 38 7.4 0.20 12.58 0.34 38.48 1.03 147854 39,40 25 8.4 0.34 14.28 0.57 43.68 1.75 147868 33 100 19.8 0.20 33.66 0.34 102.96 1.03 147869 29 25 5.0 0.20 8.50 0.34 26.00 1.04 147870 25,26,27 25 15.0 0.60 25 .50 1.02 78.00 3.12 147872 13,14 25 8.4 0.34 14.28 0.57 43.68 1.75 147873 34 167 33.1 0.20 38.98 0.23 68.38 0.41 147879 9 150 29.7 0.20 50.49 0.34 154.44 1.03 147887 36 167 33.1 0.20 56.27 0.34 172.12 1.03 147890 6 150 29.7 0.20 50.49 0.34 154.44 1.03 147891 37 167 33.1 0.20 56.27 0.34 172.12 1.03 147894 72,73 75 28.3 0.38 48.11 0.64 147.16 1.96 147902 30 75 14.9 0.20 25.33 0.34 77.48 1.03 147913 76,77,78 100 22.2 0.22 37.74 0.38 115.44 1.15 147923 80 100 19.8 0.20 33.66 0.34 102.96 1.03 147924 81 167 33.1 0.20 56.27 0.34 172.12 1.03 147925 2 300 59.4 0.20 100.98 0.34 308.88 1.03 147929 79 167 33.1 0.20 56.27 0.34 172.12 1.03 54 USC Viterbi School of Engineering PhD Thesis Table 2.24 Capacity Factors for Chatsworth 88-23 under Intermediate Loading Condition (55%) Load Bus Name CF kW CF kW CF kVA Total kW Name in DEW in Map Bench Limit 1 Limit 1 Limit 2 Limit 2 147801 65 300 32.67 0.11 121.77 0.41 389.07 1.30 147815 04,05 75 9.30 0.12 34.65 0.46 110.70 1.48 147816 67,68,69 100 32.67 0.33 121.77 1.22 389.07 3.89 147817 07,08 75 7.76 0.10 28.91 0.39 92.36 1.23 147820 11,12 75 9.30 0.12 34.65 0.46 110.70 1.48 147826 3 300 32.67 0.11 121.77 0.41 389.07 1.30 147828 10 300 32.67 0.11 121.77 0.41 389.07 1.30 147838 74,75 50 6.16 0.12 22.96 0.46 73.36 1.47 147839 15,16 50 6.16 0.12 22.96 0.46 73.36 1.47 147840 70,71 100 15.57 0.16 58.02 0.58 185.37 1.85 147841 17,18,19 50 16.34 0.33 60.89 1.22 194.54 3.89 147842 20,21 100 32.67 0.33 121.77 1.22 389.07 3.89 147849 38 38 4.07 0.11 15.17 0.40 48.47 1.29 147854 39,40 25 4.62 0.18 17.22 0.69 55.02 2.20 147868 33 100 10.89 0.11 40.59 0.41 129.69 1.30 147869 29 25 2.75 0.11 10.25 0.41 32.75 1.31 147870 25,26,27 25 8.25 0.33 30.75 1.23 98.25 3.93 147872 13,14 25 4.62 0.18 17.22 0.69 55.02 2.20 147873 34 167 18.21 0.11 30.81 0.18 68.61 0.41 147879 9 150 16.34 0.11 60.89 0.41 194.54 1.30 147887 36 167 18.21 0.11 67.86 0.41 216.81 1.30 147890 6 150 16.34 0.11 60.89 0.41 194.54 1.30 147894 72,73 75 15.57 0.21 58.02 0.77 185.37 2.47 147902 30 75 8.20 0.11 30.55 0.41 97.60 1.30 147913 76,77,78 100 12.21 0.12 45.51 0.46 145.41 1.45 147923 80 100 10.89 0.11 40.59 0.41 129.69 1.30 147924 81 167 18.21 0.11 67.86 0.41 216.81 1.30 147925 2 300 32.67 0.11 121.77 0.41 389.07 1.30 147929 79 167 18.21 0.11 67.86 0.41 216.81 1.30 55 USC Viterbi School of Engineering PhD Thesis Table 2.25 Capacity Factors for Chatsworth 88-23 under Off-peak Loading Condition (30%) Load Bus Name kVA Total kW CF kW CF kW CF Name in DEW in Map Bench Limit 1 Limit 1 Limit 2 Limit 2 147801 65 300 17.82 0.06 136.62 0.46 433.62 1.45 147815 04,05 75 5.07 0.07 38.87 0.52 123.37 1.64 147816 67,68,69 100 17.82 0.18 136.62 1.37 433.62 4.34 147817 07,08 75 4.23 0.06 32.43 0.43 102.93 1.37 147820 11,12 75 5.07 0.07 38.87 0.52 123.37 1.64 147826 3 300 17.82 0.06 136.62 0.46 433.62 1.45 147828 10 300 17.82 0.06 136.62 0.46 433.62 1.45 147838 74,75 50 3.36 0.07 25.76 0.52 81.76 1.64 147839 15,16 50 3.36 0.07 25.76 0.52 81.76 1.64 147840 70,71 100 8.49 0.08 65.09 0.65 206.59 2.07 147841 17,18,19 50 8.91 0.18 68.31 1.37 216.81 4.34 147842 20,21 100 17.82 0.18 136.62 1.37 433.62 4.34 147849 38 38 2.22 0.06 17.02 0.45 54.02 1.44 147854 39,40 25 2.52 0.10 19.32 0.77 61.32 2.45 147868 33 100 5.94 0.06 45.54 0.46 144.54 1.45 147869 29 25 1.50 0.06 11.50 0.46 36.50 1.46 147870 25,26,27 25 4.50 0.18 34.50 1.38 109.50 4.38 147872 13,14 25 2.52 0.10 19.32 0.77 61.32 2.45 147873 34 167 9.93 0.06 26.73 0.16 68.73 0.41 147879 9 150 8.91 0.06 68.31 0.46 216.81 1.45 147887 36 167 9.93 0.06 76.13 0.46 241.63 1.45 147890 6 150 8.91 0.06 68.31 0.46 216.81 1.45 147891 37 167 9.93 0.06 76.13 0.46 241.63 1.45 147894 72,73 75 8.49 0.11 65.09 0.87 206.59 2.75 147902 30 75 4.47 0.06 34.27 0.46 108.77 1.45 147913 76,77,78 100 6.66 0.07 51.06 0.51 162.06 1.62 147923 80 100 5.94 0.06 45.54 0.46 144.54 1.45 147924 81 167 9.93 0.06 76.13 0.46 241.63 1.45 147925 2 300 17.82 0.06 136.62 0.46 433.62 1.45 147929 79 167 9.93 0.06 76.13 0.46 241.63 1.45 56 USC Viterbi School of Engineering PhD Thesis 3. Load factor - the ratio of actual load to peak load. The load factors are 1, 0.55, and 0.3 for the three different loading conditions, respectively. 4. Power factor - the ratio of real useable power that can perform work to the magnitude of apparent power (current times voltage). Table 2.26 Power Factors for Chatsworth 88-23 under Peak Loading Condition (100%) Benchmark Limit 1 Limit 2 P(kW) 865.83 1764.36 4892.73 Q (kVAR) 133.26 255.53 868.97 PF 0.988 0.989 0.985 Table 2.27 Power Factors for Chatsworth 88-23 under Intermediate Loading Condition (55%) Benchmark Limit 1 Limit 2 P(kW) 474.36 1752.42 5809.72 Q(kVAR) 70.37 236.46 1072.26 PF 0.989 0.991 0.983 Table 2.28 Power Factors for Chatsworth 88-23 under Off-peak Loading Condition (30%) Benchmark Limit 1 Limit 2 P(kW) 258.22 1961.53 6559.26 Q(kVAR) 37.52 259.2 1263.21 PF 0.989 0.991 0.982 5. Power Quality - an overall rating of voltage and current wave shapes, frequency regulation, and noise or harmonic distortion levels. Table 2.29 Power Quality for Chatsworth 88-23 under Different Loading Conditions Peak Intermediate Off-peak P(kW) 4892.73 5809.72 6559.26 Q(kVAR) 868.97 1072.26 1263.21 Worst Violation 109.6V 109.7V 108.9V A Voltage (V) 10.4 10.3 11.1 57 USC Viterbi School of Engineering PhD Thesis 2.5 Summaries of the Results of the Analysis and Benefits In this section's analysis, the author assumes that each electric vehicle charging in level 1 has capacity of 2 kW (20A) at 120V with 12 hours to be fully charged. Level 2 charging has capacity of 8 kW (32A) at 240V with 6 hours to be fully charged [53-56]. The load flow analysis has been run for all three cases and the system has been checked for voltage and current violations. The limits for 120-volt base is set from 110-127V, as has already been mentioned in Section 2.3.3. Result analysis of the three systems can be found in the following subsections. 2.5.1 IEEE 34 Node Distribution Test System 2.5.1.1 EDD Results and Conclusions The result of Case 1 shows that: • For load bus 840, 860, 848, 844 under peak load condition, around 377 kW more load can be attached, which is equivalent to 186 PHEVs in level 1 charging or 47 PHEVs in level 2 charging. While for load period 2 and 3, nearly 1580 kW and 2090 kW more demand can be added, respectively. • For load bus 890 under peak load condition, only 27 kW more load can be added, which is equivalent to 13 PHEVs in level 1 charging or 3 PHEVs in level 2 charging. While for load period 2 and 3, almost 750 kW and 870 kW more demand can be attached, respectively. • For load bus 830 under peak load condition, 512 kW can be increased to reach the voltage limit, which is equivalent to 256 PHEVs in level 1 charging or 64 PHEVs in level 2 charging. While for load period 2 and 3, nearly 1880 kW and 2450 kW more demand can be increased, respectively. The result of Case 2 reflects: • Under period 1 (peak load), the system will reach its voltage limit when the three load buses (840-860-848) increase their kW demand by 10%, and the voltage limit will occur at the node bus 814. The results also show that 615 kW additional demand can be added to the whole system, which is equivalent to 307 PHEVs in level 1 charging or76 PHEVs in level 2 charging. • Under period 2 (intermediate load), the system will reach its voltage limit after all the load buses increase their kW demand by 10%, while load bus 840 and 860 increase 20%, and the voltage limit will occur at the node bus 814. The results also show that 1640 kW 58 USC Viterbi School of Engineering PhD Thesis additional demand can be handled to the whole system, which is equivalent to 820 PHEVs in level l charging or 205 PHEVs in level 2 charging. • Under period 3 (off-peak load), the system will reach its voltage limit almost when all the six load buses increase its kW demand by 20%, and the voltage limit will occur at the node 852. The results also show that 2255 kW additional demand can be added to the whole system, which is equivalent to 1127 PHEVs in level 1 charging or 281 PHEVs in level 2 charging. The result of Case 3 illustrates: • Under period 1 (peak load), the system will reach its voltage limit when all the six load buses increase its kW demand by 50% with respect to themselves, and the voltage limit will occur at the node bus 814. The results also show that 523.5 kW additional demand can be handled to the whole system, which is equivalent to 261 PHEVs in level 1 charging or 65 PHEV s in level 2 charging. While under period 2 and period 3, the system will run into voltage limit issue when all the six buses are added their demand by 140% and 163%, which is equivalent to 1465.8 kW and 1675.2 kW, respectively. 2.5.1.2 OpenDSS Results and Conclusions From different cases, it is able to identify the maximum kW that can be added to each spot bus for different test case scenario. From that, time of use rate can be used to calculate the charging and discharging price. This study and the developed software program can pave the road for a better demand response analysis either daily, weekly or yearly. Table 2.30 below summarizes the different cases total charging and discharging energy of the day and summarize the total cost for charging and discharging for each case. • The additional load on bus 830 for case 1 and case 4 seems to be the best place to have the highest possible energy and eventually the highest number of Electric Vehicles. The cost might need to be updated once having the charging/discharging costs. The cost might not be a good indication for the cheapest place to have the charging power or the most place to discharge the power because it is the same for every bus so it is going to be proportional to the amount of kW. • Case 3 and case 6 are preferred in case it is needed to cover more geographical area of customers to charge and discharge their vehicles. 59 USC Viterbi School of Engineering PhD Thesis Table 2.30 Summary of the Different Cases Charging and Discharging Costs Case 1 buses 840,860,848,844 Case 1 bus 830 Case 2 Case3 Case 4 buses 840,860,848,844 Case 4 bus 830 Case 5 Case 6 Total charging Energy (kWh) 26100 31320 26100 21861.4 26100 31320 26100 21861.4 2.5.2 USC Micro-grid Total discharging Energy (kWh) 0 0 0 0 2295 4050 2497.5 4198.5 2.5.2.1 EDD Results and Conclusions The result of Case 1 shows that: Net Energy (kWh) 21861.4 23805 27270 23602.5 17662.9 Total charging cost($) 5179.6 6183.8 7420.6 6183.8 5179.6 Total discharging cost($) 0 0 0 0 -1123 -1983 -1222 -2055.8 Net cost (S) - - ~ I I . . . . . 5179.6 5060.1 5437.6 4960.9 3123.8 • For PSA, around 500 kW, 1300 kW, 5600 kW more demand can be attached, in order to reach the first, second, and third limit, respectively. It is equivalent to 250 PEVs in level 1 charging or 62 PEVs in level 2 charging, in order to reach the first current limit at transformer attached ahead of the load bus. • For PSB, about 650 kW and 3100 kW more demand can be added, in order to reach the first and second limit, respectively. It is equivalent to 325 PEVs in level 1 charging or 81 PEVs in level 2 charging to reach the first current limit at transformer. • For PSD-Left, almost 360 kW, 950 kW, 4600 kW more demand can be put to reach the first, second, and third limit, respectively. It is equivalent to 180 PEVs in level l charging or 45 PEVs in level 2 charging to reach the first current limit. • For PSD-Right, around 138 kW, 300 kW, 1000 kW more demand can be attached, in order to reach the first, second, and third limit, respectively. It is equivalent to 69 PEVs in level 1 charging or 17 PEVs in level 2 charging, in order to reach the first current limit at transformer attached ahead of the load bus. • For PSX, nearly 400 kW, 946 kW, 4550 kW more demand can be added to reach the first, second, and third limit, respectively. It is equivalent to 200 PEV s in level l charging or 50 PEVs in level 2 charging, in order to reach the first current limit at transformer 60 USC Viterbi School of Engineering PhD Thesis before the load bus. The result of Case 2 reflects that in this test strategy, the system will reach its first current limit when all 6 load buses increase their kW demand by 10% while PSX increases 20%, and the current limit will occur at the transformer attached ahead of PSX. The results show that almost 855 kW additional demand can be added to the whole system to reach the first limit, which is equivalent to 427 PHEVs in level 1 charging or 106 PHEVs in level 2 charging. The result of Case 3 illustrates that in this test strategy, the system will increase all the 6 load buses' kW demand by 100% each time with respect to themselves, and the first current limit will occur when all the load buses increase by 300% at the transformer attached ahead of PSX, while the second and third current limit will occur when all the load buses increase by 400% and 600% at the transformer attached ahead of PSA and PSD-Left, respectively. The fourth and fifth current limit will occur at 700% increment at PSB and PSD-Right. The last limit is the voltage limit will occur when all the load buses increase 3720% at load bus PSA. The results also show that 1265 kW additional demand can be attached to the whole system to reach the first limit, which is equivalent to 632 PHEVs in level 1 charging or 158 PHEVs in level 2 charging. 2.5.2.2 OpenDSS Results and Conclusions The USC micro-grid circuit was modeled in OpenDSS software. This allowed for practical simulation of the EV impact on USC micro-grid. Different cases and scenarios were analyzed to find out the system capabilities and limitations. After that, the maximum possible additional EV charging and discharging kilowatt loads were calculated for each parking structure over the course of a complete 24 hours. This allowed for estimating the maximum possible connected number of EVs whether being charged by the grid or providing power to the grid. Cost projection was also estimated using available time of use rates to give an idea of the charging or discharging preference each hour. 2.5.3 Chatsworth Distribution System 2.5.3.1 Chatsworth Circuit 88-22 The result of Case 1 shows that: • For PSA, around 500 kW, 1300 kW, 5600 kW more demand can be attached, in order to reach the first, second, and third limit, respectively. It is equivalent to 250 PEVs in level 1 charging or 62 PEVs in level 2 charging, in order to reach the first current limit at transformer attached ahead of the load bus. 61 USC Viterbi School of Engineering PhD Thesis • For PSB, about 650 kW and 3100 kW more demand can be added, in order to reach the first and second limit, respectively. It is equivalent to 325 PEVs in level 1 charging or 81 PEVs in level 2 charging to reach the first current limit at transformer. • For PSD-Left, almost 360 kW, 950 kW, 4600 kW more demand can be put to reach the first, second, and third limit, respectively. It is equivalent to 180 PEV s in level 1 charging or 45 PEVs in level 2 charging to reach the first current limit. • For PSD-Right, around 138 kW, 300 kW, 1000 kW more demand can be attached, in order to reach the first, second, and third limit, respectively. It is equivalent to 69 PEVs in level 1 charging or 17 PEVs in level 2 charging, in order to reach the first current limit at transformer attached ahead of the load bus. • For PSX, nearly 400 kW, 946 kW, 4550 kW more demand can be added to reach the first, second, and third limit, respectively. It is equivalent to 200 PEVs in level 1 charging or 50 PEVs in level 2 charging, in order to reach the first current limit at transformer before the load bus. The result of Case 2 reflects: • In this test strategy, the system will reach its first current limit when all 6 load buses increase their kW demand by 10% while PSX increases 20%, and the current limit will occur at the transformer attached ahead of PSX. The results show that almost 855 kW additional demand can be added to the whole system to reach the first limit, which is equivalent to 427 PHEVs in level 1 charging or 106 PHEVs in level 2 charging. The result of Case 3 suggests: • In this test strategy, the system will increase all the 6 load buses' kW demand by 100% each time with respect to themselves, and the first current limit will occur when all the load buses increase by 300% at the transformer attached ahead of PSX, while the second and third current limit will occur when all the load buses increase by 400% and 600% at the transformer attached ahead of PSA and PSD-Left, respectively. The fourth and fifth current limit will occur at 700% increment at PSB and PSD-Right. The last limit is the voltage limit will occur when all the load buses increase 3720% at load bus PSA. The results also show that 1265 kW additional demand can be handled to the whole system to reach the first limit, which is equivalent to 632 PHEVs in level 1 charging or 158 PHEV s in level 2 charging. • When testing at 55% and 30% loading capability, system can withstand a few more load demand before reach those limits. 62 USC Viterbi School of Engineering PhD Thesis 2.5.3.2 Chatsworth Circuit 88-23 The result of Case 1 shows that: • When testing at peak load condition (period 1), 15 out of the all 26 loads can be attached less than or equal to 100 kW demand, which is equivalent to 50 PHEVs in level 1 charging or 12 PHEVs in level 2 charging, in order to reach the first current limit at transformer attached ahead of the load buses. • To reach the second current limit, which is 225% of the transformer current, these 15 load buses can be attached less than or equal to 225 kW demand, which is equivalent to 112 PHEVs in level 1 charging or 28 PHEVs in level 2 charging. • To reach the third voltage limit, which is at the load bus itself, these 15 load buses can be attached less than or equal to 580 kW demand, which is equivalent to 290 PHEVs in level 1 charging or 72 PHEVs in level 2 charging. • The highest demand came at load bus 86, which is at the far end. For load bus 86, 320 kW, 752 kW, and 1325 kW can be added to reach the first, second, and third limit, respectively. • The test results under intermediate load (period 2) and off-peak load (period 3) did not make much difference compared with peak load condition, as shown in Fig 6. The only big difference is the demand at the far end - load bus 86. In Case 1.2, it can be added 350 kW, 780 kW, and 1420 kW to reach the first, second, and third limit, respectively. While in Case 1.3, it can be attached 375 kW, 1149 kW, and 2180 kW to reach three limits, respectively. The result of Case 2 shows that: • Under period 1 (peak load), the system will reach its first current limit when 24 out of the all 26 load buses increase their kW demand by 10%, and the current limit will occur at the transmission transformer and voltage regulator at the substation side of circuit 88- 22. The results also show that when increasing EV demand of all the 26 load buses by 15%, the system will reach the voltage limit at load bus 04. The results show that almost 1400 kW additional demand can be added to the whole system to reach the first limit, which is equivalent to 700 PHEVs in level 1 charging or 175 PHEVs in level 2 charging. • Under period 2 (intermediate load), the system will reach its first current limit when 17 out of the all 26 load buses increase their kW demand by 15%, and the current limit will occur at the voltage regulator near the substation side of circuit 88-22. The results also show that when increasing kW demand of all the 26 load buses by 20%, the system will reach the voltage limit at load bus 904 with 108V. This implies that almost 1500 kW additional demand can be handled to the whole system to reach the first limit, which is equivalent to 750 PHEVs in level 1 charging or 187 PHEVs in level 2 charging. 63 USC Viterbi School of Engineering PhD Thesis • Under period 3 (off-peak load), the system will reach its first current limit when 14 out of the all 26 load buses increase their kW demand by 20%, and the current limit will occur at the voltage regulator near the substation side of circuit 88-22. At the same time, the system will also reach the voltage limit at load bus 04with109.6V. The results imply that 1624 kW additional demand can be added to the whole system to reach the first limit, which is equivalent to 812 PHEVs in level 1 charging or 203 PHEVs in level 2 charging. The result of Case 3 shows that: • Under the period 1 (peak load), the system will increase all the 26 load buses' kW demand by 50% each time with respect to themselves, and the first current limit will occur when all the load buses increase by 250% at the voltage regulator near substation, while the second current limit will occur when all the load buses increase by 300% at the transmission transformer near substation. The third limit is the voltage limit will occur when all the load buses increase 550% at load bus 70. The results also show that 1439 kW additional demand can be added to the whole system to reach the first limit, which is equivalent to 729 PHEVs in level 1 or 179 PHEVs in level 2 charging. While under period 2 and period 3, the testing results are almost the same. In these two cases, the system will increase all the 26 load buses' kW demand by 100% each time with respect to themselves, and the first current limit will occur when all the load buses increase by 300% at the voltage regulator near substation. The second limit is the voltage limit will occur when all the load buses increase 700% at load bus 70. The results show that 1724 kW more can be increased to the whole system to reach the first limit, equivalent to 862 PHEVs in level 1 charging or 215 PHEVs in level 2 charging. The result analysis for Chatsworth circuit 88-23 is similar to what have been done for 88- 22 above, therefore no need to do it one more time. 64 USC Viterbi School of Engineering PhD Thesis 3. CHAPTER 3: Distribution Effect of Battery Aggregation and Backfill Coupled with Renewable Energy 3.1 Research Overview There are three areas of consideration that are in this chapter: Battery Aggregation and Backfill coupled with Renewable Energy. Based on the SGRDP design document [50], they are: a) Use of EV Batteries in the Grid (G2V only); b) Use of EV Batteries in the Grid (G2V & V2G); and c) Distribution Effects of Battery Aggregation and Backfill (G2B, B2V, B2G). These three areas of considerations are discussed in detail below: a) Use of Batteries in the Grid (G2V Only) Charging aggregated EV s can result in the overload of transformers at the distribution level. Each EV consumes from two to ten times the amount of power of a single electric home during charging. Charge rates vary depending on levels of charging. Considering public adoption of Level 2 charging which requires 220 V and greater than 50 A of current [56], the amount of power that a single EV can possibly consume is very high. The challenge involved with EV integration is to balance that load when a large number of EVs need charging and working with the vehicles' onboard BMS to handle charging given that charging current varies nonlinearly with time and percent of charge. It is also important when considering an aggregate scenario to take into account consumer behavior. Periods of time exist during the day where EV loads could be heavy and optimization of charging schedules could limit issues created by severe loading. Identifying active charging stations and monitoring State of Charge (SOC) will allow for greater flexibility of charging schedules. The control center system will take into account various criteria to tum on or off a charger to ensure safety and grid needs are met. The EV Project team will install appropriate grid-tie inverters and bi-directional AMI smart meters at certain charging stations to demonstrate Vehicle-to-Grid (V2G) and Grid to Vehicle (G2V) operations. Depending on the make and model of the EV, battery backfill may not be possible without modifying the BMS and other circuits on the EV. The EV Project team will study and investigate a non-destructive approach to implement this 65 USC Viterbi School of Engineering PhD Thesis research. All EVs that are participating in the V2G demonstration may need to be equipped with additional circuits and modules. b) Use of Batteries into the Grid (G2V & V2G) Depending on the make and model of the EV, battery backfill may not be possible without modifying the battery management system (BMS) and other circuits on the EV. Introducing backfill of EV charge, or V2G discharge, has the potential to introduce new issues for the both EVs and the power system. Considering that existing EVs may not have a BMS for the reversal of charge flow, precautions must be taken to monitor and control the rate of discharge from EVs to ensure maximum battery lifetime. Taking into account a range of variables, such as time of day and battery status of charge, optimization algorithms should be developed to charge vehicles according to consumer scheduling, utility grid management, and battery safety. Simulations will be performed in a controlled environment in order to implement the algorithms to ensure safety and quality control. c) Effect of Battery Aggregation and Backfill on Distribution (G2B, B2V and B2G) Three different battery operational modes will be examined. Data collection intervals for each scenario are also listed. Current, voltage, real and reactive power, frequency and harmonic data from the charging stations will be used to model power flows on the distribution system for each of the scenarios listed below: 1. Stationary Battery charging from the grid; typically at night (G2B). 2. Stationary Battery charging EVs; morning through evening hours (B2V). 3. Stationary Battery back-feeding into the grid; peak times, no EV charging (B2G). The study results reported in this chapter are based on the assumptions and data discussed in the thesis and are conducted using MATLAB, PSCAD and EDD analysis models. Due to lack of real-world data on EV charging and discharging, only simulation results are available for this part of research. Several charging/discharging scenarios have been designed to represent G2V, G2B, B2V, B2G and V2G. They are listed as follows: • Different Scenarios for Battery Aggregation and Backfill • Scenario 1: Base Case (G2V and V2G) • Scenario 2: Simple Case (G2V and V2G) 66 USC Viterbi School of Engineering PhD Thesis • Scenario 3: High Capacity (G2V and V2G) • Scenario 4: Full Peak Shaving (G2V, G2B, B2G and B2V) • Different Scenarios for Battery Aggregation and Backfill Coupled with PV Panels • Case 1: G2V charging at period 2 (55% loading) with 30% Capacity PV • Case 2: G2V charging at period 2 (55% loading) with 55% Capacity PV • Case 3: G2V charging at period 2 (55% loading) with 100% Capacity PV • Case 4: V2G charging at period 1 (100% loading) with 30% Capacity PV • Case 5: V2G charging at period 1 (100% loading) with 55% Capacity PV • Case 6: V2G charging at period 1 (100% loading) with 100% Capacity PV • Battery Aggregation and Backfill Strategies in MATLAB Modeling • EV Coupled with Storage in PSCAD, power quality and harmonics 3.2 Overview of the Systems Utilized 3.2.1 Technologies and Systems Used To meet the general system modeling and representation requirements outline in the SGRDP design document [50], the following activities are under taken by the author: 1. Two sets of distribution network models are selected to enable analysis of EV distribution effects as described in the SGRDP Preliminary Design Document [50]. Furthermore, four software programs are selected for the development of these models. These software programs are: a. Distribution system modeling software Distribution Engineering Workstation (DEW) by Electrical Distribution Design, Inc. (EDD) is used to model the distribution network to assess the impact of EV, community energy storage systems (CES), and distributed generation (DG) such as photovoltaic (PV) on the grid with regard to system load flow, and load factor including the fully functional Micro-grid. EDD is specialized in doing power system analysis on distributed level systems but has limited capability in running harmonic and real-time analysis. Power flow analysis and load factor results are expected from EDD. b. Mathematical modeling and analysis tool MATLAB and Microsoft Excel are used to model and analyze various charging and discharging strategies associated with EV, CES, PV, and the Grid. These software are expected to perform designed charging/discharging algorithms of batteries. 67 USC Viterbi School of Engineering PhD Thesis c. Electromagnetic time domain transient simulation software Power System Computer Aided Design (PSCAD) by Manitoba Hydro is used to model high-frequency transients to assess the impact of multiple Level 2 EV chargers on the electric network. PSCAD will be used to perform power quality analysis for IEEE 34 system. The university version of PS CAD has limited buses restrictions and cannot be used to build Chatsworth system. Instead, IEEE 34 node test feeder's results will be used. 2. Since EDD is a newly developed software and not many tests have been made using this software, two distribution network models are selected to compare with the benchmark results and have a more reliable analysis for the Chatsworth system. These two systems are: a. An IEEE 34 node test feeder reflecting an industry standard distribution test model b. A Chatsworth Distribution System or Micro-grid representing a residential/industrial area in the LAD WP system with a special focus to the circuit numbers 22 and 23 3. The EDD, MATLAB, and PSCAD models and the two system representations (IEEE 34 bus and Chatsworth) allow the attachment of multiple EV-charger loads of various classifications 4. A harmonic and sensitivity analysis are setup on the test systems with the goal of revealing the substations within the LAD WP territory that have the greatest impact on the transmission grid when they are similarly subjected to EV loads. 3.2.2 Configurations of the Smart Grid Systems, Subsystems and Components The network configurations of the test systems and the system modeling considerations are discussed in this subsection. The actual system studies test results are provided in Section 3.4. The two test system selected in this chapter are: • IEEE 34 node test feeder • Chatsworth Distribution system Their configurations have already been discussed and can be found in Section 2.2.2. 68 USC Viterbi School of Engineering PhD Thesis 3.3 Description of the Methodologies and Algorithms Utilized As described in Section 3 .2, two system representations (IEEE 34 bus and Chatsworth) and four types of study programs (EDD, EXCEL, MATLAB and PSCAD) are utilized to address the key tasks of this part of the research. Due to the nature of the research activities and the work assignments, the study tasks are arranged, as follows: 1. EV Charging/Discharging Strategy: Charging/Discharging Strategy EV, CES & Grid and system interaction of EVs, CES, and grid during charging/discharging 2. V2G and G2V strategies in MATLAB and Microsoft Excel 3. Evaluation of EV loading on the distribution grid: Impact to the grid that has various loading scenarios (e.g. cable systems, voltage regulators, DG systems) Each of these tasks, the methodology utilized, and algorithms developed are discussed in the following subsections. 3.3.1 Demand Load Level The simplified three daily load periods are the same as described in Section 2.3 .1. 3.3.2 EV Charging/Discharging Strategies In this section, charging/discharging strategy of EV, CES, and the grid, and the system interaction of EVs, CES, and grid during charging/discharging are studied. The impact of distribution generation, photovoltaic applications in this case, is also assessed. The study objectives, scope, and the overall study process are discussed in the following subsections. 3.3.2.1 Study Objectives The various charging and discharging objectives are outlined below: 1. Grid-to-vehicle charging (G2V), typically at night 2. Stationary Battery charging from the grid (G2B), typically at night 3. Stationary Battery charging EVs (B2V), morning through evening hours 4. Stationary Battery back-feeding into the grid (B2G), peak times, no EV charging 5. Interaction of photovoltaic (PV), a renewable distributed generation, with the above charging/discharging considerations 69 USC Viterbi School of Engineering PhD Thesis 3.3.2.2 Study Scope In order to fulfill the requirements of battery aggregation and backfill, as well as V2G and G2V charging/discharging, extensive literature review was conducted as part of this study to review the EV technology and the overall EV charging/discharging philosophy and technical capabilities. Based on those inputs and the requirements of the study objectives, the overall scope of work of this task is described below: • Use the distributed generation, renewable energy resources (i.e. photovoltaic), Plug-in Electric Vehicles (PEV), and energy storage devices (i.e. community battery) to modify the power daily load curve. • The goal is to level the daily load curve in order to minimize the total cost of power. Also, reduce the C02 emissions by using the PEV s. 3.3.3 Battery Aggregation and Backfill Strategies in Microsoft Excel 3.3.3.1 Assumptions According to the data from LADWP, the average monthly electricity consumption for a U.S. residential utility customer in Los Angeles is approximately 500 kWh per month. In this approach, the average power is assumed to be 5 kW per house. IEEE 34 Bus has total power of 2 MW, then there will be 2 MW -7- 5 kW = 400 houses as the loads attached to IEEE 34 Bus System. Assume that level 2 charging has capacity of 8 kW (32A) at 240V with 6 hours to be fully charged. In the V2G technology, the following assumptions have been made: • People will drive to woik from 6-8 AM during which EV users can neither charge from the grid nor feed to the grid. • People will drive back home from 5-8 PM during which EV users can neither charge from the grid nor feed to the grid. • Before driving to work (6 AM) and back home (8 PM), EV users will have their EVs fully charged at 100% battery. • Each trip, from home to work place, or from work place to home will use about 50% of the electricity, which is equivalent to 50% remaining battery. • People will not use their EV s during the work, from 8 AM to 5 PM. • As soon as people arrived at work place or home, they will plug in their EV s ready for G2VorV2G. 70 USC Viterbi School of Engineering PhD Thesis 3.3.3.2 Different Scenarios for Battery Aggregation and Backfill 3.3.3.2.1 Scenario 1: Base Case (G2V and V2G) Scenario 1 for battery aggregation and backfill is described as follows: 1. EVs will be charged by the grid (G2V) from 12 AM to 6 AM, and will be fully charged before 6 AM. 2. From 6 AM - 8 AM, and 5 PM - 8 PM, people will only drive EVs to work and back home without using any type of charging. 3. During some emergency peak hours when utility needs to spend huge extra money to purchase the additional generation power to satisfy the customer needs while making the generation components running at super low power factor (about 5%), especially in hottest summer days in Los Angeles, utility companies will send out signals that allow EV owners to charge to the grid (V2G) to reduce the peak demand and compensate those customers with some credits. 3.3.3.2.2 Scenario 2: Simple Case (G2V and V2G) Scenario 2 for battery aggregation and backfill is described as follows: 1. EVs will be charged by the grid (G2V) from 12 AM to 6AM, and will be fully charged before 6 AM. 2. From 6-8 AM, people will drive EVs to work so that EV users cannot use G2V or V2G. 3. During 12 PM to 5 PM, all the EVs will become available to charge to the grid (V2G), and will have minimum 50% battery remained at 5 PM. Up to 20% of battery's electricity is used for V2G. 4. From 5 PM to 8 PM, people will drive EVs back home so that no G2Vor V2G can be used. 5. During 8 PM to 12 AM, all the EVs won't be able to receive any type of charging. 6. At 12 AM, all the EVs will be plugged in at home with almost zero battery remaining in the EVs. 3.3.3.2.3 Scenario 3: High Capacity (G2V and V2G) Scenario 3 for battery aggregation and backfill is described as follows: 1. EV s will be charged by the grid ( G2V) from 12 AM to 6 AM, and will be fully charged before 6 AM. 71 USC Viterbi School of Engineering PhD Thesis 2. From 6-8 AM, people will drive EVs to work so that EV users cannot use G2V or V2G. 3. At 8 AM, all the EVs will be plugged in at the parking space with 50% of the battery remaining in the EVs. 4. From 8 AM to 12 PM, all the EVs will be charged by the grid (G2V), and will be fully charged before 12 PM. Assume currently there will be enough chargers and advanced technology ready for EVs. 5. From 12 PM to 5 PM, all the EVs will become available to feedback to the grid (V2G), and will have minimum 50% battery remained at 5 PM. 6. From 5 PM to 8 PM, people will drive EVs to home so that no G2V or V2G can be used. 7. During 8 PM to 12AM, all the EVs won't do any type of charging. 8. At 12 AM, all the EVs will be plugged in at home with minimum 10% of the battery remaining in the EV s. 3.3.3.2.4 Scenario 4: Full Peak Shaving (G2V, G2B, B2G and B2V) Scenario 4 for battery aggregation and backfill is described as follows: 1. EV s will be charged by the grid ( G2V) from 12 AM to 6 AM, and will be fully charged before 6 AM. 2. From 6-8 AM, a community battery will be charged (G2B) from the grid. 3. At 8 AM, all the EVs will be plugged in at the parking space with 50% of the battery remaining in the EVs. 4. From 8 AM to 12 PM, all the EVs will be charged by the grid (G2V), and will be fully charged before 12 PM. Assume currently there will be enough chargers and advanced technology ready for EVs. 5. From 12 PM to 5 PM, all the EVs will become available to support the grid (V2G), and will remain minimum 50% battery at 5 PM. 6. From 5 PM to 8 PM, the grid will use the community battery as an energy source to shave the peak (B2G). 7. At 8 PM, all the EVs will be plugged in at home or to the stationary battery with minimum 10% of the battery remaining in the EVs. (B2V) 72 3.3.3.3 USC Viterbi School of Engineering PhD Thesis Different Scenarios for Battery Aggregation and Backfill Coupled with PV Panels Assume PV will receive sunlight and will be working from 8 AM to 5 PM daily. From 8 AM to 9 AM and 4 PM to 5 PM, the sunlight is weak and PV will only produce 30% of its peak power. During 9AM to 11 AM and 2 PM to 4 PM, PV will produce 55% of the peak power. From 11 AM to 2 PM, PV will working at its highest rated power. Figure 3 .1 shows the actual and simplified daily PV output curve. PV Output as the Seasons Change 3000 2500 :'5! 2000 ... Cll 0 .. ~ 1500 "' ~ ~ 1000 500 0 5:00 7 :00 9:00 11:00 13:00 15:00 17:00 19:00 S implified PV Output Curve 100 80 60 40 20 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Figure 3.1 Daily PV Output Curve: Actual (upper) and Simplified (lower) Six G2V/V2G scenarios (Cases) coupled with PV have been analyzed. These cases are defined as follows: 73 USC Viterbi School of Engineering PhD Thesis 3.3.3.3.1 Case 1: G2V charging at period 2 (55% loading condition) with 30% PV It is assumed that PV solar panels would have a total capacity of 15% of the total circuit capacity. PV panels have been modelled in EDD by adding inverter at each load with capacity equal to 15% of the load itself. Also, each PV panel was multiplied by 30% of the total power because it is currently loading under period 3 (30%). A detailed list with all the PV panels added is shown below in Table 3 .1. Table 3.1 PV Power Added at Each Load (Period 3) Phl Power Ph2 Power Ph3 Power Load (Name in DEW) (KW) (KW) (KW) 147701 0.00 0.00 0.23 147702 0.23 0.00 0.00 147709 0.23 0.23 0.23 147704 0.00 0.14 0.00 147718 0.13 0.19 0.13 147721 0.25 0.25 0.25 147719 0.45 0.45 0.45 147720 0.23 0.45 0.23 147729 0.23 0.23 0.23 147730 0.45 0.45 0.45 147731 0.89 0.89 0.89 147714 0.19 0.19 0.19 147745 0.08 0.08 0.08 147746 0.13 0.13 0.13 147749 0.67 0.67 0.67 147747 0.13 0.25 0.13 147756 0.45 0.45 0.45 147757 0.19 0.51 0.19 147758 0.51 0.51 0.51 147767 0.45 0.45 0.45 147772 0.85 0.51 0.51 147768 0.13 0.38 0.13 147769 0.38 0.38 0.38 147770 0.13 0.38 0.13 147771 0.25 0.25 0.25 147786 0.89 0.89 0.89 74 USC Viterbi School of Engineering PhD Thesis 3.3.3.3.2 Case 2: G2V charging at period 2 (55% loading condition) with 55% PV In this case, the total power added is 15% of the total system capacity multiplied by 55%. It has been assumed that PV panels are loading under period 2 in this scenario. A detailed list of the added PV panels are listed in Table 3.2. Table 3.2 PV Power Added at Each Load (Period 2) Phl Power Ph2Power Ph3Power Load (Name in DEW) (KW) (KW) (KW) 147701 0.00 0.00 0.41 147702 0.41 0.00 0.00 147709 0.41 0.41 0.41 147704 0.00 0.25 0.00 147718 0.23 0.35 0.23 147721 0.46 0.46 0.46 147719 0.82 0.82 0.82 147720 0.41 0.82 0.41 147729 0.41 0.41 0.41 147730 0.82 0.82 0.82 147731 1.63 1.63 1.63 147714 0.35 0.35 0.35 147745 0.14 0.14 0.14 147746 0.23 0.23 0.23 147749 1.23 1.23 1.23 147747 0.23 0.46 0.23 147756 0.82 0.82 0.82 147757 0.35 0.93 0.35 147758 0.93 0.93 0.93 147767 0.82 0.82 0.82 147772 1.55 0.93 0.93 147768 0.23 0.70 0.23 147769 0.70 0.70 0.70 147770 0.23 0.70 0.23 147771 0.46 0.46 0.46 147786 1.63 1.63 1.63 75 USC Viterbi School of Engineering PhD Thesis 3.3.3.3.3 Case 3: G2V charging at period 2 (55% loading condition) with 100% PV In this case, the PV will be in period 1 (100%) and the system will have 55% loading. The list of the PV panels added is provided in Table 3.3. Table 3.3 PV Power Added at Each Load (Period 1) Phl Power Ph2Power Ph3Power Load (Name in DEW) (KW) (KW) (KW) 147701 0.00 0.00 0.75 147702 0.75 0.00 0.00 147709 0.75 0.75 0.75 147704 0.00 0.45 0.00 147718 0.42 0.63 0.42 147721 0.84 0.84 0.84 147719 1.49 1.49 1.49 147720 0.75 1.49 0.75 147729 0.75 0.75 0.75 147730 1.49 1.49 1.49 147731 2.97 2.97 2.97 147714 0.63 0.63 0.63 147745 0.26 0.26 0.26 147746 0.42 0.42 0.42 147749 2.24 2.24 2.24 147747 0.42 0.84 0.42 147756 1.49 1.49 1.49 147757 0.63 1.70 0.63 147758 1.70 1.70 1.70 147767 1.49 1.49 1.49 147772 2.82 1.70 1.70 147768 0.42 1.28 0.42 147769 1.28 1.28 1.28 147770 0.42 1.28 0.42 147771 0.84 0.84 0.84 147786 2.97 2.97 2.97 3.3.3.3.4 Case 4: V2G charging at period 1 (100% loading condition) with 30% PV The PVs added in this case is similar to case 3. However, the circuit here has 100% loading and the EVs are discharging to provide power to the system. The PVs added in this case can be found in Table 3.3. 76 USC Viterbi School of Engineering PhD Thesis 3.3.3.3.5 Case 5: V2G charging at period 1 (100% loading condition) with 55% PV The circuit has the same amount of PVs as case 2. In this case the circuit has 100% loading and the cars are discharging to reduce the peak load using V2G concept. The list of all added PV is provided in Table 3.2. 3.3.3.3.6 Case 6: V2G charging at period 1(100% loading condition) with 100% PV The PV s added in this scenario is similar to case 1. The system in this case has 100% loading capacity with EVs charged from the grid. 3.3.4 Battery Aggregation and Backfill Strategies in MATLAB Modeling MATLAB will be using the input data of the daily power curve to find the best way to flat the power curve. The goal is to make the instantaneous power consumption is equal to the average daily power. Also, the algorithm will show the change of power price on hourly basis. 3.3.4.1 Variables in MATLAB Modeling Table 3 .4 below lists the necessary variables used in MATLAB modeling. Table 3.4 Variables in MATLAB Modeling Variable Description N Number ofEVs CL Charging Level (KW) TC Travel Consumption per Trip CCPl Percentage of EV s Charged 8AM-12PM CCP2 Percentage of EVs Charged 8PM-12AM CDP Percentage ofEVs Discharged 12PM-5PM BCL Battery Charging Level (KW) BCR Battery Charging Rate 6AM-8AM BDR Battery Discharging Rate 5PM-8PM pp Load Curve Peak Power (KW) BS Capacity of the EV battery (KWh) CPI Cost at Peak Load (¢/KWh) CP2 Cost at Intermediate Load (¢/KWh) CP3 Cost at Base Load (¢/KWh) 77 USC Viterbi School of Engineering PhD Thesis 3.3.4.2 MATLAB Modeling Algorithm Description Several assumptions have been made regarding the modeling. These assumptions are: 1) The power curve is divided into three periods: Peak-100% (12PM-8PM), Intermediate- 55% (8AM-12PM & 8PM-12AM), and base-30% (12AM-8AM). 2) The EVs will be charged during the periods (0-6AM, 8AM-12PM, and 8PM-12AM) 3) The EVs will be discharged during the peak load (12PM-5PM) 4) A community battery will be used to provide power to the systems during the time that people will go from work to home (5-8PM). This Battery can be charged during the time that people go from home to work (6-8AM). 5) It is assumed that half of the car battery will be consumed per trip. 6) The cost of energy per kWh is: 7 cents/kWh (base), 9 cents/kWh (intermediate), and 14 cents/kWh (peak). The program will allow the utility to enter the values of peak power of the power system, the capacity of the battery of EV, and the number of EVs. Based on this information, the algorithm will try to find the best fit of power curve compared with the average power. The average power of the system load is equal to the sum of the average power of the three periods and the average power consumed by EVs which is equal to • Average power = Average Power of 3 periods + Average power consumed by EV s • Average Power of the three periods= Peak power x (1+0.55 + 0.3) I 3 • Average power consumed by EVs =(Battery size x Number ofEVs) I 24 The EVs will be charged during the periods (12AM-6AM, 8AM-12PM, and 8PM-12AM). The three periods have different power loadings. Base load during the time ( 12AM-6AM) and intermediate loading during the time (8AM-12PM and 8PM-12AM), the total energy needed to charge the EV s is equal to the energy needed to fill the gap between base and intermediate power loadings and average power loading. • Total available energy= Energy (0-6AM) +Energy (8AM-12PM and 8PM-12AM) • Energy (0-6AM) = 6 x (Average power - 0.3 x peak power) • Energy (8AM-12PM and 8PM-12AM) = 8 x (Average power - 0.55 x peak power) Total available energy will be used to charge the EVs and provide enough power to the system during peak load (12PM-5PM). 78 USC Viterbi School of Engineering PhD Thesis During the time people go and come back from work, the community battery will be charged with the energy difference between the base load power and average power. Due to the difference between the stored energy during (6-8AM) and discharged energy during (5-8PM), there might be more or less than required energy during the discharging period. 3.3.5 EV Coupled with Storage in PSCAD The aim of this loading scenario is to study the potential mitigation of infrastructure requirements and economic benefit when battery aggregation and backfill processes are locally integrated with EVs. 3.3.5.1 Assumption Battery storage is added on bus 840, Figure 3.2. The output power of the battery storage should be around 500kW. The accurate value will be decided after the Battery Storage model is tested. Again, 5 EVs will be attached to bus 840 one by one. Power and harmonic changes on bus 800, 834, 840 will be collected for analysis. 844 I 864 842 860 836 834 840 ~ • • 86 2 888 890 'fs s2 838 Figure 3.2 Location for Battery Storage Attachment The Battery Storage model used in this scenario is almost the same as the EV model. The difference is more batteries will be integrated in one model so that the Storage model can keep providing power to the grid, instead of charging and discharging periodically. 79 USC Viterbi School of Engineering PhD Thesis a) Battery model lbatt < 0.1~18 [ohm) V~t Rese _bat 13·ph<1seln~rterBridge Figure 3.3 Battery Model Reseived energy in this Battery Storage model is set to 0.99Ah and the number of series and parallel batteries are set to 160, as shown in Figure 3 .4. [Battery] I Battery _paramter .::J reated capacity (Ah) 1 [Ah] reserved battery (Ah) 0.99 [Ah] nominal capacity (Ah) 0.85 Maximum voltage (V) 4.2 [V] voltage at exponential point (V) 3.9 [V] charge current (A) -0.5 [A] effeciency 0.995 number of seriesed battery 160 number of parallelled battery 160 nominal voltage (V) 13.6 [VJ OK Cancel Help ... Figure 3.4 Battery Storage Model Parameters 80 USC Viterbi School of Engineering PhD Thesis b) Three Phase Inverter (PI) 1. Simple P and Q Regulation The new Battery Storage has 160 batteries in series and 160 * 3.6V = 576V. In the new PI controller, Figure 3.5, battery voltage Vbat will be set to 0.58. The output of the controller will be used as an input to the firing pulse generator that will be discussed next. Vb at D 0.58 Figure 3.5 PI Controller The second PI controller sets the reactive power of the grid to zero which force the inverter to operate at utility power factor so that it produces sinusoidal voltage and current which are in phase. The output of this controller will be also used as an input to the firing pulse generator. 11. Firing Pulse Generation The switching signals of the 6 IGBT switches of the 3-legged inverter are generated using a SPWM technique shown below, as shown in Figure 3.6. It starts with creating three sinusoidal modulating waves with a frequency of 60 Hz and a phase shift equal to the output of the previous PI controller with additional shifting of -120 and 120 degrees. The magnitude of the modulating wave is equal to the output from the previous PI controller. Then, the three sinusoidal modulating waves are compared with a triangular carrier wave with magnitude between -1 and 1. Switching signals gtl, gt3, gt5 are generated by setting the output of the comparator to 1 whenever the modulating wave is greater than the carrier wave and 0 otherwise. Since the operation of the two switches in each of the three legs should be complementary to produce the final sinusoidal wave, the switching signals gt4, gt6, gt2 are generated by inverting the switching signals gtl, gt3, gt5 respectively. 81 Aflg USC Viterbi Mag Freq t !------____..,.___~ Freq School of Engineering PhD Thesis ..--~~~~---~ A ~~____, Com _par ator ~~~~~---.-~ A ~~____, Com _par ator g~4 g~6 g~ Figure 3.6 SPWM Technique to Generate Switching Signal 111. Three Phase Inverter Bridge By applying the previously generated switching signals to the 6 IGBT switches, the inverter keeps its input DC voltages at a constant value of 0.5 kV and converts this DC voltage to an AC voltage, as shown in Figure 3.7. 82 USC Viterbi School of Engineering PhD Thesis -----.___.=a 1 ~1 la_motor Ee Eb Ea Figure 3. 7 Three Phase IGBT Bridge 3.4 Summaries of the results of the Performance of the Smart Grid systems and technologies derived from lab tests, field tests, or grid-connected applications. The study results related to the system studies are presented in this section. 3.4.1 EV Charging/Discharging Strategy The conventional charging condition for EV is what is called G2V. This is where the electric vehicle battery is charged directly from the electric power grid. There are verities of G2V charging strategies discussed in the literature. The strategy mainly depends on the economic and technical objectives of the EV program and the technology that can be deployed to support the program. Based on the analysis of these charging strategies, the author has developed the G2V strategy that is considered to be the best strategy for LADWP. This strategy takes into consideration distribution system issues and limitations, time-of-use pricing, as well as demand-side management (DSM) objectives. This strategy is depicted in Figure 3 .8. 83 3.4.2 3.4.2.1 USC Viterbi ~ :Y!S -E- : No : Aalon box : Oeci~ion box ~ : One way flow Pr icing ~ / System Based Issues DSM / Syste m Based Issues School of Engineering ·[ Connect t o Grid Charging instr uction .... DS M ~ Safe Setting Exit ] PhD Thesis Charging by instruction System Issues Figure 3.8 Best V2G Charging Strategy Proposed for LADWP Battery Aggregation and Backfill Strategies in Microsoft Excel G2V, V2G, G2B and B2V Other charging strategies, generally considered "smart" charging are employed when a higher level objectives and advanced system capabilities and technology are deployed. This typically may include one or more of these considerations: V2G (vehicle-to-grid), G2B (grid-to-battery i.e. community energy storage), B2G (battery-to-grid), B2V (battery-to vehicle) and V2B (vehicle-to-battery). The objective of the studies is also to develop and test smart charging (also called charging/discharging strategy). Based on MATLAB modeling a smart charging strategy has been developed for the LAD WP system. This is depicted in the Figure 3. 9. 84 USC Viterbi School of Engineering PhD Thesis Load Curve (V2G Scenario 3 ) 2.50 2.00 '§" 1.50 ~ ~ .. :: ~ 1.00 - With EVs ..,._Without EV 0.50 0.00 0 5 10 15 20 25 Time of day (hour) Figure 3.9 Smart Charging Impact on Daily Load Curve The above developed strategy was tested using the IEEE 34 Node Distribution Test system as well as the Chatsworth Circuits 22 and 23. The study results are shown in Figure 3.10, Figure 3.11 and Figure 3.12 for the IEEE 34 test case, Chatsworth Circuit-22, and Chatsworth Circuit-23, respectively. As can be seen by these figures the charging/discharging strategy utilized is successful in leveling the system load profile thus effectively "shaving the peak and filling the valley." 3.4.2.2 G2V N2G Combined with PV Modeling Scenarios Test cases were also conducted to study the integration of PV into the smart charging strategy. The study result for the Chatsworth Circuits 22 is shown in Figure 3 .13. The effect of simply adding a PV to the above strategy is to simply reduce the peak load condition. Depending on the size of the PV, further integration of PV is possible to provide for a higher optimization of all of the components of the charging/discharging strategy. 85 3 :!. Cl. 2500 2000 1500 1000 500 0 2 3 4 USC Viterbi School of Engineering IEEE 34 Daily Load Curve 5 6 7 8 9 10 11 12 13 1 4 15 16 17 18 19 20 21 22 23 24 PhD Thesis ~·ith v~ tech - f 0 V2G tech Figure 3.10 24 Hour Charging Load Curve with and without V2G for IEEE 34 700 600 500 j' ~ 400 Cl. 300 100 Daily Load Curve 88-22 --WtthV2G -e-W/O V2G 3 4 s 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Tlme(hr) Figure 3.11 Hour Charging Load Curve with and without V2G for CW88-22 900 800 700 600 l soo 400 Cl. 300 200 100 0 Daily Load Curve 88-23 • • • l 2 3 4 5 6 7 8 9 JO ll U 13 14 JS 16 l7 18 19 20 2J 22 23 24 T1me (hr) Figure 3.12 24 Hour Charging Load Curve with and without V2G for CW88-23 86 USC Viterbi School of Engineering PhD Thesis Daily Load Curve 88-22 With PV - WV2G --W/ OV2G ~WV2G+PV 1 2 3 4 s 6 7 8 9 10 11 12 13 14 15 16 1 7 18 19 20 21 22 23 24 Figure 3.13 24 Hour Charging Load Curve with V2G and PV for CW88-22 3.4.3 Battery Aggregation and Backfill Strategies in MATLAB Modeling 3.4.3.1 Three Testing Cases with/without EV Three testing cases for daily load power cmve with and without EV have been done and the results can be found below, Figure 3.14. • When N=l 70, BS=48 kWh, PP=2000 kW, the resulting plot can be found in Figure 4.7 below. 3000r.=================.-------.~~~.------, - Power w ithout EV 2500 - Power with EV ........ .: .................. ~ ... ... . ---T-Average consumed power - 2000 ~ :ii=: - 1500 ... ~ ~ 0 Q. 1000 500 00 5 10 15 20 Time (hours) Figure 3.14 Resulting Plot in MATLAB Case 1 87 USC Viterbi School of Engineering PhD Thesis • When N=l 70, BS=48 kWh, PP=2500 kW, the resulting plot can be found in Figure 3.15 below. 3500r.================,-----------,~~~,--------, - Power without EV 3000 - Power with EV ~Average consumed power 2500 ~ 2000 ................. . - ... G.J s 1500 0 Q.. 1000 . . . . . . . . . . . . . . . . - . . . . . . . . . . . . . . . . . ................... ~ ............... ..... .... . . 500 ·················· ······································ 5 10 15 20 Time (hours) Figure 3.15 Resulting Plot in MATLAB Case 2 • When N=220, BS=48 kWh, PP=2000 kW, the resulting plot can be found in Figure 3.16 below. 3000r.==================;-~-,---~~---.--~--, - Power without EV 2500 - Power with EV ~Average consumed power - :5: ::.=:: - 1500 ... G.J == 0 Q.. 1000 500 00 5 10 15 20 lime (hours) Figure 3.16 Resulting Plot in MATLAB Case 3 88 USC Viterbi School of Engineering PhD Thesis 3.4.3.2 Economic Saving in MATLAB The economic savings have also been calculated and tested in MATLAB. The cost of kWh during different loading period are assumed $0.07, $0.09 and$ 0.14 for Period 1, Period 2, and Period 3, respectively. The average cost per kWh under this assumption would be $0.10/kWh without any EV. Two different cases after having EVs in the system have been run to find out the average cost per kWh, the results can be found below: • When peak power is fixed at 2MW and battery size 48kWh, the savings at different number of EVs are shown in Table 3.5. The resulting plot for different numbers of EV s attached into the system can be found in Figure 3 .17. Table 3.5 Economic Savings for Fixed System Peak Power in MATLAB Base Intermediate Cost no Cost with Capacity Number Peak Cost Savings Cost Cost EVs EVs (kW) ofEVs (c/kWh) (c/kWh) (c/kWh) (c/kWh) (c/kWh) (c/kWh) 50 7 9 14 11.4 10.6 0.7 100 7 9 14 11.4 10.3 1 150 7 9 14 11.4 10.1 1.3 200 7 9 14 11.4 9.9 1.5 2000 250 7 9 14 11.4 9.7 1.7 300 7 9 14 11.4 9.5 1.9 350 7 9 14 11.4 9.3 2.1 400 7 9 14 11.4 9.2 2.2 •• 0 0 0 ~~ ~~ 0.5 0 50 100 150 200 250 300 350 400 Number of E Vs Figure 3.17 Economic Savings Versus the Number of Electric Vehicles 89 USC Viterbi School of Engineering PhD Thesis • When number of EVs is fixed at 170 and battery size 48kWh, the savings at different power system capacities are shown in Table 3.6. The resulting plot for different peak load system can be found in Figure 3 .18. Number of EVs 170 Table 3.6 Economic Savings for Fixed Number ofEVs in MATLAB Base Intermediate Cost no Cost with Capacity Cost Cost Peak Cost EVs EVs (kW) (c/kWh) (c/kWh) (c/kWh) (c/kWh) (c/kWh) 1200 7 9 14 11.4 9.4 1400 7 9 14 11.4 9.6 1600 7 9 14 11.4 9.7 1800 7 9 14 11.4 9.9 2000 7 9 14 11.4 10 2200 7 9 14 11.4 10.1 2400 7 9 14 11.4 10.2 2600 7 9 14 11.4 10.3 2.5 2 0 0 ~ ~ 0 0 ~~ 0.5 0 1200 1400 1600 1800 2000 2200 2400 2600 Capacity (kW) Figure 3.18 Economic Savings Versus the Capacity of the Power System 3.4.4 EV Coupled with Storage in PSCAD Savings (c/kWh) 2 1.8 1.6 1.5 1.4 1.3 1.2 1.1 The test results for EV coupled with storage for IEEE 34 node distribution system in PSCAD can be found in this section. 90 USC Viterbi School of Engineering PhD Thesis The tables below, Table 3.7, shows the changes of active and reactive power at each bus when different numbers of EVs are plugged in. Table 3.7 Changes of P & Q at Selected bus when EVs are connected 1 1.4799 0.5309 -0.0893 0.5137 -0.4582 0.3062 2 1.6127 0.7049 -0.0028 0.6721 -0.3700 0.4626 3 1.7410 0.8863 0.0909 0.8204 -0.2796 0.6116 4 1.8649 1.0479 0.1765 0.9423 -0.1918 0.7348 5 2.0013 1.2114 0.2606 1.0664 -0.1049 0.8594 Instead of drawing a picture to show the variation trend, MATLAB is used to generate a regression equation for each case. The independent variable is EV number, dependent variable is P or Q. Then the 'coefficient of determination', denoted 'R 2 ', is designed to indicate how well the data points fit the equation, Table 3.8. Table 3.8 Variation Trend in MATLAB p y = 0.1295x + 1.3515 0.9998 BUS 800 Q y= 0.1704x + 0.3651 0.9995 p y = 0.0879x - 0.1766 0.9997 BUS 834 Q y = 0.1376x + 0.3903 0.9964 p y = 0.0885x - 0.5463 1 BUS 840 Q y = 0.1379x + 0.1813 0.9968 Table 3. 9 and Table 3 .10 present the harmonic results for this scenario. Table 3.9 Charging Starting Time Maximum Harmonics at Selected buses Charging Start Time max Harmonics bus 800 834 840 one EV 6.89E-03 2.88E-02 2.21E-01 two EV 8.23E-03 4.21E-02 3.09E-01 three EV 7.96E-03 3.36E-02 2.48E-01 four EV 1.24E-02 7.79E-02 5.17E-01 five EV 3.07E-02 1.0lE-01 6.02E-01 91 USC Viterbi School of Engineering PhD Thesis Table 3.10 Charging Steady State Average Harmonics at Selected buses Charging Steady State A\·ernge Harmonic bus 800 834 840 one EV T63E-04 2.88E-03 427E-02 two EV 8.94E-04 3.62E-03 5.96E-02 three EV 9.16E-04 5.0lE-03 6.82E-02 four EV 2.33E-03 5.97E-03 7.39E-02 five EV 3.09E-03 6.34E-03 9.03E-02 3.5 Summaries of the Results of the Analysis and Benefits In this section, summary of the study results is provided. 3.5.1 EV Charging/Discharging Strategies From the testing results shown in Section 3.4.2 and 3.4.3, conclusion can be drawn that interaction ofEVs, CES, and grid during the charging/discharging could have great impact on the power grid. Based on the V2G and EV technology, allocation the proportions of EVs and interactive strategies of centralized and decentralized are significantly important to impact on the grid. The key point of these strategies are managing peak load condition with EVs discharging to the grid to save electricity usage and shave the peak of the load curve. Also, methods of recommendation and investment suggest that commercialize EVs benefits for the better future of developing EVs industry. Smart charging (G2V) for the purpose of "valley-filling" and V2G are very effective charging options compared to others. Moreover, Vehicle to Stationary Battery (V2B) is considered in a future EVs charging system with storage device. Many of EV researches have worked to establish certain standards or guide line which is not build completely yet. Through simulating and modeling, EVs charging standards required with strategies. In this report, the basic EV charging technologies that are either available in the market or coming to market in the near future have been reviewed. Generalized charging strategy combined with G2V, G2B, B2V, B2G and V2G concepts and PV installation are described. Algorithms, program in MATLAB, and simulation in EDD are given and tested. Now, the question becomes how quickly the transition will occur in the real life. In order to make this transition as seamless as possible, these efforts must be coordinated to ensure that EV charging equipment and charging strategies are available and reasonable, and the benefits are maximized while minimizing the cost. 92 USC Viterbi School of Engineering PhD Thesis 3.5.2 EV Coupled with Storage in PSCAD • The relationship between number of EV and the power change is very linear. This can be used to predict the power needed when more EVs are in the system. • The reactive power needed will increase dramatically after the Battery Storage and EVs are plugged in, which may cause some problems in real life. • In this transient period, the Battery Storage (BS) will generate higher harmonic than when no BS in the system which may cause damages to the system and the devices. However, the harmonics in this scenario are much lower, nearly half the magnitude, than the scenario which has a PV system (with few exceptions). • With the increase of EVs, the harmonic in the system in different point will increase. It may because the BS model itself is another EV with larger capacity, so the number of 'EV' increased, the harmonics will increase accordingly. • The harmonic test is largely based on how the models are built. Different EV models may also take us to different outcomes. Thus, all the conclusions in this paper can only be regarded as a reference. Practice is the sole criterion for testing truth. • According to the data in the time when no BS is connected to the, the harmonic in the grid is larger. This means the BS could produce large harmonic in the grid. 3.5.3 Conclusion This chapter's analysis and study results are based on the assumptions and data discussed in the report and are conducted using MATLAB, PSCAD and EDD analysis models. Due to lack of real world data on EV charging and discharging, only simulation results are available for this part of the research. Several charging/discharging scenarios have been designed to represent G2V, G2B, B2V, B2G and V2G. They are listed as follows: • Different Scenarios for Battery Aggregation and Backfill • Scenario 1: Base Case (G2V and V2G) • Scenario 2: Simple Case (G2V and V2G) • Scenario 3: High Capacity (G2V and V2G) • Scenario 4: Full Peak Shaving (G2V, G2B, B2G and B2V) • Different Scenarios for Battery Aggregation and Backfill Coupled with PV Panels • G2V charging at period 2 (55% loading condition) with 30% Capacity PV • G2V charging at period 2 (55% loading condition) with 55% Capacity PV • G2V charging at period 2 (55% loading condition) with 100% Capacity PV 93 USC Viterbi School of Engineering PhD Thesis • V2G charging at period 1 (100% loading condition) with 30% Capacity PV • V2G charging at period 1 (100% loading condition) with 55% Capacity PV • V2G charging at period 1(100% loading condition) with 100% Capacity PV • Battery Aggregation and Backfill Strategies in MATLAB Modeling • EV Coupled with Storage in PSCAD, power quality and harmonics It can be concluded that battery aggregation and backfill could have positive potential impact on the power grid, if charging/discharging algorithms, together with the concepts of G2B, G2V, B2G, B2V, and V2G, have been properly designed and implemented. 94 USC Viterbi School of Engineering PhD Thesis 4. CHAPTER 4: EV Charging Patterns and Power Usage of Charging Stations 4.1 Research Overview In this chapter, a study is done on charging station and EV usage in Los Angele, CA The data has been classified into three major categories: large corporate fleet, local residential EV and LADWP ride share EVs. The usage of EV and charging stations, as well as the charging patterns have been recorded and analyzed for a time period of one year. The EV usage data and station records came from two sources: LAD WP and UCLA. LAD WP data together with UCLA data will be used to represent large corporate fleet. The results will be used to provide information and practical insights for LAD WP and other similar utility companies for resource planning and future installations of wired and wireless charging infrastructures for EV [ 61-63]. EV charging data was collected either through meters or through a third party (FleetCarma). Power consumption data are stored using timestamp information. In addition, the miles driven between each charging session are collected for statistical analysis and EV driving characterization. Once enough power consumption data is collected to determine usage patterns and load curve shapes, analysis will be performed to investigate the feasibility of using EV chargers to levelize load curve shapes by removing/reducing peaks and valleys. Another benefit of charger monitoring will be the ability to detect a charger failure in timely manner. There will be 3 types of EV charging stations that monitored by this part of the research: 1. Two groups of stations will be for use by the large corporate fleet. EVs in the large corporate fleet will be used by employees for daily operations. These vehicles will be distinguished by short-range, frequent usage during working hours, as follows: a. LAD WP fleet b. UCLAfleet 2. The second category of station will be for local residential use. These cars will be allotted to the public for a portion of the work day. This scenario will be distinguished by very short-range and infrequent use, relative to the previously described station. 3. The third category of station will be used by LADWP employees for Ride Share use. Usage in this scenario will be distinguished by longer range, daily usage. These 95 USC Viterbi School of Engineering PhD Thesis vehicles will be left overnight without charging, indicating that the battery depletion per ride should be significantly larger in comparison to the other two scenarios. 4.2 Data Collection and Test Fleet This analysis was conducted in Los Angeles, CA with a total population of around 3.9 million. The data received comes from two sources: John Ferraro Building's (JFB) parking lots in LADWP, and Parking Structures in UCLA The data collection included a range of operational use cases and conditions, charging stations, coiporation fleet and personal EV s. 4.2.1 Large Corporate Fleet and Charger Information 4.2.1.1 LADWP Fleet and Charger Information In LADWP, 251 chargers are available, including 14 DC fast charger and 237 level 2 chargers. 36 assigned EVs have been equipped with FleetCarma trackers in order to record the daily usage and charging events. The data was collected since February 2015, and the data used in this analysis was from May 20th 2015 to May 20th 2016 because only one EV was found from February to May 2015. Effective data recorded for these 36 EVs started from middle of May 2015. 3,073 charging events have been recorded in this time period. 4.2.1.1.1 EV Status In this study, 36 EVs in 5 brands and 8 different models have been used in LAD WP. The specific model and battery size are shown in Table 4 .1. Table 4.1 EV Model and Battery Size in LADWP 7.6 Focus 23 Fusion Energi PHEV 3 7.6 i-MiEV BEV 3 16 Leaf BEV 4 24 RAV4 BEV 4 42 Prius PHEV 3 4.4 96 USC Viterbi School of Engineering PhD Thesis 4.2.1.1.2 Charging Stations and Data Collection Multiple EV charging stations are available in LADWP consisting of Level 1, 2 and 3 chargers. In total, 251 chargers are available, including 14 DC fast charger and 237 level 2 chargers. These chargers can be found in the Parking Levels 1, 2 and 3 of the John Ferraro Building for the LADWP workers and visitors to charge their EVs. Parking Level 3 arrangement is shown in Figure 4.1. J FB-P3 ~ - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ .. ~ 51 52 53 54 55 56 57 I 50 I 49 I 4a I I 41 I 4s I 451 1 «1431421 1 41I40 I 39 I I 3a I 31 I 3s I D BCRDP Charser • non-SCRDP Charser I I 18 19 20 Figure 4.1 LADWP JFB Parking Level 3 Charging Station Arrangement Level 2 stations are using US/Japan standard type Jl 772. One level 3 charging station was installed and it complied with CHAdeMO standard connector. The output power for level 2 and 3 chargers are 6.6 kW and 24 kW, respectively. 4.2.1.1.3 Charging Events and Data Interpretation EV charging records received from FleetCarma have start time, duration, charging level, charging energy (kWh), charging loss (kWh), starting and ending state-of-charge (SOC) (%). A charging record is defined as a session that starts when an EV's charging hatch is opened and ends when the hatch is closed. 3,073 charging events have been filtered and analyzed in this paper for a time period of one year, from May 20th, 2015 to May 20th, 2016. All data were recorded and updated in FleetCarma. 97 USC Viterbi School of Engineering PhD Thesis 4.2.1.2 UCLA Fleet and Charger Information In UCLA, there are 19 charging stations, including 1 DC fast charger, 13 level 2 and 9 level 1 chargers. All level 1 stations were complied with NEMA 5-20R connectors, while level 2 stations were using US/Japan standard type Jl 772. One level 3 charging station was installed and it complied with CHAdeMO standard connector. The output power for level 1, 2 and 3 were 1.9 kW, 6.6 kW and 24 kW respectively. There were more than 17 EV models selected to analyze the usage and charging patterns. Similar to LADWP, the data was collected since July 2014, and the data used in this analysis was from July 2014 to June 2015. All meters recorded and updated the chargers' information every 15 minutes. The pilot data collected in UCLA includes 19,617 charging events and one year charging station record. 4.2.1.2.1 EV Status In this study, 11 brands and 17 different EV models were used in UCLA by daily commuters. The specific model and battery size can be found below, Table 4.2. Table 4.2 EV Model and Battery Size in UCLA 22 21.3 16 24 C-Max Energi 7.6 Ford - Focus 23 .. Fusion Energi 7.6 Accord Hybrid 6.7 Fit EV 20 Honda n - Kia - Soul EV 27 .. Mitsubishi i-MiEV 16 Nissan .. Leaf 24 .. Smart ED 13.2 Models 60 Models 85 Tesla .. .. Toyota RAV4 42 98 USC Viterbi School of Engineering PhD Thesis 4.2.1.2.2 Charging Stations and Data Collection Several EV Charging Stations are available on UCLA campus consisting of Levell, 2 and 3 chargers. These chargers are available in Parking Structures 1, 5, 9 and 32 for the UCLA community and visitors for EV charging as shown in Figure 4.2. • Pauley Pavilion ~ H Ronald Reagan UCLA ~ Medical Center t~ WE"STWOOD VILLAGE: Hammer Museum s Dickson Ct Figure 4.2 Location of UCLA Charging Stations 4.2.1.2.3 UCLA Schematic The schematic of UCLA campus is shown in Figure 4.3 below. Red color represents the location of all chargers. 99 i f i I BUS BD·H K\'S.C 52·1 2 ~ USC Viterbi School of Engineering c. 12 ,l C-lH ~•• + CRJ· 13().!C ~ Plln! .Old ~ ~ I T r t1~ ~r~ z . i • Cl<J.01C 18 ~ ~ 0 ~ C · r8 f Figure 4.3 Schematic of UCLA 4.2.1.2.4 Charging Events and Data Interpretation PhD Thesis BUSt-r.'S-0 1247kV / M~]:, SH·5 TT•v T L EV charging record received from EVs have start and end time, voltage, current, power factor, active power, main power and apparent power. A charging record is defined as a session that starts when anEV's charging hatch is opened and ends when the hatch is closed. 19,617 charging events have been filtered and analyzed in this paper for a time period of more than one year, from July P\ 2014 to August 3P\ 2015. However, only one years' worth of data was analyzed. All data were recorded every 15 minutes. 100 USC Viterbi School of Engineering PhD Thesis 4.2.2 LADWP Local Residential Fleet and Charger Information In the residential areas within the LADWP service territory, 893 EV rebate chargers have been installed. 223 EV rebate meter data were collected and analyzed to represent for local residential use of EV. Each meter reflects information of one home and will be recording 240V level 2 EV charging energy. The data used in this analysis was from February 1st 2016 to April 30th 2016. All meters recorded and updated the chargers' information every 15 minutes. The pilot data collected from these 223 rebate meters includes more than 1 million records of 15-minute reads of EV charging meter data. Table 4.3 below lists information of all meters and their associating zip code. Table 4.3 EV Rebate Meter Information Meter Zip Code Meter Zip Code Meter Zip Code 73AED00009-01499272 90732 73AED00009-01499386 91607 7AED00009-01499074 90035 73AED00009-01499273 90046 73AED00009-01499387 91606 7AED00009-01499075 90064 73AED00009-0149927 4 90049 73AED00009-01499388 91401 7AED00009-01504913 90042 73AED00009-01499275 90035 73AED00009-01499390 90027 7AED00009-01504914 90042 73AED00009-01499276 91344 73AED00009-01499391 90039 7AED00009-01504915 90042 73AED00009-01499277 91344 73AED00009-01499394 90011 7 AED00009-01504916 91423 73AED00009-01499278 91403 73AED00009-01499396 91326 7AED00009-01504917 90042 73AED00009-01499279 91403 73AED00009-01499398 91335 7AED00009-01504919 90027 73AED00009-01499280 91411 73AED00009-01499399 91367 7AED00009-01504924 91364 73AED00009-01499282 90039 73AED00009-01499400 90026 7AED00009-01504926 91604 73AED00009-01499286 91344 73AED00009-01499403 90210 7AED00009-0150492 7 91364 73AED00009-01499287 91325 73AED00009-01499404 91403 7AED00009-01504932 90065 73AED00009-01499288 91316 73AED00009-01499405 90048 7AED00009-0150493 7 91406 73AED00009-01499290 91344 73AED00009-01499406 91606 7AED00009-01504938 91405 73AED00009-01499291 91344 73AED00009-01499407 91607 7AED00009-01504939 91042 73AED00009-01499292 91343 73AED00009-01499408 91345 7AED00009-01504940 91423 73AED00009-01499295 91344 73AED00009-01499410 91607 7AED00009-01504941 90046 73AED00009-01499306 90049 73AED00009-01499414 90065 7AED00009-01504942 90039 73AED00009-01499308 90046 73AED00009-01499415 90065 7AED00009-01504943 90027 73AED00009-01499312 91307 73AED00009-01499440 90066 7AED00009-01504945 90035 73AED00009-01499313 91304 73AED00009-01499443 91403 7AED00009-01504948 90036 73AED00009-01499315 91411 73AED00009-01499444 90064 7AED00009-01504949 90004 73AED00009-01499316 91401 73AED00009-01499448 91436 7AED00009-01504956 91325 73AED00009-01499317 90046 73AED00009-01499450 91436 7AED00009-01504958 90039 73AED00009-01499318 90039 73AED00009-01499451 91316 7AED00009-01504959 90039 73AED00009-01499322 91423 73AED00009-01499458 90045 7AED00009-01504961 91401 101 USC Viterbi ' School of Engineering PhD Thesis 73AED00009-01499328 91401 73AED00009-01499459 90045 7AED00009-01504962 91423 73AED00009-01499330 91325 73AED00009-01499473 90292 7AED00009-01504963 91604 73AED00009-01499331 91325 73AED00009-01499476 90292 7AED00009-01504964 91405 73AED00009-01499333 90049 73AED00009-01499478 90272 7AED00009-01504965 91436 73AED00009-01499334 91604 73AED00009-01499479 90049 7AED00009-01504967 90046 73AED00009-01499335 91423 73AED00009-01499488 91316 7AED00009-01504968 91401 73AED00009-01499336 91367 73AED00009-01499489 91335 7AED00009-01504969 91401 73AED00009-01499338 90035 73AED00009-01499492 90049 7AED00009-01504970 91401 73AED00009-01499339 91604 73AED00009-01499493 90049 7AED00009-01504971 91401 73AED00009-01499340 91326 73AED00009-01499494 90049 7AED00009-01504972 91304 73AED00009-01499341 91335 73AED00009-01499496 90272 7AED00009-01504973 91306 73AED00009-01499342 91367 7AED00009-01498909 90272 7AED00009-01504978 90046 73AED00009-01499343 91367 7AED00009-01498910 90049 7AED00009-01504983 90039 73AED00009-01499345 91342 7AED00009-01498911 90049 7AED00009-01504984 91344 73AED00009-01499346 91042 7AED00009-01498912 90049 7AED00009-01504986 91326 73AED00009-01499347 91042 7AED00009-01498913 90049 7AED00009-01504987 91335 73AED00009-01499348 90065 7AED00009-01498915 90291 7AED00009-01504988 91406 73AED00009-01499349 91367 7AED00009-01498916 90049 7AED00009-01504989 90041 73AED00009-01499350 91345 7AED00009-01498917 90272 7AED00009-01504991 91401 73AED00009-01499352 91401 7AED00009-01498918 90049 7AED00009-01504994 90065 73AED00009-01499354 91403 7AED00009-01498920 90402 7AED00009-01504996 90045 73AED00009-01499355 91411 7AED00009-01498921 90402 7AED00009-01504998 90018 73AED00009-01499357 91406 7AED00009-01498924 90025 7AED00009-01504999 90068 73AED00009-01499359 91406 7AED00009-01498925 90064 7AED00009-01505000 91601 73AED00009-01499360 91401 7AED00009-01498926 90064 7AED00009-01505001 91401 73AED00009-01499361 90036 7AED00009-01498927 90034 7AED00009-01505002 91367 73AED00009-01499362 90019 7AED00009-01498936 90066 7AED00009-01505008 91604 73AED00009-01499363 91343 7AED00009-01498937 90045 7AED00009-01505009 91607 73AED00009-01499364 91364 7AED00009-01498938 90045 7AED00009-01505010 90046 73AED00009-01499365 91364 7AED00009-01498940 90034 7AED00009-01505011 90048 73AED00009-01499366 90029 7AED00009-01498943 91436 7AED00009-01505012 90046 73AED00009-01499367 91316 7AED00009-01498944 90066 7AED00009-01505015 91606 73AED00009-01499368 91364 7AED00009-01498945 90066 7AED00009-01505016 91364 73AED00009-01499370 90004 7AED00009-01498946 90066 7AED00009-01505017 91344 73AED00009-01499371 90004 7AED00009-01498947 90292 7AED00009-01505018 91344 73AED00009-01499373 90210 7AED00009-01498952 90064 7AED00009-01505019 91344 73AED00009-0149937 4 90039 7AED00009-01498957 91316 7AED00009-01505020 90045 73AED00009-0149937 5 90008 7AED00009-01498966 91356 7AED00009-01505021 90732 73AED00009-01499376 90039 7AED00009-01498968 90077 7AED00009-01505023 90039 73AED00009-01499377 90039 7AED00009-01498970 90025 7AED00009-01505032 90036 102 USC Viterbi School of Engineering ' PhD Thesis 73AED00009-01499378 90026 7AED00009-01498971 90731 7AED00009-01505035 90036 73AED00009-01499379 90068 7AED00009-01498972 90049 7AED00009-01505036 91423 73AED00009-01499380 91403 7AED00009-01498973 90077 7AED00009-01505038 90068 73AED00009-01499381 91423 7AED00009-01499062 90291 7AED00009-01505040 91306 73AED00009-01499382 91403 7AED00009-01499063 90291 7AED00009-01505041 91306 73AED00009-01499383 91423 7AED00009-01499064 90272 7AED00009-01505042 91306 73AED00009-01499384 91607 7AED00009-01499065 90272 7AED00009-01505044 90065 73AED00009-01499385 91311 7AED00009-01499067 90272 7AED00009-01505045 91604 7AED00009-01505046 90021 I 4.2.3 LADWP Ride Share Fleet and Charger Information In LADWP, 251 chargers are available, including 14 DC fast charger and 237 level 2 chargers. 30 ride share EVs were equipped with FleetCarma trackers in order to record the daily usage and charging events. The data was collected since February 2015, and the data used in this analysis was from May 20th 2015 to May 20th 2016 because only one EV was found from February to May 2015. Effective data recorded for these 30 EVs started from middle of May 2015. 8,211 charging events have been recorded in this time period. 4.2.3.1 EV Status In this study, 30 EVs in 2 brands and 2 different models were used in LAD WP. The specific model and battery size are shown in Table 4.4. 4.2.3.2 Table 4.4 EV Model and Battery Size in LADWP Brand Model Number of Vehicles Battery Size (kWh) l§l@ji!i Leaf @i,ij.Jd RAV4 24 6 Charging Stations and Data Collection 24 42 Multiple EV charging stations are available in LADWP consisting of Level 1, 2 and 3 chargers. In total, 251 chargers are available, including 14 DC fast charger and 237 level 2 chargers. These chargers can be found in the Parking Levels 1, 2 and 3 of the JFB for the LADWP workers and visitors to charge their EVs. Parking Level 3 arrangement is shown in Figure 4.4. 103 J FB-P3 ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~ - ~ ~~~ USC Viterbi School of Engineering I 3s I 34 I 33 I I 32 I 31 I30 I I 29 I 28 I 21 I I 2s I 2 s I 24 I I 23 I 22 I 21 I I so 149148 I I 41 I 4s I 4s I I« I 431 42 I I 4 1 I 40139 11 38 I 31 I 3s I D SCRDP C harger • non-SCRDP Charger PhD Thesis I ~ 3 4 ~ 6 7 18 19 20 Figure 4.4 LADWP JFB Parking Level 3 Charging Station Arrangement Level 2 stations were using US/Japan standard type J1772. One level 3 charging station was installed and it complied with CHAdeMO standard connector. The output power for level 2 and 3 chargers were 6.6 kW and 24 kW, respectively. 4.2.3.3 Charging Events and Data Interpretation EV charging records received from FleetCarma have start time, duration, charging level, charging energy (kWh), charging loss (kWh), starting and ending state-of-charge (SOC) (%). A charging record is defined as a session that starts when an EV's charging hatch is opened and ends when the hatch is closed. 8,211 charging events have been filtered and analyzed in this paper for a time period of one year, from May 20th, 2015 to May 20th, 2016. All data were recorded and updated in FleetCarma. 104 USC Viterbi School of Engineering PhD Thesis 4.3 EV Charging Pattern Charging data were obtained from the receiver equipped in all EV s. Charging patterns were analyzed based on charging plug-in time, and total energy transferred per charging event. 4.3.1 Charging Pattern of Large Corporate Fleet 4.3.1.1 Charging Pattern in LADWP Figure 4.5 shows the energy delivery for all assigned EVs charging at LADWP from May 20th, 2015 to May 20th, 2016. It can be noticed that 90% of charging events (out of 3,073 events) transferred less than 11 kWh. Also, there were few charging happen in 2015 than in 2016. This may because of the less assigned EV usage in 2015. It can also be found that there existed certain patterns that more charging events happen in weekdays than weekends. ~ ... Energy Transfer per Charging Event of LADWP 36 Assigned EVs May 2015 - May 2016 35 30 f..(;ha < l1 kWh 25 20 15 10 oi\lte~~~~~~LtllJfiR!~M 2015/ 5/ 20 2015/7/9 2015/ 8/ 28 2015/10/17 2015/12/ 6 Dat e 2016/ 1/ 25 • • • 2016/3/ 15 2016/ 5/ 4 Figure 4.5 Energy Delivery per Charging Event ofLADWP Assigned EVs from May 2015 to May 2016 In Figure 4 .5, no adjustments were made for the type of charger used or the type of vehicles. As outlined in Section 4.2.1.2.2 above, the three types of chargers used have different energy transfer rate capacities. Furthermore, the various classes of vehicles in the study have different battery sizes and thus different energy storage capacity. Data analysis with adjustments for each charger type and vehicle class is not useful as this time due to the sample size of the data. The histogram of the charging plug-in time intervals over the whole 3,073 charging events is plotted in Figure 4.6. Out of all charging events, 94% (2,889/3,073 events) of EV users 105 USC Viterbi School of Engineering PhD Thesis parked and plugged in their EVs for less than 4 hours. Around 1 % (34/3,073 events) of the time EV users would plug in their EVs for more than 5 hours. Included in the plug-in time data of Figure 4.6 is the actual charging time of the EVs. This is a function of the battery SOC at the time of plug-in, the level of charger used, the vehicle battery size, and the SOC at the time of disconnection. The data collection devices don't have these data to make adjustments to accurately describe the charging time. Other indirect methods such as the miles driven between charges, the charger type used and the battery size can be used to make approximations for the actual charging time. Histogram of Plug-in Tim e of LADWP {36 Assigned EVs) May 2015 - May 2016 :i 1000 .. !! w 800 b.O .. ·~ 600 ti ~ v .... 400 0 .. ii 200 E = z 0 0-1 Plug- in Time !hours) Figure 4.6 Histogram of Charging Plug-in Time ofLADWP Assigned EVs from May 2015 - May 2016 4.3.1.2 Charging Pattern in UCLA Charging data were obtained from the receivers in all of the EVs. Charging patterns were analyzed based on charging plug-in time, and total energy transferred per charging event. Figure 4.7 shows the energy delivery for every EV charging at UCLA from July 1st, 2014 to August 31st, 2015. It can be noticed that 90% of charging events transferred less than 12 kWh. Because of the Christmas and Spring Break holiday, very few charging events took place during these two time periods. In Figure 4.7, no adjustments are made for the type of charger used or the type of vehicles. As outlined in Section 4.2.1.2.2 above, the three types of chargers used have different energy transfer rate capacities. Furthermore, the various classes of vehicles in the study 106 USC Viterbi School of Engineering PhD Thesis have different battery sizes and thus different energy storage capacity. Data analysis with adjustments for each charger type and vehicle class is not useful as this time due to the sample size of the data. 80 70 60 90% of charges 50 < 12 w ..c • :!: 40 .... 30 • 20 10 0 Energy Transferred per Charging in UCLA • • • * + Christmas and New Year • • • Spring Break • • • 2014/ 6/ 30 2014/ 8 / 19 2014/ 10 / 8 2014/11/ 27 2015/ 1/ 16 20 15/ 3/ 7 2015/ 4/ 26 2015/ 6/ 1 5 2015/ 8 / 4 Dat e Figure 4.7 Energy Delivery per Charging Event in UCLA from June 2014 to August 2015 The histogram of the charging plug-in time intervals over the whole 19 ,617 charging events is plotted in Figure 4.8. Out of all charging events, 67% (13143/19617 events) of EV users parked and plugged in their EVs for less than 4 hours, while 87% (17067/19617 events) of EV drivers plugged in their EVs for less than 7 hours. The reason that their plug-in time is longer is probably that the EV users sometimes leave their vehicles plugged in to the charger beyond the time required to fully charge the vehicle. Included in the plug-in time data of Figure 4.8 is the actual charging time of the EVs. This is a function of the battery state of charge (SOC) at the time of plug-in, the level of charger used, the vehicle battery size, and the SOC at the time of disconnection. The data collection devices don't have these data to make adjustments to accurately describe the charging time. Other indirect methods such as the miles driven between charges, the charger type used and the battery size can be used to make approximations for the actual charging time. 107 .. c 6000 ! 4500 b,11 c '[li3000 Ill ... u 1 500 0 USC Viterbi School of Engineering Histogram of t he UCLA Plug-in Time from July 2014toAugust 2015 ~ .., ,------.... ~ ................ ' .. . •pi37 \ • • PhD Thesis 87% of users charge< 7 hrs 0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8"9 > 9 Plug-in Time (hrs) Figure 4.8 Histogram of Charging Plug-in Time in UCLA from June 2014 to August 2015 4.3.2 Charging Pattern of LADWP Ride Share EVs Figure 4.9 shows the energy delivery for all assigned EVs charging at LADWP from May 20th, 2015 to May 20th, 2016. It can be noticed that 90% of charging events (out of 8,211 events) transferred less than 8.77 kWh. Because of the Christmas and New Year holiday, very few charging events took place during these two time periods. Also, there was almost no charging in the weekends for ride share EVs. This may because the ride share EVs can only be used in the weekdays for working purpose. 30 25 20 i 15 10 0 2015/5/20 • • • Energy Transfer per Charging Event - LA DWP R ide Share (30 E Vs) May 2015 to May 2016 • • • • • •• •• • 90% of C harges • • < 8.77kWh • • • 1ianksgivinc • • • • • Christmas& + • • • \ • + NewYear + • • • • 2015/7/9 2015/8/ 28 2015/10/17 2015/12/6 2016/1/25 2016/3/15 2016 / 5/4 Date Figure 4.9 Energy Delivery per Charging Event of LADWP Ride Share from May 2015 to May 2016 108 USC Viterbi School of Engineering PhD Thesis In Figure 4.9, no adjustments were made forthe type of charger used or the type of vehicles. As outlined in Section 2 above, the three types of chargers used have different energy transfer rate capacities. Furthermore, the various classes of vehicles in the study have different battery sizes and thus different energy storage capacity. Data analysis with adjustments for each charger type and vehicle class is not useful as this time due to the sample size of the data. The histogram of the charging plug-in time intervals over the whole 8,211 charging events is plotted in Figure 4.10. Out of all charging events, 84% (6,897/8,211 events) of EV users parked and plugged in their EVs for less than 2 hours. Less than 1 % ( 49/8,211 events) of the time EV users would plug in their EVs for more than 5 hours. Compared with assigned EVs in LADWP, the charging plug-in time is much less because users in the ride share program need to use EV more frequently. Included in the plug-in time data of Figure 4.10 is the actual charging time of the EVs. This is a function of the battery SOC at the time of plug-in, the level of charger used, the vehicle battery size, and the SOC at the time of disconnection. The data collection devices don't have these data to make adjustments to accurately describe the charging time. Other indirect methods such as the miles driven between charges, the charger type used and the battery size can be used to make approximations for the actual charging time. Histogram of Plug-in Time of LADWP Ride Share {30 EVs) May 2015 to May 2016 .. .. " 5000 ! 4000 "° " ·~ 3000 .. .: u 'Ci 2000 ~ 1000 :I z 0 0-1 1-2 \ I 2-3 84 o o users charge < 2 hrs 125 3-4 Plug-in Time (hours) >5 Figure 4.10 Histogram of Charging Plug-in Time ofLADWP Ride Share from May 2015 to May 2016 109 USC Viterbi School of Engineering PhD Thesis 4.4 Monitor of Large Corporate Charging Station The data of charging station were obtained primarily through the AMI network with additional communication of all chargers. Power consumption data will be stored using timestamp information. 24 hour charging energy distribution were plotted and analyzed hourly, daily, monthly, quarterly, and yearly. 4.4.1 Station Usage in LADWP 4.4.1.1 Methodology FleetCarma has 24-hr plot for each EV of specified time period, Figure 4 .11. However, the exact value of the data cannot be provided by FleetCarma. Additionally, the 24-hr cumulative charging data for all 36 EVs is not available. An algorithm based on the original data in FleetCarma, Figure 4.12, has been created in Excel to plot the same 24-hr charging energy distribution. Ti me of day charging energy profile • Charging Energy Consumed In Target Time Period • Charging Energy Consumed Outside Target Time Period 10 11 12 13 14 1~ 16 17 18 19 2ll 21 22 2l Time of Day Figure 4.11 Example of 24-hr Charging Energy Distribution for one EV 110 USC Viterbi School of Engineering PhD Thesis Percentage of electricity consumed during target time period Goal : 80% Act ual : 2% Charge Charge r Energy Cha rging Loss Start SOC End SOC Stan Date Duration Leve l (kWh) (kWh) (% ) (% ) Lati1ude Longi1ude June 03 2015 02:57·10 PM 01:26:24 Level 2 6.15 0.49 70 86 34 08222 -118.21836 ~ June 09 2015 12:39:27 PM 00:50:39 Level 2 3.47 0.28 77 86 34.08223 -118.218151 June 15 201511:01:53 AM 01:49:29 Level 2 7.87 0.63 66 86 34.08230 -118.21819 June 23 2015 12:43:27 PM 03:31 :40 Level 2 15.93 1.27 46 86 34.08207 -118.21830 June 29 2015 08:21:42 AM 00:51:38 Level 2 3.59 0.29 77 86 34.08226 -118.21823 July 07 201512:33:20 PM 03:01 :51 Level 2 13.80 1.10 51 86 34.08211 -118.21817 July 09 201510:18:53 AM 00:38:29 Level 2 265 0 21 79 86 34 08225 -118.21821 July 16 2015 01 :34:59 PM 01:24:37 Level 2 6.11 0.49 70 86 34.08225 -118.21817 July 28 2015 02:44 35 PM 03:57:01 Level 2 18 01 1.44 40 86 34 08223 -118.21821 10 July 29 2015 10:43:27 AM 01:10:03 Level 2 5.02 0.40 73 86 34.08241 -118.21817 August 04 2015 04.03:57 00:53:46 Level 2 3.74 0.30 76 86 34 08226 -118.21817 PM 11 o\11n11~t 1') '>n1s;; ns;;·1o··u;; Figure 4.12 Example of Original Charging Record for one EV The algorithm is described as follows: • Convert Start Time (ST) and Duration (D) into seconds • Convert all integer hour into seconds (e.g. 1 :00:00 = 3600s), and let Yi= seconds of ith hour • Let Xi= Yi - ST, z =seconds charging during (i-1)1h to i 1 h hour If xi < 0, ST is not within (i-1)1h to ith hour, z = 0 • • If O < Xi < 3600, ST within (i-1) 1 h to ith hour, z =min (xi, D) If 3600 < Xi < D, charging all the time from (i-1 )1h to ith hour, z = 3600 Ifxi> D );> D < yi-1-ST,z=O );> D > Yi-I - ST, z = D - [Yi-I - ST] Charging energy during (i-1)1h to ith hour= z x Energy ID Repeat for all charging records and plot cumulative result After applying the above algorithm to the charging records listed in Figure 4.12, the resulting plot is shown below, Figure 4.13. The results are summarized in the following sections, Section 4.4.1.2 to 4.4.1.5. 111 ..c 80 70 60 50 s 40 ..:.: 30 20 10 0 USC Viterbi School of Engineering PhD Thesis 24 Hour Charging Energy Distribution - FleetCarma • _ I • - • I I - 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time Interval Figure 4.13 Example of 24-hr Charging Energy Distribution Plot for one EV using Designed Algorithm 4.4.1.2 Hourly Charging Distribution Analysis Figure 4 .14 records hourly energy distribution for all chargers of LAD WP 36 assigned EV s from May 20th 2015 to May 20th 2016, for a total of 8808 hours have been recorded. These data represented the cumulative energy being transferred from the grid to 36 EVs in one year period. Hourly Energy D istribution of LA DWP 36 Assigned EVs from May 2015 to May 2016 25 20 15 ~ s ... ,. ,I~ i,~~/1 iLil1I~~ ,~d~'~lji U~1 11 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""" ~~~~~~~B~~~~~~~~~~~~~~~~~ffi~~~~~~~~~~m~~§~~~$~§R~~~~~~£~~~~ Each H our Figure 4.14 Hourly Charging Distribution ofLADWP 36 Assigned EVs, May 2015-May 2016 As can be seen from the figure, there exists certain patterns for every day and every week in this one year period. The charging energy transferred in the weekdays was much more 112 USC Viterbi School of Engineering PhD Thesis than that of the energy in the weekends. Also, the energy used for charging in 2016 is higher than that of 2015. 4.4.1.3 Weekday vs Weekend Energy Distribution Figure 4.15 is a comparison of 24-hour average energy distribution between weekday and weekend. It can be seen that there were big difference of the energy transferred from grid to all charging stations in weekdays, compared with weekends. The reason is because all assigned EVs in LADWP would need to be used and charged for work purposes more frequently in the weekdays. 6 5 4 ~ 3 .. 2 0 Weekday vs Weekend 24hr Energy Distribution of LA DWP 36 Assigned EVs May 2015 - May 2016 (Averaged) 0 1 2 3 4 5 6 7 8 9 10 11 12 B 14 l5 16 17 18 19 20 21 22 Tim e of Day - Weekday Aver~e - - Weekend Aver~e Figure 4.15 Comparison of Averaged Weekday and Weekend 24 hour Energy Distribution ofLADWP 36 Assigned EVs from May 2015 to May 2016 4.4.1.4 Energy Distribution in Different Seasons Figure 4.16 is a comparison of 24 hour normalized energy distribution in four seasons in 2015-2016, each curve represents one season's energy distribution. As seen in Figure 4.16, the charging energy profiles in all four seasons have two peaks, one in the morning from 7-8 am (hour-ending 08 to hour-ending 09) and the other in the afternoon between 1-4 pm. In summer and fall 2015, the charging energy curves were similar to each other and the energy consumed was about 1/2 of the winter season in 2015 and 1/4 of the spring season in 2016. The reason of this can be found in Section 4.3.1.1, as more charging events occurred in 2016. Also, there were not much charging energy flow in the evening and early morning (6pm to Sam next day) because the assigned EVs were being charged mainly during work hours. 113 900 800 700 600 ..<: 500 $ -" 400 300 200 100 0 USC Viterbi School of Engineering Seasonal Comparison of 24hr Charging Energy Distribution of LADWP 36 Assigned EVs May 2015 - May 2016 (Cumulative) ' ' /\ I ' /...._ I ,....... '"\ I ,. ..... , --.. \ I I .......... , , ..... ,' ',,.,' --:, \ I I , .,,. -- PhD Thesis . - 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of day - • Summer -- Fall --- Wiiter - Spring Figure 4.16 Seasonal Comparison of 24 Hour Cumulative Charging Energy Distribution of LADWP 36 Assigned EVs from May 2015 to May 2016 4.4.1.5 One Year Cumulative Energy Distribution Figure 4.17 is the cumulative 24 hour energy distribution of LADWP 36 assigned EVs' charging energy in one year, from May 2015 to May 2016. It can be seen from the figure below that the cumulative load curve has two peaks, one in 7-8 am and the other in 1-3 pm. The first peak happened when most EV users arrived at work and started charging their EVs, while the second peak occurred in the afternoon after lunch time. The cumulative peak value of all testing EVs charging in LADWP reached around 1,400 kWh. 1400 1200 1000 .., 800 $ ... 600 400 200 0 One Year 24hr Charging Energy Distribut ion of LADWP 36 Assigned EVs May 2015 - May 2016 I • -··· ••••• 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of the Day Figure 4.17 One Year Cumulative 24 hour Energy Distribution of LAD WP 36 Assigned EVs from May 2015 to May 2016 114 USC Viterbi School of Engineering PhD Thesis 4.4.2 Station Usage in UCLA 4.4.2.1 Methodology Since charging energy is recorded for all meters in increasing time order, it should reflect the energy transferred from grid to vehicles. In order to plot the actual 24 hour energy distribution curve of all charging stations in UCLA, the following steps were taken: • Delete bad data • Extract data • Results and Plots Details of each step can be found in the following sections. 4.4.2.1.1 Delete bad data Bad data were deleted as follows: • Delete duplicates in Main Energy because energy would be transferred only when there was an energy increment • Delete the row that contains 0 in Main Energy • Sort Timestamp in increasing order, while keeping Station ID in increasing order • Split last 7 digits of Station ID to another column, since it is named in Hexadecimal, we convert it back to Decimal • Calculate Main Energy difference, and extract all data points ranging from (0, 30], greater than 0 and less than or equal to 30. Since the meter data are recorded every 15 minutes, the energy transferred from grid cannot exceed 0.5kWh, 2kWh, and 30kWh for Levell, 2, and 3, respectively • Calculate difference of that Decimal column, 99% of them should be zero because they reflect data of the same charger • Delete all rows that contain non-zero values in the difference Decimal column, because those are the first row of each station and the difference of this row with the one above are bad data 4.4.2.1.2 Extract Data After deleting all bad data, the remaining data were left for extraction and plot the 24 hour energy distribution. Here, effective data was extracted for every hour time interval. Three steps were under taken as described below: 115 USC Viterbi School of Engineering PhD Thesis • Since the meter data was recorded every 15 minutes, it is assumed to be the average data of the past 15 minutes. Therefore, the data for each hour are filtered out from 15 minutes of an hour to 15 minutes of the next hour, e.g. 0: 15 to 1: 15 • Record the sum of Main Energy Difference to be the cumulative charging energy of that hour interval • Repeat same thing for the rest 23 hours 4.4.2.1.3 Results The results are summarized in the following sections, Section 4.4.2.2 to 4.4.2.5. 4.4.2.2 8760-Hour Charging Distribution Analysis Figure 4.18 records hourly energy distribution for all chargers in UCLA from July 2014 to June 2015, for a total of 8760 hours have been recorded. As can be seen from the figure, there exists certain patterns for every day and every week in this one year plot. There were almost zero energy transferred in one week of December 2014 and also in March 2015, and that was because of Christmas and Spring Break. 8760 Hourly C harging Distribution 07 /2014 to 06/2015 Each Hour Figure 4.18 8760 Hourly Charging Distribution in UCLA from July 2014 to June 2015 4.4.2.3 Weekday vs Weekend Energy Distribution Figure 4.19 is a comparison of 24-hour average energy distribution between weekday and weekend. 116 25 20 15 ""' ~ 10 USC Viterbi School of Engineering PhD Thesis Averaged Weekday and Weekend 24hr Energy Distribution 0 1 2 3 4 5 6 7 8 9 10 11 1 2 13 14 15 16 17 18 19 20 21 22 23 Time of Day - weekday - weekend Figure 4.19 Comparison of Averaged Weekday and Weekend 24 hour Energy Distribution in UCLA from July 2014 to June 2015 It is clear that there were much more energy transferred from grid to all charging stations in weekdays. In the early morning, from 0 am to 4 am (hour-ending 01 to hour-ending 05), and evening, from 8 pm to 12 am, the average energy transferred to all stations in weekdays were twice as much of the weekends. While in the daytime, from 4 am to 8 pm, average energy transferred to all stations was around 4 - 6 times more than that of in the weekends. The reason is because all typical EV users from UCLA would seldom drive the vehicles to campus and charge their EVs during weekends. 4.4.2.4 Summer vs Winter Cumulative Energy Distribution A comparison of 24 hour cumulative energy distribution between winter 2014 and summer 2015 is shown below, Figure 4.20. It can be noticed that the charging energy distribution are almost identical in these two seasons except two slight differences. One is that the winter distribution curve was shifted to be one hour earlier than that of in summer. This is because of the daylight saving time used in summer. Another difference is that there was only one peak in the summer season, at 9-10 am when people arrived at the parking structure and connected the EV s to the chargers. While in the winter, the first peak happened at around 8am and there was a second peak, at 3 pm. 117 USC Viterbi School of Engineering PhD Thesis 24-hr C harging E nergy Distribution Summer vs Winter ~ 400+-~~~~-----4'-~#---~___:: .... 911ii,.:~----='111r-~~~~~~ $ - summer .Iii: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day Figure 4.20 Comparison of Winter 2014 and Summer 2015's Cumulative Charging Energy Distribution in UCLA 4.4.2.5 One Year Cumulative Energy Distribution Figure 4.21 is the cumulative 24 hour energy distribution for all charging stations in one year, from July 2014 to June 2015. It can be seen from the figure that the cumulative load curve has two peaks, one is at 9am when EV users arrived at UCLA campus and plugged in EVs to the charging station. The cumulative peak value of all testing EVs charging in UCLA reached 6500 kWh. The second peak occurs in the afternoon, at around 4pm and the cumulative value is almost 5000 kWh for one year. 7000 6000 5000 4000 ~ .... 3000 2000 1000 0 One Year 24-hr Charging Energy Distribution 07 /2014 - 06/2015 I - • I I I I I I I I I 0 1 2 3 4 5 6 7 8 9 10 11 12 1 3 14 15 16 1 7 18 19 20 21 22 23 Time of Day Figure 4.21 One year Cumulative 24 hour Energy Distribution in UCLA from July 2014 to June 2015 118 USC Viterbi School of Engineering PhD Thesis 4.5 Monitor of LADWP Local Residential Chargers The data oflocal residential chargers were obtained primarily through the EV rebate meters in residential areas within the LADWP service territory. Power consumption data will be stored using timestamp information. However, EV rebate program is a newly launched program since late 2015 and only three months data were available for the analysis, from February pt 2016 to April 30th 2016. Each meter reflects information of one home and will be recording 240V level 2 EV charging energy. Data from 223 rebate meters recorded charging energy every 15 minutes for the EVs in LADWP service territory. 24 hour charging energy distribution were plotted and analyzed hourly, daily, monthly and quarterly. The results can be found in the following sub sections. 4.5.1 Hourly Charging Distribution Analysis Figure 4.22 records hourly energy distribution for all chargers in LADWP residential area from February pt to April 30th 2016, for a total of 2160 hours have been recorded. These data represented the cumulative energy being transferred from the grid to the EVs in three month period. LADWP Residential Hourly Charging Energy Distribution (200+ Meters) Feb - Apr 2016 250 1----.--------t-------------------- 200 t+-r,,---H---f-.-- -t-t-t----.-HH---t-t-- t-t+- ,....,+,,---H- --t-t--t-t-t+- --,-Hl-+rl-,-I- Hour Interval Figure 4.22 Hourly Charging Distribution in LADWP Residential Area, Feb - April 2016 As can be seen from the figure, there exists certain patterns for every day and every week in this three month plot. The charging energy transferred in the weekdays was much more than that of the energy in the weekends. 4.5.2 Weekday vs Weekend Energy Distribution Figure 4.23 is a comparison of 24-hour average energy distribution between weekday and weekend recorded from the 223 meters. 119 .t:: $ .... USC Viterbi School of Engineering PhD Thesis Weekday vs. Weekend 24-hr Charging Energy Distribution (Normalized) Feb to Apr 2016 25000 200.00 150.00 100.00 50 .00 0 .00 r.---------------+----==--- - weekday 0 1 2 3 4 5 6 7 8 9 10 1112 1 3 14 1 5 16 1718 19 20 2122 23 Time Interval - weekend Figure 4.23 Comparison of Normalized Weekday and Weekend 24 hour Energy Distribution in LADWP Residential Area, February to April 2016 It can be seen that the energy charged to EVs were almost the same from midnight until 5 am. During 6-8 am, there were more charging in the weekdays and it is probably because some people would prefer to charge their EVs when they wake up until time to go to work. The charging in weekends between 9 am to 6 pm were more than weekdays. It makes sense because the time period is the normal working hours when people cannot charge at home. After 6 pm, the charging energy in weekdays was far more than that of in the weekends because people finished their work and arrived at home. They need to charge their EVs for the next day use. 4.5.3 Energy Distribution in Different Month Figure 4.24 is a comparison of 24 hour normalized energy distribution in months in 2016. Each curve represents an averaged one day energy distribution in different month from 223 rebate meters. It can be seen that there were only tiny difference of the energy transferred from grid to all chargers in these three months. Certain patterns can be observed in the three months and the curves are desirable from utility's perspective. EV users would use their EV s in the day time and charge in the night. 120 180.00 160.00 140.00 120.00 ..i::: 100.00 i1 80.00 6 0 .00 40.00 20.00 0 .00 USC Viterbi School of Engineering 24-hr Charging Energy Distribution {Normalized) Feb to Apr 2016 0 1 2 3 4 5 6 7 8 9 10111213 1415 1 6 1718 19 20212223 Time Interval PhD Thesis Figure 4.24 Monthly Comparison of 24 Hour Cumulative Charging Energy Distribution in LAD WP Residential Area, February to April 2016 4.5.4 Three Month Cumulative Energy Distribution Figure 4.25 is the cumulative 24 hour energy distribution for all charging in three months, from February to April 2016. It can be seen from the figure that the cumulative load curve has one peak in the evening, between 8-lOpm. The cumulative peak value of all testing EVs charging in LAD WP residential areas reached 15 MWh. This EV charging load curve looks desirable to the utility companies because people will be charging in the evening which may help reduce the actual peak load time occur in the afternoon. It will be even better if the utility companies can apply V2G technology to direct the EV customers to start charging in the midnight to increase off-peak demand and help optimize the load profile. 1 6000.00 14000.00 1 2000.00 10000.00 ..i::: :s: 8 CXJO.OO ... 6000.00 4000.00 2CXJO.OO 0 .00 24-hr Charging Energy Distribution (Cumulative) Feb to Apr 2016 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 5 1 6 17 1 8 1 9 20 21 22 23 Time Int erval Figure 4.25 Cumulative 24 hour Energy Distribution in LADWP Residential Area, February to April 2016 121 USC Viterbi School of Engineering PhD Thesis 4.6 Monitor of LADWP Ride Share Charging Station The data of charging station were obtained primarily through the AMI network with additional communication of all chargers. Power consumption data will be stored using timestamp information. 24 hour charging energy distribution were plotted and analyzed hourly, daily, monthly, quarterly, and yearly. 4.6.1 Methodology Since the data on LADWP ride share EV charging also comes from FleetCarma, the methodology in this section is the same as described in Section 4.4.1.1. 4.6.2 8760-Hour Charging Distribution Analysis Figure 4.26 records hourly energy distribution for all chargers of LADWP ride share program (30 EVs) from May 2015 to May 2016, for a total of 8760 hours have been recorded. These data represented the cumulative energy being transferred from the grid to 30 carpool EVs in one year period. Hourly Energy Distribution of LADWP Ride Share (30 EVs) from May 2015 to May 2016 ~ch Hour Figure 4.26 8808 Hourly Charging Distribution ofLADWP Ride Share (30 EVs) May 2015 to May 2016 As can be seen from the figure, there exists certain patterns for every day and every week in this one year plot. The charging energy transferred in the weekdays was much more than that of the energy charged in the weekends. 122 USC Viterbi School of Engineering PhD Thesis 4.6.3 Weekday vs Weekend Energy Distribution Less than 1 % of charging events happened in the weekends from May 2015 to May 2016, therefore, only weekday energy distribution was plotted. Figure 4.27 is the 24-hour average energy distribution of all weekdays. The curve would almost be the same as the whole year energy distribution, with one peak in the afternoon between 3-4 pm. 24 Hour Weekday Energy Distr ibution of LADWP Ri d e Share (30EVs) from May 2015 t o May 2016 (Average) 0 1 2 3 4 s 6 7 8 9 10 11 12 1 3 14 1 5 1 6 1 7 1 8 1 9 20 2 1 22 23 Time of Da y Figure 4.27 Averaged Weekday 24 hour Energy Distribution of LADWP Ride Share (30 EVs) May 2015 to May 2016 4.6.4 Energy Distribution in Different Seasons Figure 4.28 is a comparison of 24 hour cumulative energy distribution in four seasons in 2015-2016. Each curve represents a 24-hour cumulative energy distribution in different season. As can be seen from the Figure 4.28, the charging energy profiles in all four seasons have the similar shape. All four curves have a peak in the afternoon between 3-4 pm. The energy consumed in the 2016 spring season is the most while the energy used for EV charging in the 2015 winter season is the least. Also, there were not much charging energy flow in the evening and early morning (10 pm to 6 am next day) because the assigned EVs were being charged mainly during work hours. 123 1600 1400 1200 1000 ~ ~ 800 600 400 200 0 USC Viterbi School of Engineering Seasonal Comparison of 24 Hour EV Charging Distribution - LADWP Ride Share (30EVs) May 2015 to May 2016 ,- PhD Thesis 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 7 18 19 20 21 22 23 Time Interval --summer ••••• Autumn --- W inter - Spring Figure 4.28 Seasonal Comparison of 24 Hour Cumulative Charging Energy Distribution of LADWP Ride Share (30 EVs) May 2015 to May 2016 4.6.5 One Year Cumulative Energy Distribution Figure 4.29 is the cumulative 24 hour energy distribution of LAD WP ride share 30 EVs' charging in one year, from May 2015 to May 2016. It can be seen from the figure that the cumulative load curve has one peak, between 3-4 pm. The cumulative peak value of all 30 EV s for the carpool use in LAD WP reached 4,000 kWh. This EV charging load curve looks pretty similar to the daily load profile in California. 4000 3500 3000 2500 3: 2000 .... 1500 1000 500 0 One Year 24hr Charging Distribut ion LA DWP Ride Share 30 EVs 05/2015- 05/2016 r.. - .... Ii I I 11 Cl 0 1 3 4 6 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day Figure 4.29 Cumulative 24 hour Energy Distribution ofLADWP Ride Share (30 EVs) May 2015 to May 2016 124 USC Viterbi School of Engineering PhD Thesis 4. 7 Load Curve Prediction in California with Different EV Penetration Rates Since enough power consumption data has been collected to determine usage patterns and load cmve shapes, analysis will be petformed in this section to investigate the feasibility of using EV chargers to levelize load curve shapes by removing/reducing peaks and valleys. 4.7.1 Original Load Profile in California According to CAISO [59], a typical daily load variation in California in different seasons is shown in Figure 4.30 below. It can be seen that the daily peak time occurred in the afternoon from 2-6 pm, while the off-peak hours were in the early morning, between 2-6 am. x · 10" «ai lY Loa di .,.. -a'li ons ~5 -------~--------r-------,--------T- 1 I 4 -------~--------~--- ' I I I I ~5 -------~------ • I I -- e bDEI -- Apcr1 2 -- J mOB -- JIMJQ I . --r-------,--------T----- 1 I I 1-5 _______ 1 ________ ._ _______ J ________ 1 ______ . I I I I I I I I I I I I 5 10 15 2 Figure 4.30 Daily Load Variation in California by CAISO 4.7.2 California EV Information Plug-in Electric Vehicle registrations per thousand people by state in 2014 is shown below, Figure 4.31. 125 ' •• ""'>- USC Viterbi 0.12 0.75 0.31 . . School of Engineering 0.12 0.19 0.31 0.26 0.21 0.37 PEV per 1,000 People ,_ • !9> 2.15 • • 2.1 and> • 1.51 to2.o D 1.1to1.5 • o.51to1 .0 D Oto0.5 0 PhD Thesis Figure 4.31 Plug-in Electric Vehicle Registrations per Thousand People by State, 2014 In 2014, California had the most PEV registrations of all states with 126,283 PEV registrations among 30 million residents, 3.25 PEVs/1000 people [60]. However, with continuing rapid growth of the adoption rate of EV, the demand in distribution networks is expected to increase. 4.7.3 California Load Curve Prediction According to the data analyzed above, the California daily load curve can be predicted based on large corporate fleet and typical EV. 4.7.3.1 Load Curve Prediction based on Large Corporate - LADWP Fleet Data on LAD WP analyzed in the previous sections represents a combination of 36 assigned EVs and 30 ride share EVs and can be used to estimate the potential demand increments brought by EV charging. Figure 4.32 compares the total daily load curve of different EV penetration rate at 10%, 20% and 30%, respectively, in California summer season. 126 50000 45000 40000 35000 30000 i 25000 20000 15000 10000 5000 0 USC Viterbi School of Engineering Daily Load Variation with Different EV Penetration Rates in California - - ~ --...... ~ "'" _/ ' ~ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Tim e of Day PhD Thesis - oaity Load - • Total (EV 10%) - Total (EV20%) - - • Total (EV 30%) Figure 4.32 Daily Load Variation Prediction based on Large Corporate Fleet data Figure 4.32 suggests that EVs have the potential to increase the demand in the daytime when large corporate fleet EV users arrive at work and start connecting the EVs to the charging station. If coupled with renewable energies, such as installing PV panels and wind turbines, the EV charging increments could be offset [64). 4.7.3.2 Load Curve Prediction based on Large Corporate - UCLA Fleet Data analyzed in the previous sections represents a group of around 20 EV daily commuters and it can be used to estimate the potential demand increments brought by EV charging. Figure 4.33 compares the total daily load curve of different EV penetration rate at 10%, 20% and 30%, respectively, in California summer season. 60000 50000 40000 ~ 30000 20000 10000 0 Daily Load Variation with Different EV Penetration Rates in California - D aily Load , .... iiiil .... ~~rl""'--------------- --Total(EV10%) >----------------------- --Total(E V20%) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day - Total(E V30%) Figure 4.33 Daily Load Variation Prediction based on Typical EV data 127 USC Viterbi School of Engineering PhD Thesis Similar to Figure 4.32, Figure 4.33 suggests that EVs have the potential to increase the demand in the morning when EV users arrive at work and start connecting the EVs to the charging station. Also, there will be a second large load increment in the afternoon at around 4pm. If coupled with renewable energy resources, such as installing photovoltaic panels and wind turbines, the EV charging increments could be offset. 4.8 Conclusion This chapter presents a detailed analysis based on SGRDP to EV usage as well as the charging station usage on the grid. Conclusions and summaries for large corporate fleet and typical EV and the associated charging stations are given in this section. 4.8.1 4.8.1.1 Conclusion for Large Corporate Fleet and Associated Charging Stations LADWP Fleet and Charging Station 36 EVs on 5 brands and 8 models, and multiple chargers at level l , 2 and 3 were selected and tested in LADWP. EV charging events were recorded from May 20th, 2015 to May 20th, 2016. Energy transferred for all 3,073 charging events were studied and the plug-in time per charging were examined. As for charging station usage, 24 hour energy distribution were plotted and analyzed daily, monthly, seasonally, and yearly. The statistic results will be used to provide information and practical insights for utility companies to make city planning and future installations of the EV charging infrastructures. The result from EV charging events implement that 90% of charging events transferred less than 11 kWh. Regarding the EV plug-in time, 94% of the time EV was connected for less than 4 hours, while less than 1 % of the time EV users plugged in their EVs for more than 5 hours. From the charging station usage, it can be concluded that it may be possible to use EV chargers to levelize load curve shapes by removing or reducing peaks and valleys. Two energy peaks brought by EV charging from the grid were between 7-8am and 1-3 pm, respectively. The 24 hour charging energy distribution curve is pretty similar to the daily load curve in California. Charging stations delivered most of the energy to all EVs during daytime, which could be offset by installing PV panels and solar systems. 4.8.1.2 UCLA Fleet and Charging Station More than 17 EVs and 19 chargers at level 1, 2 and 3 were selected and tested in UCLA campus. EV charging events were recorded from June 1st, 2014 to August 31st, 2015 while 128 USC Viterbi School of Engineering PhD Thesis charging station data were collected every 15 minutes since July 1st, 2014. Energy transferred for all 19,617 charging events were studied and the plug-in time per charging were examined. As for charging station usage, 24 hour energy distribution were plotted and analyzed hourly, daily, monthly, quarterly, and yearly. The statistic results will be used to provide information and practical insights for utility companies to make city planning and future installations of the EV charging infrastructures. The result from EV charging events implement that 90% of charging events transferred less than 12 kWh. Regarding the EV plug-in time, 67% of the time EV was connected for less than 4 hours, while 87% of the time EV users plugged in their EVs for less than 7 hours. From the charging station usage, it can also be concluded that it may be possible to use EV chargers to levelize load curve shapes by removing or reducing peaks and valleys. Energy peak brought by EV charging from the grid was between 7-lOam when EV users came to work. The 24 hour charging energy distribution curve is similar to the daily load curve in California only except the EV charging peak time occurred in the morning. Charging stations delivered most of the energy to all EVs during daytime, which could be offset by installing PV panels and solar systems. 4.8.2 Conclusion for Residential EV and Associated Charging Stations In the residential areas within the LAD WP service territory, 223 EV rebate meter data were collected and analyzed to represent for local residential use of EV. Each meter reflects information of one home and will be recording 240V level 2 EV charging energy. The data used in this analysis was from February pt 2016 to April 30th 2016. All meters recorded and updated the chargers' information every 15 minutes. The pilot data collected from these 223 rebate meters includes more than 1 million records of 15-minute interval reads of EV charging. From the charging station usage, it can be concluded that it may be possible to use EV chargers to levelize load curve shapes by removing or reducing peaks and valleys. Energy peak brought by EV charging from the grid was between 8-10 pm. The 24 hour charging energy distribution curve is pretty desirable to the utility company because there was small amount of energy used for EV charging in the daytime while large amount used in the evening. It will be even better if the utility companies can apply V2G technology to direct the EV customers to start charging in the midnight to increase the off-peak demand and help optimize the actual daily load curve. 129 USC Viterbi School of Engineering PhD Thesis 4.8.3 Conclusion for LADWP Ride Share EV and Associated Charging Stations 30 EVs on 2 brands and 2 models, and multiple chargers at level 1, 2 and 3 were selected and tested in LADWP. EV charging events were recorded from May 20th, 2015 to May 20th, 2016. Energy transferred for all 8,211 charging events were studied and the plug-in time per charging were examined. As for charging station usage, 24 hour energy distribution were plotted and analyzed daily, monthly, seasonally, and yearly. The statistic results will be used to provide information and practical insights for utility companies to make city planning and future installations of the EV charging infrastructures. The result from EV charging events implement that 90% of charging events transferred less than 8.77 kWh. Regarding the EV plug-in time, 84% of the time EV was connected for less than 2 hours, while less than 1% of the time EV users plugged in their EVs for more than 5 hours. From the charging station usage, it can be concluded that it may be possible to use EV chargers to levelize load curve shapes by removing or reducing peaks and valleys. Energy peak brought by EV charging from the grid was between 3-4 pm. The 24 hour charging energy distribution curve is pretty similar to the daily load curve in California. Charging stations delivered most of the energy to all EVs during daytime, which could be offset by installing renewable energy resources. 130 USC Viterbi School of Engineering PhD Thesis 5. CHAPTER 5: Analysis on EV User Behaviors 5.1 Research Overview This chapter evaluates EV users charging habits (type of EV, time of connection, duration, consumption, and if possible utilization). The following data was obtained from charging stations and was used to study EV customer's charging habits: charging times, duration, extent of charging and what effects these particular charging profiles would have on the distribution system. Once the customer profiles have been analyzed and categorized, models are created to study the impacts of large scale deployment of EV chargers. Additionally, charging patterns of EV usage are analyzed by the following categories: 1. Large corporate fleets that have charging patterns similar to typical gas stations, 2. Typical EVs that are employed at a university, and 3. Typical owners ofEVs in the city and suburbs. The available sources of data to study the customer's charging habit are listed below: 1. One year survey data of LAD WP employees driving assigned EVs, (BE Vs and PHEVs) from June 2014 to July 2015 2. EV usage data in FleetCarma of both assigned and carpool EVs (BEVs), from May 2015 to May 2016 3. LADWP dispatchers ' data from May 2015 to May 2016 consisting of EV and non-EV data 4. Charging records of EVs (BEVs) at the University of California Los Angeles from July 2014 to August 2015 5. LADWP residential EV (BEV) rebate meter data from February 2016 to April 2016 In this chapter, customer profiles are analyzed and categorized. Based on this profile's analysis, models are created to forecast the EV charging demand in Chapter 6. 131 USC Viterbi School of Engineering PhD Thesis 5.2 LAD WP EV User Habit Analysis Based on One Year Survey In this section, a study on a one-year survey of EV users in LADWP, from June 2014 to July 2015, is presented. (This is the source data Item 1, identified in Section 5.1, above.) It is assumed that the EV users in LAD WP to be the representatives of large corporate EV users. An analysis has been done to study the LADWP EV users' habits on three parts: overall analysis, individual user analysis, and comparison of all users. To study the EV user habit under all applicable circumstances, the users were required to drive the assigned EVs for all different purposes at all the time. 5.2.1 Raw Data and Overall Analysis The first step is to review the one year LAD WP EV Usage Survey data and delete any bad or problematic data. Criteria to determine the problematic data are listed as follows: • Stop time is earlier than start time; • Stop mileage is less than start mileage; • Missing data on start time, stop time, start mileage or stop mileage. Original file had 1,221 pieces of data, but 977 left with the removal of bad data. In the overall analysis, data including missing information, e.g. no user name or other unnecessary information, were kept when filtering data. The first thing analyzed is the 24- hour distribution of trip start time with one-hour interval, shown in Figure 5 .1. 160 140 120 100 80 60 40 20 0 - v ... '} - - 24-Hour Distribution of the Trip Start Time 1 LIR l,jb 131 . -· 1 "" 93 --- ::10 c:c: 31 'JC: - 11 • ~~ . '} - ... . ... . - - ~ - • • • - ~ - ~ v v - ~~ ':l,~(:j 'h~ ,.,.~ "~ "°~ -x~ ~~ ?i~ (;j~(:j "-~ ~~ ~~ ~~ <->~ ro~ ~~ q;,~(:j o;~ ~~ r'I,~ ri;~ ~~ ~~~~-~~~~~~~~~~~~~~~~~~~ ~· • .,.. ,,. 'lj• ~- 'j• (c)· '\. 'I)• o;'T ~~ ...:,,~ ~';;J ~~ ~~ ')~ r.:,'T ~-'T '6~ o;~ ~';;J r'f,~ ri;~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ "' "' "' Figure 5.1 24-Hour Distribution of the Trip Start Time 132 USC Viterbi School of Engineering PhD Thesis From Figure 5.1 above, it is seen that most of the trips happen during 6am to 5pm and the EVs are seldom used after 5pm, which is because that the EVs are only used by LADWP users during working hours. Furthermore, it can be seen that most frequent driving are during 8am to 9am, the reason can be that most LAD WP workers are dispatched or driving to various job sites during 8-9am. The second thing that is analyzed is the distribution of trip duration shown in Figure 5.2. It can be seen that the trip within half an hour occupies the biggest part, around 40%, and within 2 hours is about 83%. It can be included that EVs are usually used for short trips by users in LADWP, which is the result of battery limitation and work requirements. Furthermore, it is unusual that there are some trips that appear to last much longer than a typical EV's mileage range. For instance, there are 24 trips with more than 8 hours duration. This trip duration is the time from Check-out to Check-in of the vehicle, which can be assumed that EVs are taken home overnight or charged at places other than LADWP's main building. 400 350 Ii 300 ·.::: I- 250 ..... ~ 200 CIJ ..c 150 E ~ 100 50 0 'Y7n 191 Distribution of Trip Duration 129 82 • 45 32 40 J~ I I • • I 0 I I ... • Figure 5.2 Distribution of Trip Duration Ji:'.l. ... ... L • The third thing that is analyzed is the distribution of average mileage with 3-hour interval, which is shown in Figure 5.3 and the trip number in that time interval is shown in Figure 5 .4. It can be seen how far in average the EV users would like to drive during different time period from Figure 5.3. During 3am to 6am, the average mileage is 6.5 miles, which may be some special uses of EV From 6am to 9am, the average distance is 9.2 miles, which is a reasonable average distance between office and home, and EVs may be used to drive to work. From 9am to 12pm, the average mileage is 14 miles, which is more probable that users drive EVs to some places for working purpose. From 12pm to 3pm, the average mileage is 3.2 miles. Users may drive EVs to nearby job sites. From 3pm to 6pm, the average mileage is 8 miles, which is similar to the distance from 6am to 9am. It is probable 133 USC Viterbi School of Engineering PhD Thesis that the EV users are heading back to work to return the vehicle. From 6pm to 9pm, the average mileage is 24.75 miles, which is the highest. However, there are only a few numbers of trips, which means that the EVs are seldom used for the long trip and may possibly be used for special purpose. 30 25 20 VI ~15 ~ 10 5 0 Distribution of Average Mileage 24.75 o~-- Hours Figure 5.3 Distribution of Trip Duration Pie Chart of Trip Number in 24 Hours Figure 5.4 Pie Chart of Trip Number in 24 Hours 134 0 • 0:00-3:00 • 3:00-6:00 • 6:00-9:00 • 9:00-12:00 • 12:00-15:00 • 15:00-18:00 • 18:00-21:00 • 21:00-24:00 500 ._400 QJ ..c E 300 :I 2 200 100 0 <;:) <;:)~ <::)<-,· <::)~ <;:)• <-,· USC Viterbi School of Engineering PhD Thesis Trip Mileage Distribution 8 <-,~ ~~ n~ ~~ n~ i;:,~ ~~ <::)"'). <::)~ <::)"'!, <::),,, <::),,, n~ ..., <::)· '°)• ~· n· ~· ~ "'). ~ "I, ,,, '>j miles Figure 5.5 Trip Mileage Distribution The information on EVs usage is shown in another way - trip mileage distribution, Figure 5.5. Most trip mileages are less than 10 miles, about 73% of the total number of trips. For trips of more than 10 miles, the trip number drops rapidly. Therefore, 10 miles is assumed to be a dividing line between short-distance usage and long-distance usage. 5.2.2 Comparison of all EV Users After each EV user's habit analyzed (detailed analysis for each EV user can be found in the full deliverables provided to DOE and LADWP), a comparison between all users is made in this section. In order to compare their EV using habits, trip number and the average miles per trip of each user were recorded during June 2014 to July 2015. The result is shown Table 5.1. Figure 5.6 and Figure 5.7 are the comparison between each EV user. Figure 5.6 presents the number of trip of each individual user, and Figure 5.7 displays the average miles for each individual user. After data filtered, there are 21 EV users in total. From Figure 5.6, it can be seen that User #6 is the heaviest EV user among 21 users, with 140 trips. The second most frequent driver is User #16, with 85 trips. Most of the users' had around 20 trips for the whole year. According to Figure 5. 7, it can be seen that average miles driven are mostly within 10 miles, and only 6 people drove more than 10 miles. 135 USC Viterbi School of Engineering PhD Thesis Table 5.1 Trip Number and the Average Mileage for one Trip of Each EV User Number Average miles Name of Trips per trip User #1 43 5.53 User #2 14 22.21 User #3 20 6.8 User #4 11 5.03 User #5 43 2.7 User #6 140 7.2 User #7 48 5.37 User #8 7 10.1 User #9 14 4.8 User #10 19 2.94 User#ll 20 18.2 User#12 13 24.2 User#13 13 9.1 User#14 43 7.2 User#15 14 2.6 User#16 85 3.56 User #17 13 18.3 User#l8 13 6.9 User#19 29 4.2 User #20 22 8.6 User#21 19 28.2 Trip Number 160 140 140 120 100 85 80 I 60 43 43 48 43 40 I I I I 29 22 14 20 11 14 19 20 19 20 7 13 13 14 13 13 I I 0 I I • • I I I I I I I I I Figure 5.6 Number of Trips for Each User 136 USC Viterbi School of Engineering PhD Thesis Average Miles per Trip 30 28.2 24.2 25 22.21 20 18.2 18.3 15 10.1 9.1 8.6 10 'i 'I',~ '1 ' 'I' I 1 ,~· l'i 6.9 5.53 11' I 5 I 2.6 3.56 0 I I Figure 5.7 Average Miles per Trip for Each User 5.3 LAD WP EV User Habit Analysis Based on FleetCarma 5.3.1 Methodology The analysis here is based on the source data sets 2 and 3 that are identified in Section 5 .1 and consisting of both the assigned and carpool EVs. In order to analyze the Carpool data, the following procedure is used: 1. Use the dispatch data to find the heavy users: Battery Electric Vehicle (BEV), Plug-in Hybrid Electric Vehicle (PHEV), and Internal Combustion Engine (ICE). But there are only few records of PHEV in dispatch data, so only the BEV cars and ICE car will be analyzed. (Heavy Users means that the users take a large number of trips in the whole year.) 2. After find the heavy users in dispatch data, the records in dispatch table should be located in FleetCarma data to find out the accurate number of trips, trip duration, start SOC and etc. 3. Select heavy BEV users # 1 to #5 according to the total mileages in dispatch data. Then plot distribution of the number of trips in 24 hours, Average miles of trips started in this hour. 4. Select heavy ICE users #1 to #5 according to the total mileages in dispatch data. Then plot the distribution of the number of trips in 24 hours and average miles of trips started in this hour. 5. Find heavy used BEV cars, and analyze the cars. 137 USC Viterbi School of Engineering PhD Thesis In order to analyze Assigned Car data the following procedure is used: 1. Use the FleetCarma data find the heavy assigned BEV car a) Heavy assigned BEV car user #1 to #5 b) Plot the number of trips in 24 hours, average miles of trips started in this hour, Sample route and Rout Map. 2. Use the FleetCarma data find the heavy assigned PHEV car 5.3.2 a) Heavy assigned PHEV car user # 1 to #5 b) Plot the number of trips in 24 hours, Average miles of trips started in this hour, Sample route and Rout Map. Overview of all EV Usage and Charging The EV usage information will be presented and analyzed in this section, including average trip details, equivalent MPG, electricity usage and etc. Since only limited EV usage data were logged during February 2015 and April 2015, the study analysis in this section will only focus on data recorded from May 20th 2015 until May 20th 2016. Table 5.2 lists EV journeys per vehicle. Table 5.3 and Table 5.4 list detailed information on LADWP EV usage by brand and type, respectively. Table 5.5 and Table 5.6 list EV usage on weekdays and weekends, respectively, and Table 5. 7 lists average EV usage on weekdays & weekends. Figure 5.8 presents average miles per vehicle in the weeks during May 2015 and May 2016. Table 5.2 EV Journeys per Vehicle Total Average Fuel Fuel Electricity Standard Idle EV# Distance Daily Efficiency Usage Usage Charge (mi) Distance( mi) (MPG) (gal) (%) (kWh) (kWh) 1 3522.4 36.7 81.76 20.1 25 775.6 968.1 2 8033.6 33.1 91.78 39.6 25 1616.4 1912.8 3 12118.5 73 53.4 225.6 12 44 91.4 4 4197.3 45.6 55.69 70.2 19 174 228.3 5 9637.1 47.7 54.02 150.6 16 937.4 1076.7 6 18850.5 83.8 42.63 414.5 17 933 1202.1 7 15003.7 54.2 53.09 242.5 11 1352.3 1670.4 8 3824.7 19.1 76.26 11.2 57 1312.4 1510.5 9 30302.2 168.3 38.24 761.6 13 1037.5 1351.6 10 3141 21.1 64.82 24.2 43 817.3 963.6 11 15255 81.6 48.71 259.1 18 1820.8 2148.7 12 1062.5 11.2 102.3 1.2 34 309.5 382.5 13 794.8 8.9 62.36 6.9 41 198.7 265.2 14 15391.7 119.3 35.77 417.3 21 435.6 598.1 15 776.6 10.2 66.55 4.6 43 237 326.8 138 USC Viterbi ' School of Engineering PhD Thesis 16 29124.5 137.4 43.58 631.9 39 1228.3 1523.1 17 5650.8 57.1 48.28 110.1 11 234.5 291.2 18 10193.2 60 49.96 185 18 639.8 847.1 19 401.4 3.1 104.62 0 82 129.3 171.8 20 411.8 6.8 109.63 0 76 126.6 175.8 21 125.7 6.6 112.41 0 98 37.7 66.7 22 495.1 9.7 123.54 0 96 135.1 158.4 23 470.1 9.6 109.45 0 47 144.8 166.7 24 1610.5 6.1 69.51 0 68 780.9 668.9 25 725.2 2.3 64.84 0 72 377 318.8 26 2534 17.6 131.61 0 43 648.9 771.3 27 4166.8 19.7 139.24 0 37 1008.7 1168.5 28 3848.6 17 130.64 0 40 992.9 1023.5 29 2845.5 17.9 131.17 0 39 731 .1 818.1 30 3546.5 18.2 133.22 0 40 897.3 953 31 2314.6 14.9 132.97 0 41 586.7 677.5 32 4181.7 19.4 128.06 0 38 1100.6 1169 33 610.7 10.5 129.65 0 54 158.8 175.7 34 2439.5 13.5 130.38 0 33 630.7 674.8 35 4739.7 22.8 138.97 0 34 1149.5 1235.8 36 3993.5 19 132.52 0 53 1015.7 1140.3 37 1234.3 21.3 133.02 0 40 312.8 341.5 38 1779.7 20 144.67 0 39 414.6 525 39 2705.7 15.3 138.45 0 38 658.7 693.1 40 2737 17.3 124.49 0 45 741 762.5 41 3334.2 19.7 134.09 0 41 838.1 887.6 42 682.3 5.9 97.92 0 54 234.9 284.3 43 713 12.5 98.45 0 23 244.1 203 44 7361.4 32 92.47 0 28 2683.1 1710.9 45 7796.5 32.2 94.26 0 28 2787.9 1755.9 46 7452.6 31.3 93.87 0 29 2676.1 1666.1 47 7282.7 31.5 93.04 0 29 2638.3 1664.6 48 2313.6 14.6 92.22 0 29 845.6 680.9 49 1098.4 10.7 89.13 0 35 415.4 304.7 50 3978.5 35.5 67.96 34.7 25 802.7 990.9 51 387.9 7.5 54.56 5.3 34 62.3 78.9 52 945.4 11.1 83.36 2.9 44 283.3 351.7 53 1320.3 27.5 67.06 11.7 31 270.3 328.1 54 2235.9 23.8 134.91 0 37 558.6 620.8 55 923.6 7.6 143.97 0 45 216.2 224.2 139 USC Viterbi ' School of Engineering PhD Thesis 56 4160.7 18.5 132.22 0 38 1060.7 1146.8 57 3163.7 18.3 131.37 0 55 811.7 894.9 58 4172.8 18.5 129.19 0 39 1088.6 1100.1 59 3242.1 17.2 134.36 0 40 813.3 880.5 60 4707.6 20.7 143.82 0 39 1103.2 1436.2 61 4704.8 21 135.44 0 38 1170.9 1228.9 62 4426.9 20.1 127.33 0 39 1171.8 1231.3 63 4847.4 21.4 140.31 0 36 1164.5 1345.4 64 4119.6 26.8 91.55 0 30 1516.7 996.9 65 7286.2 30.7 92.59 0 28 2652.3 1687.9 66 10359.9 52.3 98.39 0 22 3549 2447.5 From the above Table 5.2, following details can be summarized: The analysis is done on 66 EV vehicles. The highest amount of electricity usage is by Vehicle 45 using around 2,787 kWh travelling for 7796 miles for the duration of one year (May 20, 2015 until May 20, 2016). The average number of miles travelled by this car is 32 miles. The average electricity consumption by the above 66 cars is 836.50 kWh for the same one year period. Table 5.3 EV Journeys by Brand Average Average Electricity Electric Level Vehicle Total Daily Idle MPGeq Usage Distance 2 Brand Distance Distance (%) (mi) (mi) (kWh) (%) (kWh) Chevrolet Volt 3686 39 67 34 479 75 582 Nissan Leaf 1536 11 128 45 395 100 430 RAV4EV 2911 27 93 28 1057 100 685 Toyota Prius 3707 43 53 22 61 27 86 C-Max 6193 55 52 15 500 31 597 Ford Fusion 7825 75 49 24 385 19 487 Focus 204 8 84 63 63 100 73 Mitsubishi iMiEV 706 5 88 62 325 100 280 From the above Table 5.3, following details can be summarized: The vehicle brands are been compared and the highest average total distance here is by car Ford Fusion which travelled for 7,825 miles with 385 kWh electric usage and lowest 11PG equivalent of 49. Since there are several cars for each car model, the total mileage for the 140 USC Viterbi ' School of Engineering PhD Thesis model is the average among all the vehicles within that group. Table 5.4 EV Journeys by Type Total Average Electricity Level 2 Type Distance (mi) Daily MPGeq Idle(%) Usage (kWh) Distance (mi) (kWh) PHEV 8156 52 60.41 23.74 632 798 BEV 4435 23 114 42 726 563 From the Table 5.4, we can summarize the journeys by type i.e. PHEV & BEV. The PHEV travelled for 8156 total distance and consumed an "MPG equivalent of 60.41with632 kWh electricity usage, whereas BEV travelled for 4,435 miles and consumed "MPG equivalent of 114 with 726 kWh electricity usage. Table 5.5 EV Usage on Weekdays Daily Benchmark Average Average Total Daily Benchmark- Weekday Vehicle On Daily On Start End mi mi Daily mi/V eh hour Hours SOC% SOC% 1 30 3700 31 41 0 2 98 96 2 44 4628 28 41 1 2 86 67 3 47 4651 27 41 2 2 86 71 4 56 5176 26 41 1 2 84 63 5 53 7155 35 41 2 2 79 63 6 56 3628 27 41 1 2 81 63 7 52 6028 32 41 1 2 75 60 8 55 6878 34 41 2 2 73 57 9 56 6091 30 41 2 2 76 60 10 58 4486 27 41 1 2 78 67 11 56 6387 41 41 2 3 77 60 12 58 6440 31 41 1 3 80 67 13 60 7184 31 41 3 3 77 59 14 56 7655 33 41 2 3 75 59 15 57 6213 27 41 1 3 78 63 16 56 5694 29 41 1 3 79 59 17 56 6729 32 41 2 3 75 58 18 54 6354 29 41 2 3 78 63 19 56 6749 36 41 2 3 80 58 20 57 7101 32 41 2 3 78 59 21 56 5612 30 41 1 3 77 65 141 USC Viterbi ' School of Engineering PhD Thesis 22 63 6467 29 41 1 3 80 61 23 59 7516 33 41 1 3 79 64 27 61 5743 25 41 1 3 80 65 24 61 4550 24 41 1 3 82 66 25 60 5324 25 41 1 3 84 63 26 55 4485 28 41 1 3 82 66 27 61 5941 32 41 1 3 79 64 28 63 7227 30 41 1 3 79 62 29 59 6326 27 41 1 3 83 66 30 52 3319 24 41 1 3 76 68 31 52 3658 29 41 1 3 78 64 32 57 4984 24 41 1 3 79 65 33 58 7058 30 41 1 3 83 64 34 60 5779 28 41 1 3 83 67 35 60 7121 31 41 2 3 78 64 36 59 7326 32 41 2 3 80 66 37 57 6879 23 41 2 3 79 72 38 60 5662 33 41 2 3 82 74 39 61 4367 37 41 1 3 75 67 40 58 6248 29 41 1 3 83 74 41 63 7908 33 41 2 3 80 65 42 62 7044 33 41 2 3 80 64 43 59 4,900 31 41 2 3 82 63 44 63 5439 33 41 2 3 83 77 45 61 4773 31 41 2 3 79 72 46 62 8994 36 41 2 3 78 69 47 60 6775 32 41 2 3 80 67 48 63 5743 31 41 1 3 82 70 49 60 7284 31 41 1 3 81 75 50 61 8344 34 41 2 3 79 68 51 59 6027 29 41 1 3 74 63 52 60 4366 32 41 2 3 76 70 Average 58 6265 36 41 1 3 80 64 From the above table, Table 5.5, it can be summarized that on an average cars travelling for 6,265 miles have the average Start SOC 80% and average end SOC 64%. 142 USC Viterbi ' School of Engineering PhD Thesis Table 5.6 EV Usage on Weekends Ben chm a Daily Benchmark- Total Daily Start End Weekends Vehicle rk - Daily On Daily On Hours mi mi SOC% SOC% mi/Veh hour (hrs/day) 1 11 23 9 41 0 2 96 94 2 38 78 14 41 1 2 93 86 3 33 20 8 41 0 2 96 94 4 37 72 16 41 1 2 89 83 5 38 135 19 41 1 2 69 65 6 40 16 6 41 0 2 82 79 7 33 115 17 41 0 2 76 72 8 38 100 8 41 1 2 72 69 9 40 75 4 41 0 2 80 79 10 39 75 3 41 0 2 61 59 11 36 26 1 41 0 2 78 77 12 44 182 11 41 1 2 70 68 13 40 256 16 41 1 3 70 67 14 46 347 23 41 1 3 63 61 15 45 47 2 41 1 3 82 73 16 19 36 2 41 0 3 74 66 17 44 192 10 41 1 3 60 57 18 44 15 41 0 3 84 79 19 40 156 17 41 1 3 64 63 20 47 340 19 41 1 3 80 76 21 21 80 8 41 0 3 81 77 22 47 77 5 41 0 3 90 83 23 42 89 8 41 0 3 82 80 24 45 52 6 41 0 3 80 79 25 46 48 4 41 0 3 87 86 26 43 76 5 41 0 3 74 69 27 45 88 7 41 0 3 86 80 28 51 62 4 41 0 3 83 82 29 49 91 7 41 1 3 77 68 30 46 19 2 41 0 3 93 92 31 38 182 12 41 0 3 84 76 32 36 76 7 41 0 3 88 84 33 41 94 10 41 0 3 78 74 34 21 64 9 41 0 3 87 87 35 49 104 20 41 0 3 90 88 36 41 83 15 41 0 3 82 78 143 USC Viterbi ' School of Engineering PhD Thesis 37 40 65 9 41 0 3 84 75 38 30 77 13 41 0 3 83 78 39 49 146 16 41 0 3 82 67 40 43 37 5 41 0 3 91 88 41 44 68 11 41 0 3 89 86 42 44 38 6 41 0 3 76 72 43 50 37 9 41 0 2 83 77 44 53 30 8 41 1 3 85 80 45 39 20 7 41 0 3 81 76 46 48 24 6 41 0 3 80 75 47 50 22 9 41 0 3 74 69 48 48 88 14 41 0 3 85 81 49 51 34 11 41 0 2 93 91 50 48 23 4 41 1 3 89 87 51 49 20 5 41 0 3 82 74 52 43 17 8 41 0 3 84 78 Average 39 77 8 41 0 3 87 75 From the above table, Table 5.6, it can be summarized that the highest distance travelled (347 miles) by vehicles in week 14 had Average Start SOC 87% & Average End SOC 75%. Table 5.7 Average EV Usage on Weekdays & Weekends Benchmark Benchmark- Daily Total Daily -Daily On Start End Vehicle Daily On mi mi Hours SOC% SOC% miNeh hour (hrs/day) Weekday 58 6265 36 41 1 3 80 64 Average Weekends 39 77 8 41 0 3 87 75 Average From the above table, Table 5.7, the following details can be concluded: the average EV usage on week days is 58 vehicles traveling on an average 36 miles and weekends 39 vehicles travelling on an average 8 miles. 144 l 45 40 35 Q) u 30 1: Q) 25 > USC Viterbi School of Engineering PhD Thesis Average Daily Mileage in Different Weeks, May 2015 - May 2016 -;;;- 20 Q) ~ 5.3.3 5.3.3.1 15 10 5 0 1 3 5 7 9 1113 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 Number of Weeks ~Benchmark - weekend ~weekday Figure 5.8 Average miles/vehicle in the Weeks during May 2015 - May 2016 Overview of LADWP Assigned EVs and Carpool EVs Assigned EVs The total number of assigned car is 36 consisting of22 PHEVs and 14 BEVs. The following procedures have been developed in analyzing the assigned EVs. 1. For dispatch record, the bad data were deleted from LADWP EV Usage Survey from May 2015 to May 2016. Criteria to determine the problematic data are listed as follows: • Stop time is earlier than start time; • Stop mileage is less than start mileage; • Missing data on start time, stop time, start mileage or stop mileage; • Trips is canceled; • Trips mileage is zero; • The average speed is higher than 100 miles/hour. • The start time is earlier than May 16th 2015. Original all cars dispatch table had 18,022 pieces of data, but with removal of bad data, 16,355 left. Original non-EV dispatch table included 8,509 pieces of data, but with 145 USC Viterbi School of Engineering PhD Thesis removal of bad data, 3604 left. Original EV dispatch table contained 7,846 pieces of data, but with removal of bad data, 5795 left. 2. Filter for carpool FleetCarma record: • The electricity consumption is lower than 0.1 KWh. • The trip distance is shorter than 0.5 mile. • The idle is higher than 90%. Original file had 28,350 pieces of data, but with the removal of bad data, 19,585 left. 3. Filter for assigned car FleetCarma record: • The electricity consumption is lower than 0.1 KWh. • The trip distance is shorter than 0.5 mile. • The idle is higher than 90%. Original file had 23, 14 7 pieces of data, but with the removal of bad data, 11,921 left. 5.3.3.1.1 Assigned BEV s Table 5.8 Detail of Assigned BEV UnitNo Type of Car BEV/PHEV UnitNo Type of Car BEV/PHEV P44657 Ford Focus EV BEV P48091 Toyota RAV 4 EV BEV P44656 Ford Focus EV BEV P48086 Toyota RAV 4 EV BEV P44672 Nissan Leaf BEV W44659 Mitsubishi iMiEV BEV W44651 Nissan Leaf BEV W44660 Mitsubishi iMiEV BEV P44658 Ford Focus EV BEV W48085 Toyota RAV 4 EV BEV P44652 Nissan Leaf BEV P48092 Toyota RAV 4 EV BEV P44686 Nissan Leaf BEV W44661 Mitsubishi iMiEV BEV Detailed information on LADWP assigned BEVs can be found in Table 5.8. From Table 5.9, it can be seen that total number of trips is 4,567, total mileage is 23,690.87 miles, and average miles is 5.19 miles. In the overall analysis, data including missing information, e.g. no user name or other unnecessary information, were kept when filtering data. The first thing that is analyzed is the 24-hour distribution of trip start time with one-hour interval shown in Figure 5.9. 146 VI Q. a: I- ... 0 0::: w cc 2 ::J z 700 600 500 400 300 200 100 0 USC Viterbi School of Engineering 24-Hour Distribution of the Trip Start Time of BEV 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 TIME INTERVAL • Number of Start Trips in This Hour • average mileages in This Hour PhD Thesis 20 18 16 VI w 14 I!) ~ 12 w :::! 10 2 w I!) 8 ~ 0::: 6 w > 4 ~ 2 0 Figure 5.9 24-Hour Distribution of the Trip Start Time of Assigned BEV From Figure 5.9 above, it can be seen that most of the trips happen during 6am to 5pm and the BEVs are seldom used after 5pm, because that the BEVs are mainly used by assigned users during working time. Unlike non-EV vehicles, the BEVs are not used for after-work activities. Besides, it can be seen that most frequent driving are during I lam to 12pm. Additionally, there are over 700 trips during 8am to 9am, over 500 trips during 3pm to 4pm. Figure 5.10 The Distribution of Trip Mileages of Assigned BEV 147 USC Viterbi School of Engineering PhD Thesis Figure 5.11 The Pie Chart of Trip Mileages of Assigned BEV The second thing that is analyzed is the distribution of trip mileages shown in Figure 5 .10 and Figure 5 .11. It can be seen that the trip within 5 miles occupies the biggest part, around 70%. It can be concluded that BEVs are usually used for short trips by assigned users, which may be the result of battery limitation and work requirements. Furthermore, the figures present a small peak during 17.5 to 20 miles, over 500 trips, which can be assumed that BEVs are used for commuting by assigned car users. The figures also illustrate the term "range anxiety", that is, customers are worried about travelling long distance by using BEV 3000 2500 ~2000 a: ::1500 0 Cl: ~1000 2 ~ 500 0 The Distribution of Electricity Consumed (kWh} of BEV - ""'!!!"" - - - 0 1 2 3 4 5 6 7 9 ELECTICITY CONSUMED (KWH) OverlO Figure 5.12 Distribution of Electricity Consumed (kWh) of Assigned BEVs The third thing that is analyzed is the distribution of Electricity Consumed (kWh), which 148 USC Viterbi School of Engineering PhD Thesis is shown in Figure 5.12. It can be seen that most of BEVs assigned car users would like to consume less than 2 kWh per trip from Figure 5.12. It demonstrates that most of the BEVs are used for short trips and only a few trips using over 6 kWh electricity to travel a long distance. Another reason why the majority of trips consume only in 1 kWh is that, customers may just start their car but do not drive their car. Figure 5.13 Distribution of Start SOC of Assigned BEV s Figure 5.14 Distribution of End SOC of Assigned BEVs The fourth thing that is analyzed is the Distribution of Start SOC of BEV and the Distribution of End SOC of BEV, which is shown in Figure 5.13 and Figure 5.14. It can be seen that most of BEVs assigned car users would like to start to drive their car above 85% SOC from Figure 5 .13 and Figure 5 .14. These pictures also show that most users end their driving at 65% to 85%. It shows that the majority of the assigned car users are willing to 149 USC Viterbi School of Engineering PhD Thesis keep their car SOC over 50%, which indicates that customers worry about driving when their vehicle is under 50% SOC. Furthermore, it is seen that BEV car users charge their car even though it is above 50% SOC. 5.3.3.1.2 Assigned PHEV s The detailed information on LAD WP assigned PHEVs' usage is shown below, Table 5.9. Table 5.9 Detail of Assigned PHEV UnitNo Type of Car BEV/PHEV UnitNo Type of Car BEV/PHEV P44557 Ford Fusion Energi PHEV P44553 Chevrolet Volt PHEV P44548 Chevrolet Volt PHEV P44554 Chevrolet Volt PHEV P44536 Chevrolet Volt PHEV P44555 Ford Fusion Energi PHEV P44552 Chevrolet Volt PHEV P44542 Ford C max Energi PHEV P44541 Ford C max Energi PHEV W44537 Chevrolet Volt PHEV P44543 Ford C max Energi PHEV W44544 Chevrolet Volt PHEV P44547 Chevrolet Volt PHEV P44539 Toyota Prius Plugin PHEV P44556 Ford Fusion Energi PHEV P44540 Toyota Prius Plugin PHEV P44550 Chevrolet Volt PHEV W44546 Chevrolet Volt PHEV P44549 Chevrolet Volt PHEV P44551 Chevrolet Volt PHEV P44535 Chevrolet Volt PHEV W44538 Toyota Prius Plugin PHEV From Table 5.9, it can be seen that the total number of trips of all PHEVs is 7,353, total mileage is 94,793.18 miles and average mileage is 12.89 miles. 1200 1000 V'l c.. C2 800 I- LL 0 600 a:: w al 2 400 ::::> z 200 0 24-Hour Distribution of the Trip Start Time of PHEV 0 1 2 3 4 5 6 7 8 9 10 1112 13 14 15 16 17 18 19 20 21 22 23 TIME INTERVA • Number of Start Trips in This Hour • average mileages in This Hour 100 80 V'l w c:> ct w 60 :::! 2 w 40 c:> ct a:: w 20 > ct 0 Figure 5.15 24-Hour Distribution of the Trip Start Time of Assigned PHEVs 150 USC Viterbi School of Engineering PhD Thesis The first thing that is analyzed is the 24-hour distribution of trip start time with one-hour interval shown distribution in Figure 5.15. It can be seen that most of the trips happen during 6am to 5pm and the PHEVs, which is almost the same as the distribution ofBEVs. But what differs from BEYS is that assigned car users are more willing to use the PHEV to travel a relative long distance for after-work activities. Besides, it can be seen that most frequent driving are during 1 lam to 12pm. The average mileage of trip in each start hour is much longer than that of BEVs. This is because the PHEVs have an internal combustion engine besides its electricity mode. Figure 5.16 The Distribution of Trip Mileages of Assigned PHEV Figure 5.17 The Pie Chart of Trip Mileages of Assigned PHEV The second thing that is analyzed is the distribution of trip mileages shown in Figure 5 .16 151 USC Viterbi School of Engineering PhD Thesis and Figure 5 .17. It can be seen that the trip within 10 miles occupies the biggest part, around 70%. It can be concluded that PHEVs are usually used for short trips by assigned car users. People may usually use the electricity mode to drive the PHEVs. Furthermore, the figure shows a small peak during 15 to 17.5 miles, over 600 trips. The figure also shows that there are over 15% trips that are over 30 miles and 4% trips that are over 60 miles, which is totally different from the pie chart of BEVs. The reason for this phenomenon is that the PHEVs will be able to use their internal combustion engines to drive like a non-EV car. What is more, the PHEVs users have combustion engines as alternative energy resource, which eases the range anxiety of EV and makes the PHEV users willing to use the electricity mode to travel a longer distance than the users of BEV 3500 3000 ~ 2500 a: I- 0 2000 a:: w 1500 cc 2 :::I 1000 z 500 0 The Distribution of Electricity Consumed (kWh) of PHEV - - - - - - 0 2 3 4 5 6 7 8 9 ELECTICITY CONSUMED (KWH) - Over 10 Figure 5.18 The Distribution of Electricity Consumed (kWh) of Assigned PHEVs The third thing that is analyzed is the distribution of Electricity Consumed (kWh), which is shown in Figure 5.18. It is seen that almost the distribution charts of PHEVs and BEVs are similar to each other. 152 USC Viterbi School of Engineering PhD Thesis Figure 5.19 The Distribution of Start SOC of Assigned PHEV s Figure 5.20 The Distribution of End SOC of Assigned PHEVs The fourth thing that is analyzed is the Distribution of Start SOC of PHEV and the Distribution of End SOC of PHEV, which is shown in Figure 5.19 and Figure 5.20. It is seen that most of PHEVs assigned car users like to start to drive their car above 90% SOC from Figure 5.19. These pictures also show that most users end their driving at 0% to 5% SOC. It shows that the majority of the PHEV car users are willing to use up all the electricity in their trips, which indicates that customers feel free to drive when the SOC is low. It also reflects that the alternative energy resource ease the range anxiety. 153 USC Viterbi School of Engineering PhD Thesis Figure 5.21 The Distribution of Mileage/kWh for BEV and PHEV The mileage/kWh for BEV and PHEV have also been researched. From the Figure 5.21, it can be seen that the distribution of BEV and PHEV have the similar curve. Both curves of BEV and PHEV like a normal standard distribution with their peak at about 3 miles/kWh. Due to PHEVs have combustion engines as alternative energy resource, PHEV cars have some large value in mileage/kWh in the chart. 5.3.3.2 Carpool EV The detailed information on LADWP carpool BEVs' usage is shown below, Table 5.10. From Table 5.10, it can be seen that the total number of trips of BEVs is 19,585, total mileage is 118,792.52 miles and average miles is 6.06 miles. In the overall analysis, data including all the information, the first thing that is analyzed is the 24-hour distribution of trip start time with one-hour interval shown in Figure 5.22. It can be seen that most of the trips happen during 7am to 3pm. The time range is narrower than chart of BEVs in the assigned car users. This is because these cars are belonged to LADWP and carpool car users have to return these cars. So the carpool BEVs are seldom used during the night. Besides, it can be seen that most frequent driving are during 1 lam to 12pm. There are also many trips during Sam to 2pm. It may because people used carpool BEVs during work time. 154 UnitNo P44677 P44684 W44680 W44663 P44683 P44666 P44675 P44671 P44670 W44679 W44673 W44682 P44674 P44685 P44669 3000 2500 VI ~ 2000 I- 0 1500 a: w ~ 1000 :J z 500 0 USC Viterbi School of Engineering ' PhD Thesis Table 5.10 Details of LAD WP Carpool BEV Type of Car BEV/PHEV UnitNo Type of Car BEV/PHEV Nissan Left BEV W44681 Nissan Left Nissan Left BEV W44664 Nissan Left Nissan Left BEV P44668 Nissan Left Nissan Left BEV W44678 Nissan Left Nissan Left BEV P44665 Nissan Left Nissan Left BEV P44667 Nissan Left Nissan Left BEV P44676 Nissan Left Nissan Left BEV P44662 Nissan Left Nissan Left BEV P44653 Nissan Left Nissan Left BEV P48088 Toyota RAV 4EV Nissan Left BEV P48090 Toyota RAV 4EV Nissan Left BEV P48087 Toyota RAV 4EV Nissan Left BEV P48089 Toyota RAV 4EV Nissan Left BEV W48084 Toyota RAV 4EV Nissan Left BEV W48083 Toyota RAV 4EV 24-Hour Distribution of the Trip Start Time of Carpool Users n n n.., BEV BEV BEV BEV BEV BEV BEV BEV BEV BEV BEV BEV BEV BEV BEV 16 14 12 10 8 6 4 2 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 TIME INTERVAL VI w ~ <( w :::! 2 w ~ <( a: w > <( ... Average Milage for all Trips Start in This Hour • Number of Trip Start Time in Each Hour Figure 5.22 24-Hour Distribution of the Trip Start Time of Carpool Users 155 USC Viterbi School of Engineering PhD Thesis Figure 5.23 The Distribution of Trip Mileages of Carpool BEV s Figure 5.24 The Pie Chart of Trip Mileages of Carpool BE Vs The second thing that is analyzed is the distribution of trip mileages shown in Figure 5.23 and Figure 5.24. It can be seen that the trip within Smiles occupies the biggest part, around 70%. It can be concluded that Carpool BEVs are usually used for short trips. Furthermore, the figure shows there are no peak in the long distance, which is quite different from the distribution of BEV in assigned car. This is because all the carpool cars are mainly used for work purpose. Furthermore, almost 90% trips occur during the day time. 156 USC Viterbi School of Engineering PhD Thesis Figure 5.25 The Distribution of Electricity Consumed (kWh) of Carpool BEV s The third thing that is analyzed is the distribution of Electricity Consumed (kWh), which is shown in Figure 5.25. It can be seen that almost the distribution charts of assigned car BEVs and carpool BEVs are similar to each other. All the carpool users and assigned car users have the same energy consumption habits on BEVs. Figure 5.26 The Distribution of Start SOC of Carpool BE Vs 157 USC Viterbi School of Engineering PhD Thesis Figure 5.27 The Distribution of End SOC of Carpool BEV s The fourth thing that is analyzed is the Distribution of Start SOC of BEV and the Distribution of End SOC of BEV, which are shown in Figure 5.26 and Figure 5.27. It can be seen that most of carpool BEV users would like to start to drive their car above 90% SOC from Figure 5.26. These pictures also show that most users end their driving at 65% to 95% SOC. It shows that the majority of the BEV car users are willing to keep their EVs above 50% SOC, which indicates that customers feel the range anxiety. 5.3.4 Individual EV User's Habit Analysis The individual EV usage information have been presented and analyzed, including average trip details, energy consumption, charging information and etc. Since only limited EV usage data were logged during February 1st, 2015 to April 30th, 2015, the study analysis in this section will only focus on data recorded from May 2015 to May 2016. The Assigned BEV, Assigned PHEV and Carpool BEV have been discussed separately. In order to keep this thesis short and enjoyable to be read, detailed analysis for each individual EV user will not be presented here, interested readers can contact DOE or LADWP for the full deliverables. The comparison for all EV users can be found in the following section. 5.3.5 Comparison of EV and non-EV Due to the reason that there are only dispatch data forthe non-EV, the comparison will only be based on the dispatch data provided by LAD WP. Each category contains 5 heavy users to be analyzed. 158 USC Viterbi ' School of Engineering PhD Thesis 5.3.5.1 Comparison of ICE and BEV Users Table 5.11 The Comparison ofICE and BEV Users User Number of Total Average Number Trips Miles Miles User 1 95 15356 161.6 User2 80 10980 137.25 ICE User 3 5 10599 2119.8 User4 10 9256 925.6 User 5 6 8501 1416.8 Total 196 54921 280.2 User 1 69 2942 42.6 User 2 32 1508 47.1 BEV User 3 52 1507 29.0 User4 46 1315 28.5 User 5 30 1152 38.4 Total 229 8424 36.8 Compared with Carpool ICE users and assigned Carpool BEV users, the total numbers of trips of ICE and BEV top 5 users are similar, but user 3, 4, 5 of ICE drove their cars only few times, Table 5.11. The total mileage and average mileage ofICE users is far more than BEV users', probably because the BEV user had range anxiety so that they were afraid to drive a long trip in case of the dead battery. Since ICE users could easily find the gas station and they did not need to drive back to charging station, they preferred to drive for a long time. Among 5 ICE heavy users, they had different using habit. User 3, 4, 5 seldom used their cars. It might because that they used their gas station outside the parking lot mostly, the dispatch did record the trip. User 3, 4 preferred drove for a long trip since their trip duration was long. Among 5 BEV heavy users, their user habits were similar. It may be concluded that BEV and PHEV users can have different using habit due to the factors such as charging station. 5.3.6 Comparison of all EV Users 5.3.6.1 Comparison between Assigned BEV and Assigned PHEV There are 10 assigned car heavy users chosen from 33 users. After each 10 users' charging habit analyzed, a comparison between each user is made in this section. In order to compare 159 USC Viterbi ' School of Engineering PhD Thesis their using habits and charging habit, number of trips, total miles, average miles and total charged energy for each user are recorded during May 2015 to May 2016. The result is shown in Table 5.12. Table 5.12 The Comparison between Assigned BEV and Assigned PHEV User Number of Total Average Total Charged Number Trips Miles Miles Energy(KWh) User 1 920 1502 1.63 673 User2 750 10168 13.56 2409 BEV User 3 663 2361 3.56 671 User4 434 2222 5.12 658 User 5 264 400 1.52 325 Total 3031 16653 5.49 4736 User 1 975 3569 3.66 1475 User2 640 2733 4.27 946 PHEV User 3 568 7547 13.29 2125 User4 562 7302 12.99 1884 User 5 467 5111 10.94 1080 Total 3212 26262 8.18 7510 Compared with assigned BEV users and assigned PHEV users, they have similar number of trips, but the total miles are different. And PHEV users consume 7510 kWh which is 277 4 kWh higher than energy used by BEV user. The reason may be that BEV users worry about battery low power. So they do not use BEV for long distance trips and do not use BEV when battery is not fully charged. It can be seen that the average mile of BEV users is 5.49 miles which is much shorter than PHEV users' 8.18 average miles. However, for PHEV users, they use PHEV for long distance trips and in many trips they run of energy in battery. So they consume more energy and have longer mileages. Among 5 BEV users, they have different using habit. It may be caused by type of work and distance between home and work place. Although average miles of the 5 BEV users are small, average mile of user 2 is 13.56 miles. The long distance trips may lead the users charging car frequently. It can be seen that amount of energy consumed by user 4 is highest among 10 users. Among 5 PHEV users, average miles are different. PHEV userl, 2 and 3 prefer short trips. But average miles of user 4 and 5 are higher than 10 miles. They have similar using habit with BEV user 2. It may be concluded that BEV and PHEV users can have similar using habit when some factors such as distance between home and work place effects using habits. 160 USC Viterbi School of Engineering PhD Thesis 5.3.6.2 Comparison between Assigned BEV and Carpool BEV There are 5 assigned heavy used BEVs and 5 carpool heavy used BEVs. In order to compare usage and charging state, number of trips, total mileages, average mileages and total amount of charged energy for each user are recorded from May 2015 to May 2016. The result is shown in Table 5.13. Table 5.13 The Comparison between Assigned BEV and Carpool BEV Assigned BEV No.l No.2 No.3 No.4 No.5 Total Number of Tri s 920 750 663 434 264 3031 Total Mileage(Mile) 1502 10168 2361 2222 400 16653 Average Mileage(Mile) 1.63 13.56 3.56 5.12 1.52 5.07 Charged Energy(KWh) 673 2409 671 658 325 4736 Carpool BEV No.1 No.2 No.3 No.4 No.5 Total Number of Trips 1010 996 979 976 879 4840 Total Mileage(Mile) 7803 7215 7803 7353 7217 37391 Average Mileage(Mile) 7.73 7.24 7.97 7.53 8.21 7.73 Charged Energy(KWh) 1718 1676 1762 1680 1660 8496 5 assigned BEVs had 3031 trips. 5 carpool BEVs had 4840 trips. And carpool BEVs had more miles than assigned BEVs. So BEVs in carpool used more frequently than assigned BEVs. Although assigned BEVs can also be used at night, BEVs in carpool can be used by many different users. It may be the reason of more trips. Plus, average mileage of carpool BEVs was 7.73 miles. Average mileage of assigned BEVs was 5.07 miles. It may be concluded that assigned BEVs had more short distance trips. But carpool BEVs may be used more frequently for long distance trips when users had to use them. For carpool BEVs, users hardly drove them to home. As for amount of charged energy, although carpool 5 BEVs consumed more energy, KWh/mile of assigned 5 BEVs is higher. It may be caused by higher idle time. 5.4 The Study of EV User's Habit for Large Corporate Fleet - UCLA Users In this section, a study on EV user's habits for typical EV s that are employed at a university will be presented. It is assumed that the EV users in UCLA to be the representatives of this type of EV user. The data used in this analysis come from UCLA. 161 USC Viterbi School of Engineering PhD Thesis 5.4.1 Methodology At first, the ineffective data that need to be filtered out. Below are the standards: 1. The duplicated data will be deleted. 2. The data where "Start Main Power" is smaller than "End Main Power" will be deleted. For plot of distribution of plug-in start time and duration, if the time interval of two records is less than 30 minutes and the difference of "end power" and "start power" is less than 0.005, these two records will be considered as one charging event. After that, 10 heavy users will be selected as effective data to analyze individual EV user's habit. Then, a comparison of all UCLA users will be presented. 5.4.2 UCLA User's Behavior Analysis In order to keep this thesis short and enjoyable to be read, detailed analysis for each individual EV user will not be presented here, interested readers can contact DOE or LADWP for the full deliverables. 5.4.3 Comparison Average Charging Energy/Day of 10 Users in UCLA from 07/2014 - 08/2015 16.00 14.14 14.00 12.45 12.00 10.26 10.95 10.00 ..c 8.24 s 8.00 -"'. 6.00 3.81 3.54 4.00 I 2.24 2.60 2.00 I I I 0.00 User 1 User 2 User3 User 4 User 5 User 6 User 7 User 8 User 9 User 10 Figure 5.28 Average Charging Energy per Day of 10 Users in UCLA from 07/2014 - 08/2015 From Figure 5 .28, it can be seen that users 4 with Toyota RAV 4 had the highest charging energy. The reason why Charging energy per day of User 9 was low might be that Toyota Prius Plug-in has a small battery. User !(Chevrolet Volt), User 6 (Nissan Leaf), User ?(Chevrolet Volt) were all had large capacity battery, their average charging energy were high among 10 user. 162 USC Viterbi School of Engineering PhD Thesis Average Plug-in Time of 10 Users in UCLA from 07 /2014 - 08/2015 25 20 19.39 15 .... ::::5 0 :I: 10 8.18 5.56 5.50 I 6.30 5.47 4.46 I 4.78 5 I I I 2.71 I 3.25 0 I • I User 1 User 2 User3 User 4 User 5 User 6 User 7 User 8 User 9 User 10 Figure 5.29 Average Plug-in Time of 10 Users in UCLA from 07/2014 - 08/2015 From Figure 5.29, it can be seen that the average plug-in time of them were under 10 hours, they preferred to charge their EV s during the work time and finish the charging after work. User 3 had longest average plug-in time, probably because that user liked to plug the EV overnight. 5.5 The Study of EV User's Habit for Typical Owners in the City In this section, a study on EV user's habits for typical EV owners in the city will be presented. It is assumed that the EV users in the residential areas within the LADWP service territory to be the representatives of typical EV owners. The data used in this analysis come from LAD WP. In the residential areas within the LAD WP service territory, 223 EV rebate meter data were collected and analyzed to represent for typical owners in the city. Each meter reflects information of one home and will be recording 240V level 2 EV charging energy. The data used in this analysis was from February 1st 2016 to April 30th 2016. All meters recorded and updated the chargers' information every 15 minutes. The pilot data collected from these 223 rebate meters, Figure 5.30, includes more than 1 million records of EV charging. Top 10 meters that has the largest charging energy are selected to analyze. The results will be presented in the following subsections. 163 5.5.1 3lley USC Viterbi School of Engineering Angeles ~arional Forest Meter: 7 AED00009-01504913 EV Make: Zip Code: 90042 Downey -L PhD Thesis Az I West Cov -J~ Cityof ~ Industry El Segundo Manhattan Beach @ • © _ Fuller v- Torrance Rancho Palos Verdes Lakewood Anat C2016 GOOQle- Figure 5.30 The Distribution Map of Charging Meter Methodology 10 heavy users of EV have been chosen from 223 users to be analyzed as the individual user. They will be analyzed in following aspects: 24-hr Energy Distribution, Weekday vs Weekend Energy Distribution, Charging Duration Distribution, Charged Energy Distribution and 2160 Hours Energy Distribution. Only if charging time> 30 minutes, it is considered to be one charging. When the charging energy< 0.4 kWh per 15 minutes, it means that the charging has ended. 5.5.2 Typical Owner's Behavior Analysis In order to keep this thesis short and enjoyable to be read, detailed analysis for each individual EV user will not be presented here, interested readers can contact DOE or LAD WP for the full deliverables. 5.5.3 Comparison 10 heavy users have been chosen from 223 users. After top 10 heavy users' charging habit analyzed, a comparison between each user is made in this section. The heaviest user (User 164 USC Viterbi ' School of Engineering PhD Thesis # 1) was not selected because of the unusual behavior. In order to compare their EV charging habits, number of charging, charging time, charged energy, average charging time and average charged energy for each user are recorded during 2/1/2016 to 4/30/2016. The result is shown in Table 5.14. Table 5.14 Top 10 Heavy User Behavior Comparison of Typical Owners in the City Number of Charging Charging Time(Hour) Charged Energy(kWh) Average Charging Time(Hour) Average Charged Energy(kWh) b.I) c: .§ "' ..c: u :t: ... cu ..c E ::I z Use r#2 122 275 215 4.4 2.3 17.7 Use Use Use Use Use Use Use r#3 r#4 r#5 r#6 r#7 r#8 r#9 141 159 216 46 166 119 70 268. 447. 538. 235. 452. 167. 306 8 3 3 5 5 8 208 188 174 158 149 142 138 1.3 7.7 6.8 8.6 3.1 8.8 6.6 1.91 2.8 2.5 5.1 2.7 2.6 2.4 14.8 11.9 8.1 34.5 9 12 19.8 Number of Charging II ••• User tt2 User tt3 User tt4 User ttS User tt6 User tt7 User tt8 User tt9 User User ttlO till Users User User #10 #11 58 68 114.8 151. 3 1411. 1368 5 .6 2 2.22 24.3 20.1 Figure 5.31 Number of Charging Comparison of Top 10 Typical Owners in the City 165 > ~ cu c U.I "'C cu ~ "' ..c u USC Viterbi School of Engineering PhD Thesis Charged Energy(kWh} User #2 User #3 User #4 User #5 User #6 User #7 User #8 User #9 User User #10 #11 Users Figure 5.32 Charged Energy (kWh) Comparison of Top 10 Typical Owners in the City Average Charging Time( Hour} User #2 User #3 User #4 User #5 User #6 User #7 User #8 User #9 User User #11 Users #10 Figure 5.33 Average Charging Time Comparison of Top 10 Typical Owners in the City Average Charged Energy(kWh} , I 111 •I User #2 User #3 User #4 User #5 User #6 User #7 User #8 User #9 User User #10 #11 Users Figure 5.34 Average Charged Energy Comparison of Top 10 Typical Owners in the City 166 USC Viterbi School of Engineering PhD Thesis The Figures above, Figure 5.31 to Figure 5.34, are the comparison between each user. Figure 5.31 stands for the number of charging of each individual user. Figure 5.31 and Figure 5 .3 2 stand for the total amount of charged energy and total amount of charging hours in three months for each individual user. Figure 5.33 and Figure 5.34 stand for average charging time and average charged energy of each individual user. All of total amount of charged energy of 10 users are higher than 1300 kWh. But they have different charging habit. It can be seen in Figure 5. 31 that number of charging vary largely from user to user. User 6 only has 46 times of charging. However, user 5 has 216 times of charging. Since charged energy of them are similar, average charging time and average charged energy is different. It may be caused by infrequent charging or long distance trips. In addition, in Figure 5.34, average charged energy of user 5 and user 6 are smaller than 10 kWh even though they the total amount of energy charged in three months are 1746 kWh and 1588 kWh. It is caused by frequent charging habit. It may be caused by more time at home, preference of short trips or worry about low power. According to Figure 5.33, except user 6 who prefer charging infrequently, all of others have similar average charging time. However, compared with Figure 5.34, average charged energy for those users are different. It may be caused by different charging rate. It may be concluded that user 10 have higher charging rate than other users. 5.6 Conclusion 5.6.1 LADWP EV User Habit 5.6.1.1 Sunrey Analysis from June 2014 to July 2015 A study on one year survey of EV users inLADWP, from June 2014 to July 2015, has been analyzed. Below are two key issues noticed from the Survey: • For certain users, the EV usage data are very limited (<5 times), therefore, those data can't be used to analysis their user habits. • In the original data sheet, there were around 1,200 pieces of data. After the first data error cleaning (filtrating), 977 pieces of data were left. And after the second filtering, 700 were left. So the reliability of data recording should be improved so more data can be used for analysis. It can be seen that most of the trips happen during 6am to 5pm and the EVs are seldom used after 5pm, which is probably because that the EVs are only used by LADWP users during working hours. Therefore, unlike private EV users, the LAD WP assigned EVs are not used for after-work activities. What is more, it can be seen that the trip within half an hour occupies the biggest part of the data, around 40%, and within 2 hours is about 83%. 167 USC Viterbi School of Engineering PhD Thesis It can be included that EVs are usually used for short trips by users in LADWP. The perceived range limitations of EV is a key driver on how these vehicles are used .. The LADWP data shows that EVs are effectively used for commuting purposes with average distances around 10 miles. For instance, from 6am to 9am, the average distance is 9.2 miles, which is a reasonable average distance between office and home, and EVs may be used to drive to work. From 3pm to 6pm, the average mileage is 8 miles, which is similar to the distance from 6am to 9am. It is probable that the EV users are heading back to home. From 6pm to 9pm, the average mileage is 24.75 miles, which is the highest, but there are only a few numbers of trips, which means that the EVs are seldom used for that long trip and the EVs are possibly used for special purpose. The records used in this analysis are only from the dispatch data since no complimentary data was available from the FleetCarma that was installed in June 2015. In next section, survey data complimented with FleetCarma data are analyzed in details. 5.6.1.2 FleetCarma Analysis from May 2015 to May 2016 Due to the reason that many EVs have records on FleetCarma, the EV usage information analysis focused on data recorded from May 20th 2015 until May 20th 2016. (The FleetCarma was installed in June 2015.) The total number of assigned car is 36. Twenty-two (22) of them are PHEV, 14 of them are BEV. Compared with assigned BEV users and assigned PHEV users, they have similar number of trips, but the total miles are different. And PHEV users consume 7510 kWh which is 2774 kWh higher than energy used by BEV user. The reason may be that BEV users worry about battery low power. So they do not use BEV for long distance trips and do not use BEV when battery is not fully charged. It can be seen that the average mile of BEV users is 5.49 miles which is much shorter than PHEV users' 8.18 average miles. However, for PHEV users, they use PHEV for long distance trips and in many trips they run of energy in battery. So they consume more energy and have longer mileages. As for the Carpool EV, it can be seen that most of the trips happen during 7am to 3pm. The time range is narrower than the assigned EV s. This is because these cars belong to LAD WP and carpool car users have to return these cars. 70% trips of Carpool EVs within Smiles. It can be concluded that Carpool BEVs are usually used for short trips, which may be the result of work requirements. Furthermore, the figure shows there are no peak in the long distance, which is quite different from the distribution of assigned EVs. This is because all the carpool cars are mainly used for work purpose. And almost 90% trips occur during the day time. 168 USC Viterbi School of Engineering PhD Thesis Although there are many difference in carpool users and assigned car users, .It can be seen that almost the distribution charts of assigned car BEVs and carpool BEVs are similar to each other. All the carpool users and assigned car users have the same energy consumption habits on BEVs. And all the BEV car users are willing to keep their EVs above 50% SOC, which indicates that customers feel the range anxiety. 5.6.2 UCLA EV User Habit As for UCLA EV user behavior analysis, 229 EV users were selected and their user habits have been monitored. EV charging events were recorded from 07/2014 to 08/2015. Energy transferred for top 10 heavy users were studied and the plug-in start time and duration per charging were examined. The statistic results of individual users will be used to provide information and practical insights for utility companies to make city planning and future installations of the EV charging infrastructures. The result from EV charging energy transfer implement that the energy used for charging per day was mostly under certain amount, the energy distribution is different with different EV type. The utility should take all types of EV into consideration while making system planning. From Figure 5.29, it can be seen that user 1, 5, 6, 7, 9 preferred to plug in their EVs before work time. User 2 implemented overnight charging sometimes. From the comparison of 10 users, it may be concluded that the EV user who has a small capacity battery would charge for small amount energy every day, but the plug-in time can be long. 5.6.3 Typical Owner Habit In this part, 223 residential users are involved. Date is obtained from meters of those users. Those meters record consumed electricity every 15 minutes during 2/1/2016 to 4/30/2016. To account number of charging accurately, some date which are unrealistic are filtered. Filter criteria is mentioned in Section 5.5.1. Because the user who consumes largest energy has abnormal charging behavior, he/she is not analyzed. Other 10 users who consume more energy than others are chosen and analyzed in the full deliverable. To find charging habit of each user, number of charging, charging time, charged energy, average charging time and average charged energy for each user is shown in Table 5.15. 169 USC Viterbi ' School of Engineering PhD Thesis Table 5.15 Information of Typical Owners in the City Number Charging Charged Average Average of Time Energy Charging Charged Charging (Hour) (kWh) Time (Hour) Energy (kWh) User#l 68 151.25 1368.6 2.22 20.13 User#2 122 275 2154.4 2.25 17.66 User#3 141 268.75 2081.3 1.91 14.76 User#4 159 447.25 1887.7 2.81 11.87 User#5 216 538.25 1746.8 2.49 8.09 User#6 46 235.5 1588.6 5.12 34.53 User#7 166 452.5 1493.1 2.73 8.99 User#8 119 306 1428.8 2.57 12.01 User#9 70 167.75 1386.6 2.40 19.81 User#lO 58 114.75 1411.5 1.98 24.34 It can be seen in Table 5.15 that all of total amount of charged energy of 10 users are higher than 1300kWh. Although there is no trips information, according normal value of miles/kWh (around 3. 5miles/k:Wh), it may be concluded that all of users drive their EV not less than 4000miles in three months. High usage of EV leads to frequent charging and long charging duration. According Table 5.15, 6 users charged EV more than 100 times in three months. Averagely they charge EV more than 1 time per day. But since the average charged energy of those users are not more than 17. 66 kWh. Although 17. 66 kWh is not small, the capacity of the battery is higher than 60 kWh, which is much bigger than 17.66 kWh. Frequent charging habit may be caused by worry about low power or long time at home. Comparing date of users, some users have similar amount of charged energy but the total charging time may quite different. It may be caused by different charging rate. As for start charging time, since those users may have different jobs and time schedule, preference of start charging time varies from person to person. Compared date in Table 5.15, they have different charging habit. The numbers of charging vary largely from user to user. User 6 only has 46 times of charging. However, user 5 has 216 times of charging. Since charged energy for both are similar, user 6 has larger average charged energy than user 5. It may be caused by infrequent charging or long distance trips. In addition, average charged energy of user 5 and user 6 are smaller than 10 kWh even though the total amount of energy charged in three months are 1746 kWh and 1588 kWh. It is caused by frequent charging habit. It may be caused by more time at home, preference of short trips or worry about low power. According to Table 5.15, except user 6 who prefer charging infrequently, all of others have similar average charging time. However, average charged energy for those users are different. It may be caused by different charging rate. It may be concluded that 170 USC Viterbi School of Engineering PhD Thesis user 10 have higher charging rate than other users. There are two different charging rates. The reason may be that car and charger are communicating. So the rate of charging might change because of SOC or the condition of car or capability of car's changing circuit. Another possible reason may be that charging could happen for a portion of 15 minutes' interval. 171 USC Viterbi School of Engineering PhD Thesis 6. CHAPTER 6: EV Charging Demand Estimation and Forecast 6.1 Research Overview The main objective of this chapter is to estimate and forecast the EV charging demand based on historical EV charging records, as well as EV user behaviors. In this chapter, EV s are divided into three categories: • Industrial area • Commercial area • Residential area EV charging data and analysis done in chapter 4 and 5 have been summarized in the table below, Table 6.1. Table 6.1 Summary of EV Charging Data and Analysis of Chapter 4 and 5 Assi nedEV 36 05120116 Industrial 07/01114 - UCLA EV UCLA 17+ 06130115 Commercial LAD WP FleetCanna 30 05120115 - Carpool EV 05120116 LAD WP 200+ 02/01116 - Residential Residential LAD WP meter 04/30/16 meter info The actual data provided by FleetCarma contains charging information for 36 LADWP assigned EVs, and 30 LADWP carpool EVs, and they are used to represent Industrial and Commercial EV charging models. For Residential model, the only available data obtained from LADWP are the three months' meter data recorded every 15 minutes from the period of February to April 2016. Therefore, additional assumptions need to be made in order to forecast the Residential EV demand. The complete list of all EV models and techniques used can be found in the table below, Table 6.2. 172 USC Viterbi School of Engineering PhD Thesis Table 6.2 Summary of Models and Techniques used for Demand Estimation and Forecast Category Input Variable Technique Compact Industrial Midsize Number of SUV Monte Carlo times; Charging Events; I Compact Charging Start time; Number of EV s; Monte Carlo Residential Midsize Plug-in Duration; SUV Percentage of Start SOC Compact different EV class Commercial SUV From Table 6.2, it can be seen that different EV body classes are subcategories under each of the three main category. For example, there are compact, midsize and SUV under industrial area and the same subcategories under residential area. It can also be noticed that only compact and SUV body styles are listed under commercial area, this is because of data provided by LADWP do not contain midsize EV charging records. All the models constructed in this chapter are based on the data received from LADWP using same techniques. Additionally, only level 2 charging events are selected and analyzed because they account for more than 95% of total events. Furthermore, only weekday events are analyzed because their demand weighted more than 98%. 6.2 Industrial EV Charging Demand Estimation and Forecast Based on the historical EV charging record from FleetCarma, normalized seasonal comparison and weekday vs weekend plot for LAD WP assigned EVs can be found below, Figure 6.1. 173 s 12 ..>:: .s: 10 -0 ~ 8 E Cl) 0 6 Cl c: ·o, 4 ro ..c: 0 2 USC Viterbi School of Engineering PhD Thesis Seasonal Comparison of EV Charging Demand 2015-2016 (Normalized) - Summer15 - Autumn15 Winter1 5 - Spring16 - Summer16 - Autumn16 - Winter1 6 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day Weekday vs Weekend EV Charging Demand 2015-2016 (Normalized) - Weekday - Weekend > w 0 ~±=-3c==t===t::::~=±===±===c=:::t::=::::::J::==t:::::::;:::=:;::::::=;::=:=t~c:::=:L:::=::r::::::::::::r=~;;;;;;:::~~ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day Figure 6.1 Historical EV Charging Demand Comparison ofLADWP 36 Assigned EVs, Level 2 charging, 05/20/15 - 02/04/17 Figure 6.1 reflects two facts: 1. The LAD WP assigned EVs' charging demand curves under different seasons have similar pattern to each other, and the normalized demand amount at different hour is also very close under different seasons. Summer 2015 has the lowest demand amount and relative different demand pattern, and this is because FleetCarma started to record charging events on May 2015 but all LADWP assigned EVs were not actively participated in the first three months. 2. There were almost no charging happened in the weekends, most charging were made in the weekdays. Therefore, it is assumed that the charging demand on different seasons will follow the same pattern and have same amount at different hour. Moreover, the designed estimation and forecast model will only be used to forecast the weekday charging demand. Table 6.3 reflects the detailed information on the 36 LAD WP assigned EVs in 5 brands and 8 models. Table 6.4 displays the number of vehicles for each subcategory (body class), battery size and the decomposition of PHEVs and BEVs. 174 USC Viterbi School of Engineering PhD Thesis Table 6.3 Detailed Information on 36 LADWP Assigned EVs Number of Batten' Size Brand Model Class Tvpe · • Vehicles (kWh) Chevrolet Volt Economy PHEV 13 16 /Compact C-Max Economy Energi /Compact PHEV 3 7.6 Focus Economy BEV 3 23 /Compact Fusion Midsize Energi PHEV 3 7.6 Mitsubishi i-MiEV Economy BEV 3 16 /Compact Nissan Leaf Economy BEV 4 24 /Compact Toyota RAV4 SUV BEV 4 42 Prius Midsize PHEV 3 4.4 Table 6.4 Summary ofLADWP Assigned EVs under Different Body Class Batten' Size EV Class # ofEVs · Decomposition (kWh) Com act Midsize SUV 6.2.1 Methodology 26 6 4 12 62% PHEV + 38% BEV 6.2 100%PHEV 42 100%BEV In order to forecast the Industrial EV charging demand, Monte Carlo simulation, together with probability theory, statistics sampling and stochastic elements have been used and applied [65-75]. The training and testing dataset contain 70% and 30% of the available data obtained from FleetCarma. Four random variables are designed and generated, they are: Charging Event (CE), Charging Start Time (CST), Duration (D) and battery Start State of Charge (SSOC). For each random variable, multiple probability density functions (pelf) are approximated based on training data set and validations are made on the testing set to find out the pdf with the best goodness of fit indices. After that, all random variables will be generated using inverse transform technique [78,79]. Monte Carlo simulation will be applied afterwards to find out the average EV charging demand of a very large number of 175 USC Viterbi School of Engineering PhD Thesis iterations and Bootstrap technique will be used to discover the 95% confidence interval of the demand for each hour. The final results will be plot as a daily load profile in three ways using MATLAB: normal 2-D plot, 3-D (surf) plot, and improved 3-D (sutf with spline) plot. This section will describe the detailed methodologies used for demand estimation and forecast. 6.2.1.1 Training vs. Testing dataset The first thing is to divide the available raw data into training and testing sets. In the applied modern statistic learning or machine learning, it is important to separate the data into training and testing sets before designing and evaluating the model. The training set is a set of observations used for the computer to learn and fit the model, while testing set is used to evaluate how well the model has been trained and the model petformance. Typically, the training set contains 60% - 80% of the original data and the testing set is the remaining 20%- 40% [76]. In this thesis, the raw data ranges from May 20th 2015 to February 4th 2017. The first 70% of the raw data according to the time series has been selected as the training set and the remaining 30% is used to be the testing set. • Training Set 05/20/2015 - 08/20/2016 • Testing Set 08/21/2016 - 02/04/2017 6.2.1.2 Random Variables Random Variable is a set of possible outcome values from a random experiment. There are two types of random variables: discrete random variable and continuous random variables. If a variable can take any value between a given interval of numbers, it is called a continuous random variable, otherwise it is called a discrete random variable where its possible outcomes are countable. In this EV demand estimation and forecast model, four discrete random variables are designed and generated. They are: • Charging Event (CE) • Charging Start Time (CST) • Plug-in Duration (D) • Start State of Charge (SSOC) 176 USC Viterbi School of Engineering PhD Thesis Charging Event (CE) is the number of charging events for the EV in one day. In the model, weekday CE will be summarized for different subcategories, based on EV body class, from the historical data. For example, there are 26 Industrial Compact EVs and the maximum number of CE based on historical data is 4. Each of their weekday CE from 05/20/2015 to 02/04/2017 will be counted by removing all weekends and national holidays. Then a table will be generated to record the number of weekdays that contain 0, 1, 2, 3 or 4 CE. After that, averaged frequency and percentage of these 26 Industrial Compact EVs' CE will be calculated and used to represent for the future Industrial Compact EV charging behavior. Charging Start Time (CST) of an EV charging event can happen anytime between 0 am to 23:59 pm. The 24 hours have been divided into 96 time points where each point represents a 15-minute time interval. The number of CE for each EV subcategory that started in any of these 96 points will be counted and stored to be either in the training set or in the testing set. Then, the probability of each point will be calculated and a comparison distribution of training vs testing set will be made. For example, if one charging event of the 26 Industrial Compact EVs started between 0:00 am and 0: 15 am in 05/21/2015, it will be stored in the training set's point 0. When all the CEs have been counted, the distribution of training and testing set will be plotted. Plug-in Duration (D) can range from 0 to more than 5 hours based on historical charging records. In this model, D has been divided into 10 points with each point stands for the number of CEs in these half hour interval. For example, the number stored in point 0 means the number of events that had a plug-in duration between 0 and 30 minutes, while the number stored in point 10 represents the number of events that finished for more than 5 hours. Then, the probability of each point will be calculated and a comparison distribution of training vs testing set will be made. The last random variable is charging start SOC (SSOC). It can range from 0 - 100% and it is divided into 100 points where each point represents a 1 % interval. The number of charging events that had a SSOC within each 1 % will be counted and stored in this point. For example, the number recorded in point 0 means these charging events started when the EV's SOC was between 0 - 1 %. After that, the probability of each point will be calculated and a comparison distribution of training vs testing set will be made. 6.2.1.3 Probability Density Function (PDF) Probability Density Function (PDF) is a function that describes the relative likelihood that a continuous random variable will take on a given value. It is defined in the following formula: 177 USC Viterbi School of Engineering PhD Thesis f x2 P(xl ~ X ~ x2) = f(x)dx Xl Where • [ xl, x2] = Interval in which X lies • P(xl ~ X ~ x2) = Probability that X lies in this interval • dx = xl - x2 Based on the training distribution of the discrete random variables discussed above, pdf on the training dataset can be approximated by using the curve fitting toolbox in MATLAB. 6.2.1.4 Validation on the PDF Multiple pdf on the training dataset will be provided in MATLAB and the goal is to select the one with best goodness of fit indices. In this model, three goodness of fit indices have been selected [77], they are: • Coefficient of determination (R 2 ) • Root Mean Square Error (RMSE) • Mean Absolute Percentage Error (MAPE) Coefficient of determination (R 2 ) the variance in the dependent variable (Y) that can be explained by the model (approximated pelf). It can be calculated by the following equation: Where • SSE is the "error sum of squares" which quantifies how much all the data points Yi vary from the estimated value Yi • SST is the "total sum of squares" which quantifies how much all the data points Yi vary from the mean y Root Mean Square Error (RMSE) is the standard deviation of the residuals or prediction errors. It is commonly used in machine learning, forecasting and regression analysis. The formula of RMSE can be found below: 178 USC Viterbi School of Engineering PhD Thesis RMSE = Ll=1 (Yi - .Yi) 2 n Where • Yi is the observation i of the dependent variable • Yi is the predicted ith value Mean Absolute Percentage Error (MAPE) is a measure of the prediction method in statistics and is defined by the following equation: n 100I 1A· -Fi MAPE =-- i i n k i=1 l Where Ai is the actual observation and Fi is the forecast result. 6.2.1.5 Monte Carlo Simulation Monte Carlo simulation is a technique that used to model the probability of various outcomes in the process that contains random variables [78,79]. This technique is selected to repeatedly generate the four random variables, CE, CST, D and SSOC in the EV charging demand estimation and forecast model to simulate different EV charging events happen in a weekday. A very large number of iterations, 10,000, is specified in the model to ensure convergence [80,81]. The averaged value for each hour will be calculated to be the estimation and forecast result of the EV charging demand. The detailed algorithm can be found in Section 6.2.1.8. 6.2.1.6 Bootstrapping Confidence Intervals After Monte Carlo simulations have been done and the averaged estimation demand for each hour has been calculated, bootstrapping will be applied to find 95% confidence interval for the demand of each hour [82,83]. Bootstrap is a resampling technique that draws samples from the empirical dataset {y 1 , y 2 , ... , Yn} to replicate statistic T to find the sampling distribution. If the empirical data is uniformly distributed over {y 11 y 2 , ... , Yn} , then Bootstrap is nothing but drawing identically, independently distributed samples from{y 11 y 2 , ... , Ynl Bootstrap technique is better explained by Figure 6.2. 179 USC Viterbi School of Engineering PhD Thesis Original data Resampling Bootstrap Statistic 1 e(\\ ! ! ~'/>.ce'((\ xp> x~I) x~I) ••••• x<•> T 1 XI ~\\'(\1e\l n o$\'-> \(\'b (\ $ X2 ~ xf2) x~2) xf> ••••• x(2) Ti n ~3 Xu xp> x~3) x~3) ••••• x~) T3 x }B) x~B) x~B) ••••• x~B) TB Figure 6.2 Graph that Illustrate Bootstrap Technique In order to find the 95% Confidence Interval, 2.5% and 97.5% quantiles of the B replications need to be calculated as the lower and upper bound. 6.2.1.7 Assumptions • The 26 LAD WP Assigned Compact EV s charging behavior can be used to represent Industrial Compact EV charging behavior • All future Industrial Compact EV charging will follow the same distributions as CE, CST, D, and SSOC • All charging events will be using level 2 charging power at constant speed • The model is only designed to estimation and forecast weekday EV charging demand 6.2.1.8 Algorithm The complete algorithm can be found in the following flowchart, Figure 6.3. 180 No Charging Event CST_k(i,j) = O; O_k(ij) = O ; SSOC_k(ij) =O ; (where k = 0 top) USC Viterbi School of Engineering Input n1 = #of Monte Carlo simulation times Input n2 = #of EVs for demand forecast Generate CE(ij), number of Charglng Events in 24 hrs for ith EV in jth Monte Carlo Simulation (CE = 0, 1, ... , p based on historical data; i = 1 to n2; j = 1 to n 1) Yes Generate Charging Event k (CST _k(ij) follows Charging Start Time pelt, D_k(i,j) follows Puig-in Duration pdf ; SSOC_k(i.i) follows Start SOC pdf) that does not overlap with previous event(s); No Find actual charging demand E_k(i,j) = PhD Thesis min{ D_k(i,j)*l2_Power, (100%-SSOC_k(ij))*BatlerySize, BatterySize } (where k = 1 to p) Find actual charging duration O_actual_k(i, j) = E_k(ij) I L2_Power (where k = 1 top) Calculate the cumulative EV chargtng demand No Plot and ou1put the 24-hr EV charging demand profile to I ttl row of forecast matrix, Where i from 1 to n1 Yes Find 95% Confidence Interval using Bootstrap; Plot and output the averaged 24-hr EV charging demand prorne End No Figure 6.3 Complete Algorithm of EV Charging Demand Estimation and Forecast 181 USC Viterbi School of Engineering PhD Thesis Take Industrial Compact EV model as an example, the maximum number of CE is 4 based on historical data received from FleetCarma. The detailed algorithm is described below: 1. Enter n2 = number of Monte Carlo simulation times 2. Enter n1 =number of Compact EVs to simulate 3. For the ith EV, generate a random variable CE e If CE = 0, CST= D = SSOC = O; • If CE>= 1, generate charging event 1: CS Tl, D 1 and SSOCl • If CE>= 2, generate non-overlapping charging event 2: CST2, D2 and SSOC2 4. Charging Energy • If CE >= 3, generate non-overlapping charging event 3: CST3, D3 and SSOC3 • If CE = 4, generate non-overlapping charging event 4: CST4, D4 and SSOC4 100 -ssoc = min{Duration x PowerLevelz, 100 x Battery, Battery} 5. Actual Charging Duration= Charging Energy I Plug-in Duration 6. Based on step 4 and 5, distribute Charging Energy into 24-hr 7. Repeat step 3-6 for n1 times 8. Sum up n1 charging demands to be the estimation of jth Monte Carlo simulation for nl Industrial Compact EVs 9. Repeat step 2-8 for n2 times 10. Use the average of n2 results to be the demand estimation for n1 Industrial Compact EVs, and use Bootstrap technique to find 95% confidence interval of the charging demand for each hour 11. Plot Monte Carlo Simulation results (2D, Surf, Surf with Spline) In order to check if the newly generated charging events has overlap with the previous one, the following algorithm must be run and checked, Figure 6.4, Figure 6.5 and Figure 6.6. Basically, it has to meet one of the following two conditions: • The charging start time of the new event starts later than the previous ending time • The charging start time of the new event ends before the previous one starts 182 USC Viterbi School of Engineering When CE >= 2, run the algorithm in Figure 6.4 to test overlap issue. Start Generate SSOC_2 Generate CST_2, and D_2 Output CST _2, 0_2 and SSOC_2 End PhD Thesis Figure 6.4 Algorithm to Check if the Newly Generated Charging Events has Overlap with the Previous one When CE >= 2 When CE >= 3, run the algorithm in Figure 6.5 to test overlap issue. Start Generate SSOC_3 Generate CST _3, and D_3 Output CST_3. D_3 and SSOC_3 End No No Figure 6.5 Algorithm to Check if the Newly Generated Charging Events has Overlap with the Previous one When CE >= 3 183 USC Viterbi School of Engineering When CE= 4, run the algorithm in Figure 6.6 to test overlap issue. start Generate SSOC_ 4 Generate CST_ 4 , and D_4 Output CST_3, D_3 and SSOC_3 End No No No PhD Thesis Figure 6.6 Algorithm to Check if the Newly Generated Charging Events has Overlap with the Previous one When CE = 4 6.2.2 Industrial Compact EV Demand Forecast Model This section presents the summary and results of the Industrial Compact EV model, with 26 LADWP assigned compact EVs' result and analysis period 05/20/2015 - 02/04/2017. • Training Set: 05/20/2015 - 08/20/2016 • Testing Set: 08/21/2016 - 02/04/2017 6.2.2.1 Charging Event The summary of 26 industrial compact EVs charging events are summarized below, Table 6.5, Table 6.6 and Figure 6.7. 184 USC Viterbi School of Engineering PhD Thesis Table 6.5 Individual EV Charging Event oflndustrial Compact EVs, 05/20/15 - 02/04/17 251 161 21 2 191 138 70 20 14 396 37 2 0 0 282 102 43 7 1 319 93 19 3 1 296 132 6 1 0 308 86 36 5 0 333 92 8 1 1 161 136 101 31 5 412 20 2 1 0 105 159 130 31 9 278 81 67 7 2 123 224 76 10 2 103 269 61 1 1 350 71 12 1 1 329 71 33 2 0 241 162 27 4 0 391 35 9 0 0 140 229 64 2 0 282 119 34 0 0 226 141 61 7 0 296 89 40 7 3 251 165 19 0 0 145 82 143 47 13 265 117 43 7 2 61.0% 27.0% 9.9% 1.7% 0.5% 185 USC Viterbi School of Engineering Charging Event of Industrial Compact EVs (n=26) 05/20/15 - 02/04/17 PhD Thesis CE=O CE=l CE=2 CE=3 CE=4 #of Charing Event per Day • EVl EV2 EV3 EV4 EVS EV6 EV7 • EV8 • EV9 EVlO • EVll • EV12 EV13 EV14 EV15 EV16 EV17 EV18 • EV19 EV20 EV21 • EV22 • EV23 EV24 EV25 EV26 • Average Figure 6.7 Charging Event Distribution oflndustrial Compact EVs, 05/20/15 -02/04/17 Table 6.6 Individual EV Charging Event oflndustrial Compact EVs, 05/20/15 - 02/04/17 CE Average Average Frequency Percentage 0 265 61.0% 1 117 27.0% 2 43 9.9% 3 7 1.7% 4 2 0.5% 6.2.2.2 Charging Start Time Distribution of the historical charging start time data on training and testing set are shown below, Figure 6.8. 186 0.06 0.05 > 0.04 -~ Vl c ~ 0.03 0.02 0.01 USC Viterbi School of Engineering Industrial Model (Compact EV}: Charging Start Time Distribution Training vs Testing 05/20/15 - 02/04/17 PhD Thesis 0 al•~ -..• 11 •. . .. , LLl,uj 1 11111111 I 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day • train • test Figure 6.8 Charging Start Time Distribution of Industrial Compact EV s, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pelf model on CST training set is a combination of 3 Gaussian distributions, Figure 6.9. . .... - ___ , ... ........ , .ll'fll"ll•tor..ieilloM&'ll • ol"...i•-~ll ~i-tS11.-.-. al• OJMT~tll."'IJ,.,~1111 '1• &.IM1('LlU,Ull ~:=1=1 1 1~ ~= ;;:~~= d • o.COMtlllUICl1MG..0071111i1 H• 1U7(<11a,lt. ll) d• K.JJ41.AZ.M.Oll -· S$1_11-1''1 ·-"'"' ~~·Ull& '- .1 ....... ,1111:11•.l:)•. ·oJ·-•t.w.JY>ll [)c--..... 1 :~1 I :~= ~:·--I I 0 & 10 1$ 20 ..... if·" . .1:. ,, .1 .l 1 .. , .11;~f;~:::r J 1 : ·w· -!Pi! I.. I 11· I I', r . 1, ._ .... • n 1 ., • ~ l . " Figure 6.9 CST Curve Fitting on Training Set for Industrial Compact EVs The fitted pelf function for the training set is shown below: ( x-6.647) 2 (x-12.39) 2 (x-16.97) 2 f(x) = 0.04742e- 0.4718 + 0.02051e- 3.443 + 0.004491e- 16.33 Validation on the testing set and residual plot can be found in Figure 6.10. 187 6.2.2.3 USC Viterbi School of Engineering Charging Start Time Model (pdf) vs Actual Observation 0.06 ~---~----~---~----~-~ ~0.04 "iii c: CD c 0.02 5 10 Time of Day 15 -- CST Model + Actual + 20 Residual Plot Charging Start Time 0.01 ~---~----~--=--=-~----~-~ -0.02 ~---~~---~---~----~-~ 0 5 10 15 20 Time of Day PhD Thesis Figure 6.10 CST Validation on Testing Set for Industrial Compact EVs Plug-in Duration Distribution of historical plug-in duration on training and testing set are shown below, Figure 6.11. 0.18 0.16 0.14 0.12 > -~ 0.1 l5 0.08 0 0.06 0.04 0.02 0 Industrial Model (Compact EV): Plug-in Duration Distribution Training vs Testing 05/20/15 - 02/04/17 I I 1 1 II 0 0.5 1 1.5 2 2.5 3 3.5 4 Plug-in Duration (hr) • Train • Test ·- 4.5 Figure 6.11 Plug-in Duration Distribution of Industrial Compact EV s, 05/20/15 - 02/04/17 188 USC Viterbi School of Engineering PhD Thesis Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a combination of 3 Gaussian distributions, Figure 6.12. ~ ., • .,..,. .. ,Jol<IY':tl Ck--- ---· --- ... •1'..C-eb1V<ll':i:l •t2".,.C-C.Wlltl>'ll • d'...C•~·ll :::c C"""-....... 91,. __ •I• OUJI ..,._1-'CUO)l) ,,. u:in 11.mLJ1 1> d o llL~-.n•G.a1'1 "2 · 0 1jOJl(l.11tl.ll\Qll w. °"''"102oao1G> <1• UlllJl.l!IOJ.ZDJ:D ol o OllMfllD!Ul.llllilJI .. , • ..,..o..._u,_ d• O.S'ltll(l)&U.0.7'2tl _., .. SSl:JCl .. OI ·-- ~~~ !, " ~.,. i ::c ! " T """'.o E ·~ml°"'""-Ot_ ....... l =~ I·:~..____,_~ ___ _____..___, ·'.t '---'--~ : _i_---------.1. ..... _o Figure 6.12 Plug-in Duration Curve Fitting on Training Set for Industrial Compact EVs ( x-2.075) 2 (x-0.4716) 2 (x-3.588) 2 f(x) = 0.1331e- o.s632 + 0.1503e- i.091 + 0.1268e- o.s581 Validation on the testing set and residual plot can be found in Figure 6.13. Plug-in Duration Model (pdf) vs Actual Observation 0.2 .------.--=-----,.---.------.--=----'-.------.------,.---.-------. + 0.15 ~ + + -- D Model + Actual -~ 0.1 CD Cl a; :::J + 0.05 + o~-~-~--~-~--~-~-~--~-~ 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) Residual Plot Charging Duration 0.02 .-------.----.---.----.-------,.~'""--r'~--.---.----.------.----. 0.01 ~ o~-- ~ -0.01 -0.02 '-------'-- 0 0.5 1_5 2 2_5 3 3_5 4 4_5 Plug-in Duration (hr) Figure 6.13 Plug-in Duration Validation on Testing Set for Industrial Compact EVs 189 USC Viterbi School of Engineering PhD Thesis 6.2.2.4 Start SOC Distribution of historical charging start SOC on training and testing set are shown below, Figure 6.14. 0.35 0.3 0.25 ;:- 0.2 ·;;; c ~ 0.15 0.1 Industrial Model (Compact EV} Start SOC Distribution Training vs Testing 05/20/15 - 02/04/17 0 3 6 9 121518212427303336394245485154576063666972757881848790939699 Start SOC Interval{%) • Train • Test Figure 6.14 SSOC Distribution oflndustrial Compact EVs, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a combination of 2 Exponential distributions, Figure 6.15. - __ .,,., 1!4···....-. ....... "'I ~-(Wltll--._.,, •• UHJ fl lOlllt.ll>l'l •• Ul'l l UOJ '~ (. o.ooust IC.OCIM J,UOXIMl •• 001 .. l50117l•OU11J -·k Sst;OOOIJOt •-o"" :"~0-t .,L ! "' f·'.r .... .. -.._ssoc • -*IOC-.. -MOCIJ -~-~UOCTIWI 1 ·:c - · ._.._ .... _,,.__-·-- - · 1 1 ~I 0015 i ! .::.. · , .. f ... ,, ' ' I : ' ' , ' f. '" '" . j ! .. :~ I ' I ,·' ' : I h I ' I' ' I I I I ' ! I 'I I " I :· 111111 I . . . . - .._...._ssoc Figure 6.15 SSOC Curve Fitting on Training Set for Industrial Compact EVs f(x) = 0.3172e- 2 · 279 x + 0.002259e 0 · 01843 x 190 USC Viterbi School of Engineering PhD Thesis Validation on the testing set and residual plot can be found in Figure 6.16. 6.2.2.5 ~ ~ 0.2 0.1 + o~++H-4++-4J......._...;...4.i-lo4-.).+.l..l...,.._>-4.J.-.j..4"'-l-!J"""""~~~~~~~h±!~bld±~±l±l:±i±±:!:;!::.J 02468W12MWW~~N~~~~~$~~~~~"W~M~MOO~MOOOOronNnnoo~MOOOOOO~MOOOO Charging Start SOC (%) 0.02 -0.02 0 2 4 6 8 1012141t518202224262830323436384042444e46505254565860626466687072.74767860828486819092949698 Charging Start soc (%) Figure 6.16 SSOC Validation on Testing Set for Industrial Compact EVs Goodness of fit Summary of the three goodness of fit indices can be found in Table 6.7. Table 6. 7 Goodness of Fit Indices Summary oflndustrial Compact EV Model Goodness of fit Index Train Test R-s uare 93.56% 83.43% CST RMSE 0.0027 0.0041 (Gaussian 3) MAPE 0.1429 0.1176 R-square 99.96% 93.72% Duration RMSE 0.0026 0.0121 (Gaussian 3) MAPE 0.0246 0.1105 R-square 98.69% 97.30% Start SOC RMSE 0.0037 0.0055 (Exponential 2) MAPE 0.0716 0.1221 6.2.2.6 Monte Carlo Simulation Result When number of Monte Carlo Simulation iterations is set to be 10,000 and number of Industrial Compact EV number is set to be 10,000, the results of EV charging demand estimation and forecast are provided in the following figures, Figure 6.17 to Figure 6.20. 191 USC Viterbi School of Engineering PhD Thesis sJilistogram of Monte Carlo Generated Charging Start Time Variables 1200 Histogram of Monte Carlo Generated Plug-in Duration Variables 400 ,., g 300 " ~ 200 LL 100 1 2 3 4 s e 1 8 910111213 141s1e11151920212223 Time of Day 1000 g 800 Q) ~ LL 600 400 200 0 ·1 ,,, , ff• ' I r" lj I] . 0 6 Plug-in Duration 150 6listogram of Monte Carlo Generated Charging Duration Variables Histogram of Monte Carlo Generated Charging Start SOC Variables 4000 -y-- 3000 ,., 1000 g g " " e ~ 2000 e LL 500 LL 1000 o~~~~~~~~~~~~~---~~~~~~ 0 l . ... H - -•H••· ..1...-- -· .......... · - ·-••0 """-_,,0.,...0-&0,.M'0 .......... 0,0.H~00' .... 000_U_U ...... -1 Charging Duralion o w ~ oo oo m Battery SOC (%) Figure 6.17 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for Industrial Compact EV 2500 ~~~- '"- ~~ •t_ ria_ lr Com ~ p•~ ct_ E~ V_ C~ ---r 'g~ in~ g~ ~•- •_ •ir 2)_ ~ _m_ a_ nd~ F_ «_ •~ ~• - t_ Bar St _ d~ •~ nM _~ ~ t•_ Ca~ ri_ o_ mr mu _ la_ t~ ~ "~ (1~ 00 _ 0_ 0~ EV_ s,_ 1r OO_ OO_ tlr mt _ s~ )~~~ 2000 .... H !!' e ~ HIOO "' > w o~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~ 0 10 11 12 1 3 14 15 16 17 18 19 20 21 22 23 Time of Day Figure 6.18 Monte Carlo Simulation Result for Industrial Compact EV Demand Model 192 250() .• _ ·- 10000 •• USC Viterbi School of Engineering a..t Plot of htdultrill Compact EV Cl'IWVlnfl Demand For.ult Band on Mont.- cam Simutation (1ooot EV• . 10000 ti!M•) _ ____ _ ...,,::· PhD Thesis ----<-- ------- :711 Figure 6.19 Monte Carlo Simulation Result for Industrial Compact EV Demand Model - Surf Plot in MATLAB Surf Pio! with Splineollndu1trial ConlpKt EV Ch,.rgln(i Oetnand fofKnt BaHClm Monti C8tlo llmulatioii C10000 EYs. 10000ti,,_) -------· . --c::··· Figure 6.20 Monte Carlo Simulation Result for Industrial Compact EV Demand Model - Surf with Spline Plot in MATLAB Averaged EV demand forecast in 24 hour format of 10,000 Industrial Compact EVs running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table 6.8. 193 USC Viterbi School of Engineering PhD Thesis Table 6.8 10,000 Industrial Compact EV Charging Demand Estimation and Forecast Result .. - - - .. - - - - - MliM ••• ••• •• ••• •• •• ··- MIM •• Wj1M WJM W}W WJM 336.9 276.9 270.1 293.3 327.3 370.5 1055.7 1695.0 1311.9 1234.1 1354.8 1647.6 1905.2 2002.3 1915.2 1681.8 1380.3 1093.5 874.8 734.8 656.8 615.4 591.6 528.9 Yearly Demand Upper (kWh) 336.3 337.5 276.3 277.4 269.6 270.6 292.8 293.9 326.7 327.9 369.9 371.1 1054.8 1056.6 1693.7 1696.4 1310.7 1313.0 1233.1 1235.2 1353.7 1355.8 1646.4 1648.8 1903.9 1906.6 2001.0 2003.7 1913.9 1916.7 1680.4 1683.0 1379.1 1381.4 1092.4 1094.6 873.8 875.7 733.9 735.6 655.9 657.6 614.6 616.2 590.9 592.4 528.1 529.6 6.30 GWh 6.2.3 Industrial Midsize EV Demand Forecast Model Midsize EV demand estimation and forecast model for the Industrial area uses the same methodologies and algorithm as Industrial Compact model. In order to keep this thesis short and enjoyable to be read, detailed analysis and results for this model will not be presented here, interested readers can jump to Appendix A for more information. 194 USC Viterbi School of Engineering PhD Thesis 6.2.4 Industrial Electric SUV Demand Forecast Model Electric SUV demand estimation and forecast model for the Industrial area uses the same methodologies and algorithm as Industrial Compact model. In order to keep this thesis short and enjoyable to be read, detailed analysis and results for this model will not be presented here, interested readers can jump to Appendix B for more information. 6.2.5 Complete Industrial EV Demand Forecast Model Complete Industrial EV Charging Demand Estimation and Forecast model will be present in this section, which is an aggregated model of Industrial Compact, Industrial Midsize and Industrial SUV model. The complete model could be used to forecast the Industrial EV charging demand of Los Angeles, California and USA for the future. This thesis is conducted under DOE and LADWP SGRDP and will only be used for a demonstration purpose. Therefore, the author selects 10,000 Industrial EVs with 34% compact EVs plus 33% midsize EVs plus 33% Electric SUVs to be the input of the model to estimate the EV charging demand. The Monte Carlo Simulation iteration number is also set to be 10,000 to ensure convergence. The results can be found in the following figures, Figure 6.21 to Figure 6.24. Histogram of Monte Carlo Generated Charging Start Time Variables 400 I I I I I r I I I T 1 r I r r TimeofDay Histogram of Monte Carlo Generated Charging Duration Variables 2000 I I 1 I 1500 [;' Histogram of Monte Carlo Generated Plug-in Duration Variables 2000 I r I r I 1500 >. ~ 5- 1000 !!! u. 500 0 ·1 3 Plug-in Duration 5 6 Histogram of Monte Carlo Generated Charging Start SOC Variables 3000 ' ' ' 2500 >. 2000 " ~ 1000 CT c: ~ 1500 !! u. u. 1000 500 500 o~~~~~~~~~~~~~~~ oll..-~--'~~--'-~~~.,,. , ~,,~,,,~,,ll=llr•*'U"~'~'~~ ·1 0 0 20 40 60 80 Charging Duration Battery SOC (%) Figure 6.21 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for 10,000 Industrial EVs 195 100 USC Viterbi School of Engineering PhD Thesis Industrial EV Charging (Level 2) Demand Forecast Based on Monte Carlo Simulation 3 500 .------.----,-~-,--r-----,.------.-~ (1T OO_ O_ OEV ~ s:_ 3r 4% ____, Co_ m~ pa~ c- t+_ 3T 3%~ Mr ids_ iz_ er +- 33_ %_ S_ U~ V,_ 10_ 00 T O_ ti_ mer s~ ) -r-~~-r---.-~,..-~--, ~ c "C 3000 2500 c 2000 ! g> -~1500 ~ (.) > w 1000 500 10 11 12 1 3 14 15 16 17 1 8 19 20 21 22 23 T ime of Day Figure 6.22 Monte Carlo Simulation Result for Industrial EV Demand Model (10,000 EVs) ... "'"' ~ .. ~ 2000 ! f "'"' v ~ 1000 ... ~If Plot ol lndullrlal EV Char;ln9 Dtmand Forecast lu.d on M onie Catto Slmulatlon (10000 EV1: 34% Co!r.pacl • ll 'Ill Mldsl.u • 33% SUV, IOOOO tlmes) Moriocatlosr..,auonlrTIQ -----.....:.· " " ,. Figure 6.23 Monte Carlo Simulation Result for Industrial EV Demand Model - Surf Plot in MATLAB (10,000 EVs) 196 USC Viterbi School of Engineering Surf Plot wllh Spllne ol lndustrbll Co1111pkt EV Ch•~ng Otm.lld Forecatt BaMd on Mon~ Carlo Simui.tion (10000 EV1: 3' % Comp.ct+ l3 'II. MidMz. + 33 % stN, 10000 time1) PhD Thesis " Figure 6.24 Monte Carlo Simulation Result for Industrial EV Demand Model - Surf with Spline Plot in MATLAB (10,000 EVs) Averaged EV demand forecast in 24 hour format of 10,000 Industrial EVs (34% Compact + 33%Midsize + 33% SUV) running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table 6.9. 197 USC Viterbi School of Engineering PhD Thesis Table 6.9 10,000 Industrial EVs Charging Demand Estimation and Forecast Result 1111 - ------------ - 104.9 104.6 105.3 ---~-------- - 129.4 129.1 129.8 ~~~~:~~~-=~~~~~~ .. 179.4 178.9 179.8 ---~-------- - 507.3 506.7 508.0 - -======~:-======~====~ 1714.7 1713.5 1716.0 2524.6 2528.2 - 2526.5 -=====~:-====~===': - 2631.1 2629.3 2633.1 1879.4 1882.7 - --=====~~====~~====~ 1881.0 MuM 1588.6 ••• 1768.0 ••• 2385.5 •• 2635.0 ••• 2135.0 -- 1849.4 •• 1813.5 .,. 1369.8 Ml:M 810.6 epw 624.5 MJM 346.7 MJW 216.1 MJM 211.3 Ycarh Demand 1587.1 1766.4 2383.7 2633.1 2133.4 1847.8 1812.0 1368.6 869.6 623.6 457.8 346.1 276.1 216.8 198 1590.0 1769.6 2387.3 2636.9 2136.7 1850.9 1815.0 1371.0 871.6 625.3 459.1 347.3 277.2 217.9 7.26 GWh USC Viterbi School of Engineering PhD Thesis 6.3 Commercial EV Charging Demand Estimation and Forecast Based on the historical EV charging record from FleetCarma, normalized seasonal comparison and weekday vs weekend plot for the LAD WP carpool EVs can be found below, Figure 6.25. .5 -0 ffi 10 E Q) 0 $: 20 ..:.:: .5 ~ 15 <ll E Q) 0 10 Cl c: ·~ ~ 5 u Seasonal Comparison of EV Charging Demand 2015-2016 (Normalized) Time of Day - Summer1 5 - Autumn15 Winter1 5 - Spring16 Summer1 6 - Autumn16 Weekday vs Weekend EV Charging Demand 2015-2016 (Normalized) - Weekday - Weekend > w OL_J__!__l__l___l~i=:::i__!__J__l___J_~'-="'=='==l=='==-"~L_J__l__J___:c==i 0 1 2 3 4 5 6 7 8 9 1 0 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day Figure 6.25 Historical EV Charging Demand Comparison of LAD WP 30 Carpool EVs, Level 2 Charging, 05/20/15 - 02/04/17 Similar to Section 6.2.2, Figure 6.25 reflects two facts: 1. The LADWP carpool EVs' charging demand curves under different seasons have similar pattern to each other, and the normalized demand amount at different hour is also very close under different seasons. Summer 2015 has the lowest demand amount and relative different demand pattern, and this is because FleetCarma started to record charging events on May 2015 but all LADWP carpool EVs were not actively participated in the first three months. 199 USC Viterbi School of Engineering PhD Thesis 2. There were almost no charging happened in the weekends, most charging were made in the weekdays. Therefore, it is assumed that the charging demand on different seasons will follow the same pattern and have same amount at different hour. Moreover, the designed estimation and forecast model will only be used to forecast the weekday charging demand. Table 6.10 reflects the detailed information on the 30 LADWP carpool EVs in 2 brands and 2 models. Table 6.11 displays the number of vehicles for each subcategory (body class), battery size and the decomposition of PHEVs and BEVs. Table 6.10 Detailed Information on 30 LADWP Carpool EVs Nissan Leaf BEV Toyota RAV4 SUV BEV It is noticed that there are not any carpool midsize EVs used in LADWP. Therefore, only compact EV and Electric SUV models were constructed to forecast the demand of the commercial area. Table 6.11 Summary of LADWP Carpool EVs under Different Body Class Batten· Size EV Class #of EVs • Decomposition (kWh) ComIJact 26 II 24 100%BEV SUV 4 42 100%BEV 6.3.1 Methodology In order to forecast the Commercial EV charging demand, same methodology are used as the Industrial EV Monte Carlo simulation, together with probability theory, statistics sampling and stochastic elements have been used and applied. The training and testing dataset are chosen 70% and 30% from the available data obtained from FleetCarma. Four random variables are designed and generated, they are: Charging Event (CE), Charging Start Time (CST), Duration (D) and battery Start State of Charge (SSOC). For each random variable, multiple probability density functions (pelf) are approximated based on training data set and validations are made on the testing set to find out the pdf with the best goodness 200 USC Viterbi School of Engineering PhD Thesis of fit indices. After that, all random variables will be generated using inverse transform technique. Monte Carlo simulation will be applied afterwards to find out the average EV charging demand of multiple simulations and Bootstrap technique will be used to discover the 95% confidence interval of the demand for each hour. The detailed methodology can be found in Section 6.2.1. 6.3.2 Commercial Compact EV Demand Forecast Model Compact EV demand estimation and forecast model for the Commercial area uses the same methodologies and algorithm as Industrial Compact model. In order to keep this thesis short and enjoyable to be read, detailed analysis and results for this model will not be presented here, interested readers can jump to Appendix C for more information. 6.3.3 Commercial Electric SUV Demand Forecast Model Electric SUV demand estimation and forecast model forthe Commercial area uses the same methodologies and algorithm as Industrial Compact model. In order to keep this thesis short and enjoyable to be read, detailed analysis and results for this model will not be presented here, interested readers can jump to Appendix D for more information. 6.3.4 Complete Commercial EV Demand Forecast Model Complete Commercial EV Charging Demand Estimation and Forecast model will be present in this section, which is an aggregated model of Commercial Compact and Commercial SUV model. The complete model could be used to forecast the Commercial EV charging demand of Los Angeles, California and USA for the future. This thesis is conducted under DOE and LADWP SGRDP and will only be used for a demonstration purpose. Therefore, the author selects 10,000 Commercial EVs with 50% compact EVs plus 50% Electric SUVs to be the input of the model to estimate the EV charging demand. The Monte Carlo Simulation iteration number is also set to be 10,000 to ensure convergence. The results can be found in the following figures, Figure 6.26 to Figure 6.29. 201 >- 0 c: Q) ~ c;r e u. USC Viterbi School of Engineering PhD Thesis 1 oJ6istogram of Monte Carlo Generated Charging Start Time Variables 4000 Histogram of Monte Carlo Generated Plug-in Duration Variables 800 600 400 200 0 0 1 2 3 4 5 6 7 8 9 1011 121314151617181920212223 Time of Day 3000 i;' c: Q) ~ 2000 ~ u. 1000 O '-----_...._.......,__,c-.~•...._-. ......... ~_._-----'--------' ·1 0 Plug-in Duration sJlstogram of Monte Carlo Generated Charging Duration Variables 4 Jillstogram of Monte Carlo Generated Charging Start SOC Variables 4000 1000 O'----....,__.~-...-..... ..._~~'-----~-----'-----' ·1 Charging Duration >- 0 c: 300 l ~ ---""'unMHllOO~· d 20 40 60 Battery SOC (%) BO 100 Figure 6.26 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for 10,000 Commercial EVs Commercial EV Charging {Level 2) Demand Forecast Based on Monte Carlo Simulation 6000~----~----~~--~~--~ (_ 1o_ oo_ o_ e_ vs _ :_ so _·_ ~_ co _m ~p- •_ ct_ +_ ~ _~ _ ._ su _v ~ ,~ 10_ 0_ 00_ 1_ im_ •• ~ )--~~~----~----~--~ 5000 ~ 4000 c -g ~ ~ Cl 3000 g> · ei .1! <..> i;j 2000 1000 o l........L_J_~.L---'---"""'!!!"".::.J.~1-_!__JL_--'---_J_~1-_1__J~--'---_j_~L__L_j~_1____:r:~L_~_J 0 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day Figure 6.27 Monte Carlo Simulation Result for Commercial EV Demand Model (10,000 EVs) 202 ~ ·coo-.. . ~ ~- ~ L ., ~ um~. o~ 10000 USC Viterbi School of Engineering Surf Plot of Comm«cial EV Chlrging Demand foncut Butd on Monte C1rlo SimulMiort (10000 EV1; 50% Compact+ SO% $UV, 10000 timM) .. .-::.···· -----· •5 .. --......:·· " PhD Thesis Figure 6.28 Monte Carlo Simulation Result for Industrial EV Demand Model - Surf Plot in MATLAB (10,000 EVs) 5000·-. 4000 -~. Sufi' Pio! with Spline of Commert:ill Coml)Kt EV Clm-ging Dtr111od Foree.st 8ned on M onte C-arloSlmulation(IOOOOEVs;S0"4~+SO%SVV, 100001imes) -·· --~~- ------- !5 Tim&d03y " . .. Figure 6.29 Monte Carlo Simulation Result for Industrial EV Demand Model - Surf with Spline Plot in MATLAB (10,000 EVs) 203 USC Viterbi School of Engineering PhD Thesis Averaged EV demand forecast in 24 hour format of 10,000 Commercial EVs (50% Compact + 50% SUV) running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence inteival result can also be found below, Table 6.12. Table 6.12 10,000 Commercial EVs Charging Demand Estimation and Forecast Result Upper (kWh) .. 16.9 16.8 17.1 - ~-~---~---~ 2.2 2.1 2.2 0.5 0.5 - 0.5 ----------- - 1.5 1.5 1.6 ---------------- 5.8 6.0 .. 5.9 ---~~-=-11-==--~-=~.--=-~-----=- - 22.4 22.2 22.6 77.6 78.2 - ~~~~:~~~-=~~=-~~~ 77.9 239.9 241.0 - 240.4 --------------- - 623.1 622.3 624.0 1311.6 1314.1 - ====~~==========~ 1312.8 MliM 2245.9 ••• 3205.8 ••• 3954.0 ••• 4408.9 ••• 4789.2 •• 5445.7 •• 4730.8 MIM 3107.9 MIM 2004.7 Mp+ 1285.o MJM 402.1 M}W 188.8 MJM 75.7 Yearly Demand 2244.5 3203.9 3952.0 4406.6 4786.8 5443.1 4728.4 3105.8 2002.9 1283.6 759.8 401.2 188.3 75.3 204 2247.7 3207.7 3956.2 4411.1 4791.7 5448.1 4733.5 3110.1 2006.3 1286.6 762.1 402.9 189.4 76.0 10.14 GWh USC Viterbi School of Engineering PhD Thesis 6.4 Residential EV Charging Demand Estimation and Forecast In order to estimate and forecast the EV charging demand for residential area, similar charging records on Charging Start Time, Plug-in Duration and Start SOC are necessary. However, only three months' EV rebate meter data recorded every 15 minutes on 225+ meters within LADWP seivice residential area were provided by LADWP, and these information is not useful to define and generate such random variables. In this case, additional assumptions are made to estimate and forecast residential EV charging demand. 6.4.1 Assumptions • The average battery size of residential EV is assumed to be 24 kWh. • There will only be one charging event or no charging event in a given day, and the probability for one charging event and zero charging event is 60% and 40%, respectively. • The charging start time random variable follows a Gaussian distribution with mean 21 and variance 2 [ 6 9]. • The charging start SOC random variable follows a Gaussian distribution with mean 50% and variance 30% [69]. 6.4.2 Methodology In order to forecast the Residential EV charging demand, same methodology are used as the Industrial EV Monte Carlo simulation, together with probability theory, statistics sampling and stochastic elements have been used and applied. The training and testing dataset are chosen 70% and 30% from the available data obtained from FleetCarma. Four random variables are designed and generated together with the additional assumptions, they are: Charging Event (CE), Charging Start Time (CST), Duration (D) and battery Start State of Charge (SSOC). For each random variable, multiple probability density functions (pdf) are approximated based on training data set and validations are made on the testing set to find out the pdf with the best goodness of fit indices. After that, all random variables will be generated using inverse transform technique. Monte Carlo simulation will be applied afterwards to find out the average EV charging demand of multiple simulations and Bootstrap technique will be used to discover the 95% confidence inteival of the demand for each hour. The detailed methodology can be found in Section 6.2.1. 205 USC Viterbi School of Engineering PhD Thesis 6.4.3 Monte Carlo Simulation Result Complete Residential EV Charging Demand Estimation and Forecast model will be present in this section. The complete model could be used to forecast the Residential EV charging demand of Los Angeles, California and USA for the future. This thesis is conducted under DOE and LADWP SGRDP and will only be used for a demonstration purpose. Therefore, the author selects 10,000 Residential to be the input of the model to estimate the EV charging demand. The Monte Carlo Simulation iteration number is also set to be 10,000 to ensure convergence. The results can be found in the following figures, Figure 6.30 to Figure 6.33. -~-~_J_J _____ J_J_~---L-~ 10 11 12 13 14 1 5 1 6 11 1 8 19 20 21 22 23 TtrroofD ay 600 Hist am of Monte C1rto Generaled C h in D uran on Vanables 600 ~ c ~ 400 ¥ u. 200 0 .5 1.5 2.5 3.5 ~.5 cna~~o,,.100 150 H istogram of M onte Carlo G enerated Charging Start S OC Varot>les I ,,100 I ' l 0 c • j a I f u. 50 I ' 10 20 30 41) 50 60 TO 81) 90 100 Ba11 e1y sec llll Figure 6.30 Generated CST, Charging Duration and SSOC Random Variables for 10,000 Residential EV s 206 10000 ~ 8000 . . "O c ro E ~ Cl 6000 "' c ·~ ro .c () G; 4000 2000 USC Viterbi School of Engineering PhD Thesis R esidential EV Charging (Level 2) D emand Forecast B ased on Monte Carlo Simulation (10000 E Vs, 10000 times) 10 11 12 13 14 15 1 6 17 1 8 19 20 21 22 23 Time of Day Figure 6.31 Monte Carlo Simulation Result for Residential EV Demand Model (10,000 EVs) Sutt Ploc of Rnldenlial EV C~ o.m.nd forKnl Butel on Monte c.to Slmulltion (10000 EVa, 10000 limnJ 10000 8000 6000 4000 2000 Figure 6.32 Monte Carlo Simulation Result for Residential EV Demand Model - Surf Plot in MATLAB (10,000 EVs) 207 USC Viterbi School of Engineering PhD Thesis Surf Plot wilt! Splint of RHidtr'dal E V Ch11ging Oemand Fomast Bntd on M ontt Carlo smmi1 11ion (1 0000 EVt. 10000 timt•I "'" 1 2000 .... 1000) .•.. 8000-~- ______ ... -<.-~ ., Figure 6.33 Monte Carlo Simulation Result for Residential EV Demand Model - Surf with Spline Plot in MATLAB (10,000 EVs) Averaged EV demand forecast in 24 hour format of 10,000 Residential EVs running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table 6.13. 208 USC Viterbi School of Engineering PhD Thesis Table 6.13 10,000 Residential EVs Charging Demand Estimation and Forecast Result 11111 - - - 111111 - - - - 11111 Mi!M ••• Mi# MIM MIM •LW .,. .,. Ml:M •@• Ml1M MJM MJM MJM 8967.3 7055.7 4889.7 3015.3 1666.6 831.6 378.2 145.8 42.9 10.4 2.1 0.4 0.3 2.0 12.3 61.3 242.3 767.0 1951.5 4019.0 6764.7 9416.7 11000.7 10761.6 Ycarh Demand Lower (kWh) 8964.8 8969.8 7053.2 7058.2 4887.7 4892.0 3013.5 3017.l 1665.2 1668.1 830.7 832.6 377.5 378.8 145.4 146.2 42.7 43 .1 10.3 10.5 2.0 2.1 0.3 0.4 0.3 0.3 1.9 2.0 12.2 12.4 61.0 61.5 241.8 242.8 766.2 767.9 1950.2 1952.8 4017.2 4021.0 6762.2 6767.3 9413.8 9419.2 10998.0 11003.7 10758.6 10764.5 18.77 GWh 209 USC Viterbi School of Engineering PhD Thesis 6.5 Conclusion This chapter proposed an EV charging demand estimation and forecast model with Monte Carlo simulation technique for the three customer classes: • Industrial • Commercial • Residential The historical charging records on 36 LAD WP assigned EVs, and 30 LAD WP carpool EVs have been chosen to construct the forecast model and represent for the Industrial and Commercial area. LAD WP residential EV rebate meter data have been used, together with additional assumptions, to estimate EV demand for the Residential area. Monte Carlo simulation, together with probability theory, statistics stochastic elements and Bootstrap confidence interval have been used and applied. The training and testing dataset are chosen 70% and 30% from the available data obtained from FleetCarma. Four random variables are designed and generated, they are: Charging Event (CE), Charging Start Time (CST), Duration (D) and battery Start State of Charge (SSOC). For each random variable, multiple probability density functions (pelf) are approximated based on training data set and validations are made on the testing set to find out the pdf with the best goodness of fit indices. After that, all random variables will be generated using inverse transform technique. A very large number of iterations of Monte Carlo simulation will then be applied afterwards to find out the average EV charging demand of multiple simulations and Bootstrap technique will be used to discover the 95% confidence interval of the demand for each hour. The developed model could be used to evaluate the impact of EV charging on the distribution grid. The proposed model may benefit LADWP or other similar utility operators to plan the generation and operation profiles accordingly in the future. Further research may be needed to improve the model. This can be done by considering and incorporating more variables into the current designed model and by acquiring better quality and more real-world data on EV charging. 210 USC Viterbi School of Engineering PhD Thesis 7. CHAPTER 7: Conclusions and Future Work This thesis has examined the impact of electric vehicle's integration to the distribution grid. It first evaluates the impact of EV infrastructure charging on the distribution grid using diverse system conditions. Then, various EV charging/discharging strategies have been designed and tested, and an integrated solution has been developed that combines EV, renewable and storage systems to mitigate the impact. Furthermore, statistical analyses have been presented on actual EV usage, charging patterns, station usage and also focus on the user's behavior. Finally, an EV charging demand model based on historical charging record and Monte Carlo simulation technique has been developed that can be used to estimate and forecast future EV charging demand to help utility operators to plan the generation and operation profiles accordingly and benefit the utilities. This chapter summaries the main contributions of the thesis, presents the conclusions and proposes future work. 7.1 Thesis Contribution The contributions made in this thesis are summarized as follows: 1. It presents a detailed evaluation of EV infrastructure charging on the distribution grid (Chapter 2). Daily load profile has been modified to allow load flow analysis under different load periods. Three real-world distribution systems have been selected and tested via diverse EV penetration levels and system loading conditions. The designed cases have been tested and verified by two specialized power system analysis software: EDD and OpenDSS. 2. It develops an integrated EV charging/discharging solution to obtain an optimum daily load profile (Chapter 3). Various EV charging/discharging models, based on the concept of vehicle-to-grid (V2G), are designed and tested on two distribution networks. An integrated algorithm that incoiporate EV, PV and storage system is developed and verified to achieve the goal of shaving the peak and filling the valley. 3. It analyzes the EV charging patterns and power usage of different types of charging stations from one year actual data provided by LADWP, UCLA and FleetCarma (Chapter 3). The unique data sources and analysis are important for LADWP and other similar utility companies. The results could be used to provide information and practical insights for LAD WP and other similar utility companies for resource planning and future installations of wired and wireless charging infrastructures for EV. 211 USC Viterbi School of Engineering PhD Thesis 4. It gives a detailed analysis of EV user behaviors based on one year real-world trip information and driving patterns from various EV users (Chapter 5). Traditional similar analyses mostly used survey data as the only source. The users are categorized into three groups: large corporate EV users, EV users employed at a university, and typical EV owners in the city. From the data, charging habit and key trip information, such as trip start time, duration, mileage, and energy consumed have been studied. These information reveals different pattern of EV user behavior, and the results suggest the behavior varies in different groups. 5. It constructs an EV charging demand estimation and forecast model based on historical charging record and using Monte Carlo simulation technique (Chapter 6). Conventional designed models do not have EV user behavior information and their driving behaviors, and those models were mainly rely on traffic patterns, EV battery characteristics and charging characteristics, without using historical real-world charging record. The designed model can be used to estimate future EV demand for the three customer classes: Industrial, Commercial and Residential. The model in this thesis defines four random variables based on historical 20-months' real-world data. The presented model could be used to evaluate the EV charging on the distribution grid, and may benefit LAD WP or other similar utility operators to plan the generation and operation profiles accordingly in the future. 7.2 Impact of EV Infrastructure Charging on the Grid 7.2.1 Summary of Research Work Chapter 2 analyzes the impact of EV infrastructure charging on the grid with regards to system load flow, load factor and stability. Three distribution networks have been selected: • IEEE 34 node distribution test system (allows standardized model development and testing) • USC micro-grid (represents a large urban complex) • Chatsworth distribution system (a mixture of residential and industrial area) These systems have been tested under diverse system conditions: various case scenarios and load periods. The designed cases have been tested and verified by two specialized power system analysis software: EDD and OpenDSS. Moreover, various positive and negative effects on the distribution network and micro-grid 212 USC Viterbi School of Engineering PhD Thesis have be analyzed, such as: power quality (over-voltage, voltage sages, switching surges, harmonics and noise), impact on distribution infrastructure (transformers, wires, cables, and capacitors), capacity factor, utilization factor, load factor, system reliability. The conclusions for this part of the research are summarized as follows: • The existing distribution infrastructure has sufficient capacity to carry a reasonable level of EV penetration • Distribution transformers are one of the key equipment that may need to be replaced to achieve higher penetration levels • The EV loading profile is consequential to accurately model and predict the impact of EV penetration • Predicting the correct loading profile 1s important m understanding the infrastructure system limits 7 .2.2 Limitations and Suggestions for Future Work Three distribution system have been analyzed and they are all located in the west coast of United States. More distribution networks at different locations can be selected and tested for comparison and verification. The daily load curve has been simplified into three load periods, and all the case scenarios are tested under these three periods. More periods or real-time load curve may be used for the designed cases. No real-world data have been received on the selected distribution networks to validate the results, only simulation level of analysis is available in this chapter. Further research may include actual data and test the designed scenarios. 7.3 Distribution Effect of Battery Aggregation and Backfill Coupled with Renewable Energy 7.3.1 Summary of Research Work Chapter 3 describes distribution effects of battery aggregation and backfill coupled with renewable energy. The concept of vehicle-to-grid (V2G) has been introduced and then applied into EV charging/discharging strategies to evaluate the distribution effects of battery aggregation and backfill, and the possibility of "shaving the peak and filling the valley". In addition, the designed V2G strategies have been coupled with renewable energy 213 USC Viterbi School of Engineering PhD Thesis and storage systems and becomes an EV, battery and PV integrated solution. Several charging/discharging scenarios have been designed to represent G2V, G2B, B2V, B2G and V2G. They are listed as follows: • Different Scenarios for Battery Aggregation and Backfill • Scenario 1: Base Case (G2V and V2G) • Scenario 2: Simple Case (G2V and V2G) • Scenario 3: High Capacity (G2V and V2G) • Scenario 4: Full Peak Shaving (G2V, G2B, B2G and B2V) • Different Scenarios for Battery Aggregation and Backfill Coupled with PV Panels • G2V charging at period 2 (55% loading condition) with 30% Capacity PV • G2V charging at period 2 (55% loading condition) with 55% Capacity PV • G2V charging at period 2 (55% loading condition) with 100% Capacity PV • V2G charging at period 1 (100% loading condition) with 30% Capacity PV • V2G charging at period 1 (100% loading condition) with 55% Capacity PV • V2G charging at period 1(100% loading condition) with 100% Capacity PV • Battery Aggregation and Backfill Strategies in MATLAB Modeling • EV Coupled with Storage in PSCAD, power quality and harmonics In conclusion, an integrated algorithm that incorporate EV, PV and storage system has been developed and verified on distribution systems at simulation level that is able to achieve the goal of shaving the peak and filling the valley. 7 .3.2 Limitations and Suggestions for Future Work Due to lack of real-world data on EV charging and discharging, only simulation results are available for this chapter. Further research may include actual data and validate the designed strategies. This thesis presents an integrated solution of EV, PV and storage system, further research may be focused on the possibility of incorporate other renewable energies or technologies. 214 USC Viterbi School of Engineering PhD Thesis 7.4 EV Charging Patterns and Power Usage of Charging Stations 7.4.1 Summary of Research Work Chapter 4 presents EV charging patterns and power usage of three types of stations: • Charging stations used for large corporate fleet • Charging stations for local residential use • Charging stations used by LADWP employees for ride share purpose EV charging data are collected either through meters or through a third party (FleetCarma). Analyses have been made based on historical one year EV charging and usage records. Hourly, weekday vs weekend, seasonally and yearly energy distribution curves have been generated and compared to examine if they have similar patterns. The last part of this chapter estimates the effect of EV loading on California network under different EV penetration rates. The conclusions for Chapter 4 are summarized below: • The results could be used to provide information and practical insights for LAD WP and other similar utility companies for resource planning and future installations of wired and wireless charging infrastructures for EV • The EV charging analysis results suggested the charging energy demand pattern could be much different. • Charging stations deliver most of the energy to all EVs during daytime, the peak period may be offset by installing PV panels. • Estimating and forecasting the EV charging demand is needed for LADWP and other similar utility companies. 7.4.2 Limitations and Suggestions for Future Work The study in this part of the research covers one year actual data provided by LAD WP, UCLA and FleetCarma, future work is to continue collecting EV charging record and station usage data and keep updating previous results. Three types of charging stations have been categorized, more types, such as stations for rental EVs or EV buses may be included. 215 USC Viterbi School of Engineering PhD Thesis 7.5 Analysis on EV User Behaviors 7.5.1 Summary of Research Work Chapter 5 provides an analysis on EV user behaviors. Customer profiles have been analyzed and divided into three categories: • Large corporate EV users • EV users employed at a university • Typical EV owners in the city For each user group, top 10 heavy users have been selected and their daily trips including mileage, trip start time and battery energy distribution are analyzed. The conclusions for this part of the research are summarized as follows: • Individual EV user driving pattern and behavior is similar to other members in the same group • The analysis results indicate that EV user behavior varies in different groups • Estimating and forecasting the EV charging demand is needed for LADWP and other similar utility companies 7 .5.2 Limitations and Suggestions for Future Work The study in Chapter 5 covers one year actual data provided by LADWP, UCLA and FleetCarma, future work can be continually collecting EV charging record and station usage data and keep updating previous results. Three types of EV users have been categorized, more types, such as users of rental EVs or taxi drivers, may be included. 7.6 EV Charging Demand Estimation and Forecast 7.6.1 Summary of Research Work Chapter 6 presents an EV charging demand estimation and forecast model. The historical charging records on 36 LADWP assigned EVs, and 30 LADWP carpool EVs have been chosen to construct the forecast model and represent for the Industrial and Commercial area. LADWP residential EV rebate meter data is used, together with additional 216 USC Viterbi School of Engineering PhD Thesis assumptions, to estimate EV demand for the Residential area. The selected raw data ranges from May 2015 to February 2017, the first 70% of the data has been used to be the training set and the remaining 30% has set to be testing set. Four random variables have been defined based on historical data, they are: • Charging Event (CE) • Charging Start Time (CST) • Plug-in Duration (D) • Charging Start State of Charge (SSOC) Monte Carlo simulations have been applied afterwards to find out the average EV charging demand of a very large number of iterations and Bootstrap technique has been used to discover the 95% confidence inteival of the demand for each hour. The final results have been plotted to be the estimation and forecast EV charging demand. The conclusions for Chapter 6 can be found below: • The developed model could be used to evaluate the impact of EV charging on the distribution grid. • The proposed model may benefit LAD WP or other similar utility operators to plan the generation and operation profiles accordingly in the future. • The constructed model may potentially decrease costs by reducing mismatches between day-ahead scheduling and real-time operation and benefit the market. 7 .6.2 Limitations and Suggestions for Future Work The available historical data is less than two years, more data need to be collected and used to update the model. Additional sub-models may be created on different seasons or weekdays once having enough data, and then integrated into the model to obtain better estimation accuracy. The designed model in Chapter 6 contains four random variables, more variables may be incorporated in the future for model improvement. 217 USC Viterbi School of Engineering PhD Thesis REFERENCE [1] M. Kevin, K. Donald, and F. James, "U.S. Department of Energy Vehicle Technologies Program - Advanced Vehicle Testing Activity: Plug-in Hybrid Electric Vehicle Charging Infrastructure Review", Nov. 2008. [Online]. Available: https ://avt.inl. gov /sites/ default/files/pdf/phev /phev lnfrastructureReport08 .pdf [2] Department of Energy, "One Million Electric Vehicles by 2015. [Online]. Available: https://energy.gov/sites/prod/files/edg/news/documents/l Million Electric Vehicle Repo rt Final.pdf [3] S. Rezaee, E. Farjah, B. Khorramdel, "Probabilistic analysis of plug-in electric vehicles impact on electrical grid through homes and parking lots", IEEE Transactions on Sustainable Energy, vol.4, no. 4, 2013, pp. 1024-1033. [ 4] Department for Business Enterprise and Regulatory Reform (BERR), Department for Transport (DIT), "Investigation into the scope for the transport sector to switch to electric vehicles and plug-in hybrid vehicles", 2008 [5] P. Denholm and W. Short, "An evaluation of utility system impacts and benefits of optimally dispatched plug-in hybrid electric vehicles, Tech. Rep. NREL/TP- 620-40293, July 2006. [Online]. Available: http://www.mel.gov/docs/fy07osti/40293.pdf [6] L. P. Fernandez and T. G. S. Roman et al., "Assessment of the impact of plug-in electric vehicles on distribution networks," IEEE Transactions on Power Systems, vol. 26, no. 1, 2010, pp: 206-213. [7] K. Schneider, C. Gerkensmeyer, M. Kintner-Meyer and R. Fletcher, "Impact assessment of plug-in hybrid vehicles on pacific northwest distribution systems," 2008 IEEE Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century, Pittsburgh, PA, 2008, pp. 1-6. [8] S. Acha, T. C. Green and N. Shah, "Effects of optimised plug-in hybrid vehicle charging strategies on electric distribution network losses," IEEE PES T &D 2010, New Orleans, LA, USA, 2010, pp. 1-6. [9] S. Rezaee, E. Farjah, B. Khorramdel, "Probabilistic analysis of plug-in electric vehicles impact on electrical grid through homes and parking lots", IEEE Transactions on Sustainable Energy, vol.4, no. 4, 2013, pp. 1024-1033. [10] N. Leemput, F.Geth, J. Van Roy, A. Delnooz, J. Buscher, J. Driesen, "Impact of electric vehicle on-board single-phase charging strategies on a flemish residential grid" IEEE 218 USC Viterbi School of Engineering PhD Thesis Transactions on Smart Grid, vol. 5,no. 4, 2014, pp. 1815-1822. [11] F.H.P. Zamora, H. Kagan, M.A.Pelegrini, L.G. Gardiman, L. Zamboni, M. Gavazzi, M.A.P.Fredes, C.A.M. Gon9alves, J.P. Niggli Silva, "Evaluation of the impact of electric vehicles on distribution systems combining deterministic and probabilistic approaches", 22nd International Conf.and Exhib. on Electricity Distribution, 2013, pp. 1-4. [12] A. Briones, J. Francfort, P. Heitmann, M. Schey, S. Schey, J. Smart, "Vehicle-to-Grid (V2G) Power Flow Regulations and Building Codes Review by the AVTA", Idaho National Laboratory, 2012 [13] Letendre, S. and P. Denholm, 2006, "New load, or new resource?", Public Utilities Fortnightly,2006: 28-37. [14] Yang Hong ming, Xiong Luo cheng, Liu Bao ping. "Analysis of Charging and Discharging for Plug-in Hybrid Electric Vehicles". Journal of Electric Power Science and Technology, 2010, 25(3): 8-12. [15] K. Mets T. Verschueren F. De Turck C. Develder "Exploiting V2G to optimize residential energy consumption with electrical vehicle (dis )charging" Smart Grid Modeling and Simulation (SGMS) 2011 IEEE First International Workshop on 17-17 Oct. 2011 pp.7- 12 [ 16] P.cStroehle S. Becher S. Lamparter A. Schuller C. Weinhardt "The Impact of Charging Strategies for Electric Vehicles on Power Distribution Networks" Energy Market (EEM) 2011 8th International Conference on the European 25-27 May 2011 pp.51-56 [17] S. W. Hadley, "Evaluating the impact of plug-in hybrid electric vehicles on regional electricity supplies", IEEE iREP Symposium on Bulk Power System Dynamics and Control, Charleston, SC, USA, pp. 1-12, 2007. [18] X. Yu, "Impacts assessment of PHEV charge profiles on generation expansion using energy modeling system", IEEE Power and Energy Society 2008 General Meeting, Pittsburgh, Pennsylvania, USA, pp. 1-5, 2008. [19] G.T. Heydt, "The impact of electric vehicle deployment on load management strategies", IEEE Transactions on Power Apparatus and Systems, Vol. 102, pp. 1253-1259, 1983. [20] M. Kintner-Meyer, K. Schneider, R. Pratt, "Impacts assessment of plug-in hybrid electric vehicles on electric utilities and regional U.S. power grids. Technical analysis", 10th Annual EUEC Conference, Tucson, AZ, USA, Pacific Northwest National Laboratory (PNNL), pp. 1-20, 2007. 219 USC Viterbi School of Engineering PhD Thesis [21] S. Rahman, Y Teklu, "Role of the electric vehicle as a distributed resource", IEEE Power Engineering Society Winter Meeting, Vol. 1, pp. 528-533, 2000. [22] A. Heider, H.J. Haubrich, "Impact of wide-scale EV charging on the power supply network", IEE Colloquium on Electric Vehicle, A Technology Roadmap for the Future, pp. 6/1-6/4, 1998. [23] P. Denholm, W. Short, An evaluation of utility system impacts and benefits of optimally dispatched plug-in hybrid electric vehicles, Technical Report, National Renewable Energy Laboratory (NREL), 2006. [24] V Marano, G. Rizzoni, "Energy and economic evaluation of PHEVs and their interaction with renewable energy sources and the power grid", IEEE International Conference on Vehicular Electronics and Safety, pp. 84-89, 2008. [25] Y Gurkaynak, Z. Li, A. Khaligh, "A novel grid-tied, solar powered residential home with plug-in hybrid electric vehicle (PHEV) loads", IEEE Vehicle Power and Propulsion Conference, pp. 813-816, 2009. [26] Y Gurkaynak, A. Khaligh, "Control and power management of a grid connected residential photovoltaic system with plug-in hybrid electric vehicle (PHEV) load", IEEE Applied Power Electronics Conference and Exposition, pp. 2086-2091, 2009. [27] Tu Yiyun; Li Can; Cheng Lin; Le Lin, "Research on Vehicle-to-Grid Technology," in Computer Distributed Control and Intelligent Environmental Monitoring (CDCIEM), 2011 International Conference on, vol., no., pp.1013-1016, 19-20 Feb. 2011 [28] Yifeng He; Venkatesh, B.; Ling Guan, "Optimal Scheduling for Charging and Discharging of Electric Vehicles," in Smart Grid, IEEE Transactions on , vol.3, no.3, pp.1095-1105, Sept. 2012. [29] M. Allan, "Load model for prediction of electric vehicle charging demand", International Conference on Power System Technology (POWERCON), Zhejiang, China, 2010. [30] Z. Luo, Y. Song, Z. Hu, Z. Xu, X. Yang, and K. Zhan, "Forecasting charging load of plug-in electric vehicles in China", IEEE Power and Energy Society General Meeting, Michigan, USA, 2011. [31] K. Qian, C. Zhou, M. Allan, and Y. Yuan, "Modeling of Load Demand Due to EV Battery Charging in Distribution Systems", IEEE Transactions on Power Systems, vol. 26, pp. 802-810, May 2011. 220 USC Viterbi School of Engineering PhD Thesis [32] J. Kiviluoma and P. Meibom, "Methodology for modelling plug-in electric vehicles in the power system and cost estimates for a system with either smart or dumb electric vehicles", Energy, vol. 36, no. 3, pp. 1758-1767, Mar. 2011. [33] T. K. Kristoffersen, K. Capion, and P. Meibom, "Optimal charging of electric drive vehicles in a market environment", Applied Energy, vol. 88, pp. 1940-1948, May 2011. [34] N. Juul and P. Meibom, "Optimal configuration of an integrated power and transport system", Energy, vol. 36, no. 5, pp. 3523-3530, May 2011. [35] Liang H, Sharma I, Zhuang W, Bhattacharya K. Plug-in electric vehicle charging demand estimation based on queueing network analysis. IEEE power and energy soc general meeting, National Harbor, :MD. p. 1-5. [36] Bae S, Kwasinski A. Spatial and temporal model of electric vehicle charging demand. IEEE Trans Smart Grid 2012; 3(1):394-403. [37] Wang H, Song Q, Zhang L, Wen F, Huang J. Load characteristics of electric vehicles in charging and discharging states and impacts on distribution systems. In: Sustainable Power Generation and Supply (SUPER GEN) Int Conference, Hangzhou, China. p. 1-7. [38] Mu Y, Wu J, Jenkins N, Jia H, Wang C. A spatial-temporal model for grid impact analysis of plug-in electric vehicles. Appl Energy 2014; 114:456-65. [39] Dong X, Mu Y, Jia H, Wu J, Yu X. Planning of fast EV charging stations on a round freeway. IEEE Trans Sustain Energy 2016:1-10. [40] Majidpour M, Qiu C, Chu P, Pota HR, Gadh R. Forecasting the EV charging load based on customer profile or station measurement? Appl Energy 2016; 163: 134-41. [ 41] Lojowska A, Kurowicka D, Papaefthymiou G, van der Sluis L. From the transportation patterns to power demand: Stochastic modeling of uncontrolled domestic charging of electric vehicles. IEEE power and energy soc general meeting, San Diego, CA. p. 1-7. [42] Xydas E, Marmaras C, Cipcigan L, Hassan A, Jenkins N. Forecasting electric vehicle charging demand using support vector machines. In: Power engineering conference (UPEC), Dublin, Ireland. p. 1-6. [43] Luo Z, Song Y, Hu Z, Xu Z, Yang X, Zhan K. Forecasting charging load of plug-in electric vehicles in China. IEEE power and energy soc general meeting, San Diego, CA. p. 1-8. [ 44] Fiori C, Ahn K, Rakha H. Power-based electric vehicle energy consumption model: 221 USC Viterbi School of Engineering PhD Thesis model development and validation. Appl Energy 2016; 168:257-68. [45] Kelly J, MacDonald J, Keoleian G. Time-dependent plug-in hybrid electric vehicle charging based on national driving patterns and demographics. Appl Energy 2012; 94:395- 405. [ 46] Wang H, Zhang X, Ouyang M. Energy consumption of electric vehicles based on real world driving patterns: a case study of Beijing. Appl Energy 2015; 157:710-9. [47] KaraE, Macdonald J, Black D, Berges M, Hug G, Kiliccote S. Estimating the benefits of electric vehicle smart charging at non-residential locations: a data- driven approach. Appl Energy 2015; 155:515-25. [48] Xydas E, Marmaras C, Cipcipgan L, Jenkins N, Carroll S, Barker M. A data-driven approach for characterizing the charging demand of electric vehicles: a UK case study. Appl Energy 2016; 162:763-71. [49] Li Y, Davis C, Lukszo Z, Weijnen M. Electric vehicle charging in China's power system: energy, economic and environmental trade-offs and policy implications. Appl Energy 2016; 173:535-54. [50] EV SGDP Preliminary Design Document, 2012 [51] IEEE 34-bus Feeder Standard. [Online]. Available: https://ewh.ieee.org/soc/pes/dsacom/testfeeders/ [52] IEEE PES Distribution System Analysis Subcommittee, Radial Distribution Test Feeders. [Online]. Available: https://ewh.ieee.org/soc/pes/dsacom/testfeeders/testfeeders.pdf [53] C. Botsford and A. Szczepanek, "Fast charging vs. slow charging: Pros and cons for the new age of electric vehicles,'' presented at the 24th Electric Vehicle Symposium, Stavanger, Norway, May 2009. [54] SAE Electric Vehicle and Plug-in Hybrid Electric Vehicle Conductive Charge Coupler, SAE Standard Jl 772, Jan. 2010. [55] Electric Vehicle Infrastructure Installation Guide, Pacific Gas and Electric Company, San Francisco, CA, USA, Mar. 1999. [56] Jim Francfort, "Electric Vehicle Charging Levels and Requirements Overview" Idaho National Laboratory, Dec. 2010 222 USC Viterbi School of Engineering PhD Thesis [57] Electric Vehicle Infrastructure Installation Guide, Pacific Gas and Electric Company, San Francisco, CA, USA, Mar. 1999. [58] S. Philipp, B. Silvio, L. Steffen, S. Alexander, and W. Christof, "The Impact of Charging Strategies for Electric Vehicles on Power Distribution Networks," International Conference on the EEM, May 2011 [59] Ralph Masiello, Khoi Vu, Lisa Deng, Alicia Abrams, Karin Corfee and Jessica Harrison, "Research evaluation of wind generation, solar generation, and storage impact on the California grid", KEMA. Inc, San Francisco, CA. [Online]. Available: http://www.energy.ca.gov/2010publications/CEC-500-2010-010/CEC-500-2010-010.PDF [60] PEV registrations - National Renewable Energy Laboratory analysis, R.L. Polk, POLK_ VIO_DETAIL_2014, May 2015. [61] M. Rastegar, A. Safdarian, M. Fotuhi-Firuzabad, and F. Aminifar, "Impacts of Plug-in Hybrid Electric Vehicle Uncertainty and Grid Unavailability on Home Load Management," in the International Conference on Environment and Electrical Engineering (EEEIC), 2012, pp. 693-698. [62] S. Rezaee, E. Farjah, B. Khorramdel, "Probabilistic analysis of plug-in electric vehicles impact on electrical grid through homes and parking lots", IEEE Transactions on Sustainable Energy, vol.4, no. 4, 2013, pp. 1024-1033. [63] A. Ashtari, E. Bibeau, S. Shahidinejad, T. Molinski, "PEV charging profile prediction and analysis based on vehicle usage data", IEEE Transactions on Smart Grid, vol.3, no. 1, 2012,pp. 341-350. [64] M. Zhang, J. Chen, "The energy management and optimized operation of electric vehicles based on microgrid", IEEE Transactions on Power Delivery, vol.29, no. 3, 2014, pp. 1427-1435. [65] Younes Z, Boudet L, Suard F, Gerard M, Rioux R. Analysis of the main factors influencing the energy consumption of electric vehicles. In: 2013 international electric machines & drives conference, 2013. p. 247-53. [66] Dutta S. Data mining and graph theory focused solutions to Smart Grid challenges. Ol-Dec-2012. [ 67] Selbas R, Sencan A, Kucuksille E. Data mining method for energy system applications. In: Knowledge-oriented applications in data mining, InTech, 2011, p. 147-66. [68] Zhang P, Qian K, Zhou C, Stewart B, Hepburn D. A methodology for optimization of 223 USC Viterbi School of Engineering PhD Thesis power systems due to electric vehicle charging load. IEEE Trans Power Syst 2012; 27(3): 1628-36. [69] Zhang P, Wu X, Wang X, Bi S. Short-term load forecasting based on big data technologies. CSEE J Power Energy Syst 2015; 1(3):59-67. [70] Kelly J, MacDonald J, Keoleian G. Time-dependent plug-in hybrid electric vehicle charging based on national driving patterns and demographics. Appl Energy 2012; 94: 395- 405. [71] M. Ghiassi, D. K. Zimbra, and H. Saidane, "Medium term system load forecasting with a dynamic artificial neural network model", Electric Power Systems Research, vol. 76, no. 5, pp. 302-316, Mar. 2006. [72] J. A. Jardini, C. M. V Tahan, M. R. Gouvea, S. U. Ahn, and F. M. Figueiredo, "Daily load profiles for residential, commercial and industrial low voltage consumers", IEEE Transactions on Power Delivery, vol. 15, no. 1, pp. 375-380, 2000. [73] A. Aabrandt et al., "Prediction and optimization methods for electric vehicle charging schedules in the EDISON project," IEEE Power Energy Soc. Innov. Smart Grid Tech. Conf., 2012, pp. 1-7. [74] A. Ashtari, E. Bibeau, S. Shahidinejad, and T. Molinski "PEV charging profile prediction and analysis based on vehicle usage data," IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 341-350, Mar. 2012. [75] A. Khosravi and S. Nahavandi, "Load forecasting using interval type-2 fuzzy logic systems: Optimal type reduction," IEEE Trans. Ind. Informat., vol. 10, no. 2, pp. 1055- 1063, May 2014. [76] G. Bontempi, S. Ben Taieb, and Y Le Borgne, "Machine learning strategies for time series forecasting," in Business Intelligence. Berlin, Germany: Springer-Verlag, 2013, pp. 62-77. [77] E. S. Xydas, C. E. Marmaras, L. M. Cipcigan, A. S. Hassan and N. Jenkins, "Forecasting Electric Vehicle charging demand using Support Vector Machines," Power Engineering Conference (UPEC), 2013 48th International Universities', Dublin, 2013, pp. 1-6. [78] S. M. Ross, Simulation: Elsevier Science, 2012. [79] R. Y Rubinstein and D. P. Kroese, Simulation and the Monte Carlo Method: Wiley, 2011. 224 USC Viterbi School of Engineering PhD Thesis [80] Grahn, P. (2013) Electric Vehicle Charging Impact on Load Profile. PhD thesis. School of Electrical Engineering, Royal Institute of Technology, Stockholm, Sweden. [81] Papadopoulos, P. (2012) Integration of Electric Vehicles into Distribution Networks. PhD thesis. Institute of Energy, School of Engineering, Cardiff University. [82] B. Efron, "Bootstrap methods: another look at the jackknife," The Annals of Statistics, Vol. 7, No. 1, pp. 1-26, 1979. [83] B. Efron, "Better bootstrap confidence intervals," Journal of the American Statistical Association, Vol. 82, No. 397, pp. 171-185, 1987. [84] Cenex, "Vehicle to Grid", available online: http://www.cenex.co.uk/vehicle-to-grid/ 225 USC Viterbi School of Engineering PhD Thesis APPENDIX A. Industrial Midsize EV Demand Forecast Model This section presents the summary and results of the Industrial Midsize EV model, with 6 LADWP assigned midsize EVs' result and analysis period 05/20/2015 - 02/04/2017. • Training Set: 05/20/2015 - 08/20/2016 • Testing Set: 08/21/2016 - 02/04/2017 A.1 Charging Event The summary of 6 midsize compactEVs charging events are summarized below, Table Al, Figure Al. Table A.1 Individual EV Charging Event of Industrial Midsize EV s, 05/20/15 - 02/04/17 Average Average Percentage 374 210 345 255 58.8% 57 44 67 116 26.8% 226 4 0 0 110 58 10 13 7 3 45 15 2 10.5% 3.5% 0.5% CE=O USC Viterbi School of Engineering Charging Event of Industrial Midsize EVs (n=6) 05/20/16 - 02/04/17 CE=l CE=2 CE=3 #of Charging Event per Day • EVl EV2 EV3 • EV4 EVS EV6 . Average PhD Thesis CE=4 Figure A.1 Charging Event Distribution oflndustrial Midsize EVs, 05/20/15 - 02/04/17 A.2 Charging Start Time Distribution of the historical charging start time data on training and testing set are shown below, Figure A.2. > ...., 0.12 0.1 0.08 -~ 0.06 ~ Cl 0.04 0.02 0 Industrial Model (Midsize}: Charging Start Time Distribution Training vs Testing 05/20/15 - 02/04/17 ' .J 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day • train • test Figure A.2 Charging Start Time Distribution of Industrial Midsize EV s, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on CST training set is a combination of 3 Gaussian distributions, Figure A.3. 227 ---Gou.J. ... ·•·oopl-f>. b•J,'1:11"2) •"2"'""'-C.~2) · ~-=.!!=...- • \ • Cl.Ol6l'«l.056Jl. ll07'tl. ~11: :::.=..':~ ol • O.O.Z.Z7•to.010'1.CUll~ ~ : ~:;~~ ~ ~;~~:::iSll ol • llOU1SIOO'IH . G.01fJ) ~= ~~;;: SSl . OJm.lll ~-•7724 ::"o.a::o.n 11 USC Viterbi School of Engineering PhD Thesis i..-- 01'-.ll(«o 1>111tW2l• - ••>·MJ11.,_l>lV<l)~l) 0c- ........ -,,...._UIXlll-CIT·- J . :.001 i4'0:1--~-~...,.._.~,,......_._.,...._~,...,._.-...... ~.,..,.., ....... "-'--mcoo-rr-,......_.~___, '- -"02 L . ,_ " Figure A.3 CST Curve Fitting on Training Set for Industrial Midsize EVs The fitted pdf function for the training set is shown below: ( x-6.166) 2 (x-15.93) 2 (x-13.16) 2 f(x) = 0.06624e- o.7674 + 0.02274e- o.s137 + 0.01515e- 5.307 Validation on the testing set and residual plot can be found in Figure A.4. Charging Start Time Model (pdf) vs Actual Observation o.1....--.---...---.--=.--T--.-.....-...-.---.----..---r--'-'r--+-...,...-..--.......... ---...---.---.---.--.-----. ~ -~ 0.05 Q) c 0.02 m ::::J + -- CST Model + Actual 0 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day Residual Plot Charging Start Time ~ 0 1---- -..-..... 1.'A.J ..... ..--......J ....... 1'-............ ~ -0.02 0 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day Figure A.4 CST Validation on Testing Set for Industrial Midsize EVs 228 USC Viterbi School of Engineering PhD Thesis A.3 Plug-in Duration Distribution of historical plug-in duration on training and testing set are shown below, Figure A.5. >- 0.45 0.4 0.35 0.3 ·~ 0.25 ai 0.2 0 0.15 0.1 0.05 0 Industrial Model (Midsize): Plug-in Duration Distribution Training vs Testing 05/20/15 - 02/04/17 .I 11 II .1 I-_ _ - .. -_ _ - 0 0.5 1 1.5 2 2.5 Plug-in Duration (hr) • Train • Test 3 3.5 4 4.5 Figure A.5 Plug-in Duration Distribution of Industrial Midsize EVs, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a combination of 2 Gaussian distributions, Figure A.6. -- - ·· ·~ . _ .. .... ,. =-'~ 1 :--cttH I !,,:: .. 0.15 .. ~ .1·~·"111o1:•Z1• .l'"""'4 ~ZI Dc- ...ii- .L ' ....... ...,_o ... l'":l--'----~------~-~-~----1 ... .L ' L " Figure A.6 Plug-in Duration Curve Fitting on Training Set for Industrial Midsize EVs ( x-2.17) 2 (x-0.8723) 2 f(x) = 1.509e- 0.1492 + 0.2227e- o.6463 229 USC Viterbi School of Engineering PhD Thesis Validation on the testing set and residual plot can be found in Figure A. 7. Plug-in Duration Model (pdf) vs Actual Observation 0.3 ~-~-~-~--~-~-~--~-~-~ + -~ 0.2 + - D Model + Actual en c Q) 0 0.1 + + 0 '----- _.L_ _ __.L_ _ __J_ __ .L..__--1=~--..L.-~F----*------+ 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) Residual Plot Plug-in Duration 0 .1~--~-~~-~-~~~~-~-~~-~ (ii 01-- :::J -c "iii Q) 0:: -0.1 -0.2 '-----'--.L..___J_ _ __.L_ _ _.L__c___._ _ _,___..__.. _ _, 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) Figure A.7 Plug-in Duration Validation on Testing Set for Industrial Midsize EVs A.4 Start SOC Distribution of historical charging start SOC on training and testing set are shown below, Figure A.8. 0.7 0.6 0.5 c 0.4 ·v; c ~ 0.3 0.2 0.1 0 Industrial Model (Midsize} Start SOC Distribution Training vs Testing 05/20/15 - 02/04/17 bL_ ... ....... - .. . . . . . . . . . ·-· -· . -·· ... • .. -· ---- .. -· ... o • • • • • • • - - • · - · '' - •••• J 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75 78 81 84 87 90 93 96 99 Start SOC Interval(%) • Train • Test Figure A.8 SSOC Distribution oflndustrial Midsize EVs, 05/20/15 - 02/04/17 230 USC Viterbi School of Engineering PhD Thesis Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a combination of 2 Exponential distributions, Figure A 9 . ~-'.Z ~-··.-·.,·e...wio ~(Wllfo~_........"°"'"""'" •• Ol•Mll-1.0l•ftll a.. 11 .. Ptlfi.Zl••I ~:=~'-::1o:°=.. -·· , .. ..,,..,, ·-- =--:.,:;:;--ottN ......... 1~1 [ : : : ~=~= 0 :10 40 so " '" --"""" . ...._uo;a-.cx:.-1 Figure A.9 SSOC Curve Fitting on Training Set for Industrial Midsize EVs f(x) = 0.6136e- 2 · 784 x + 0.002848e 0 · 003818 X Validation on the testing set and residual plot can be found in Figure A.10. ~ 0.2 0 '-="'"-'""44-µJ.l ................ ~ ......................................................... -'4-;o ......................................................................... ..._........,......, ....... ........... 0.1 0 2 4 6 8 10 i2 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 B2 64 66 B8 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 Charg;ng Stort SOC(%) ! 0.05 &! 0 2 4 6 8 101214161820222426283032343638404244464850525456586062646668707274767880828486889092949698 Charg;ng Stort SOC(%) Figure A.10 SSOC Validation on Testing Set for Industrial Midsize EVs A.5 Goodness of Fit Summary of the three goodness of fit indices can be found in Table A.2. 231 USC Viterbi School of Engineering PhD Thesis Table A.2 Goodness of Fit Indices Summary of Industrial Midsize EV Model Goodness Train Test of fit Index R-square 77.26% 71.71% CST RMSE 0.0078 0.0075 (Gaussian 3) MAPE 0.0863 0.0880 R-square 84.44% 69.34% Duration RMSE 0.0274 0.0587 (Gaussian 2) MAPE 0.0976 0.1296 R-sguare 99.82% 98.61% Start SOC RMSE 0.0026 0.0153 (Exponential 2) MAPE 0.1661 0.1032 A.6 Monte Carlo Simulation Result When number of Monte Carlo Simulation iterations is set to be 10,000 and number of Industrial Midsize EV number is set to be 10,000, the results of EV charging demand estimation and forecast are provided in the following figures, Figure A.11 to Figure A.14. aJtiistogram of Monte Carlo Generated Charging Start Time Variables 600 >- g .. :J 400 ~ ~ u.. 200 O '--_,,_,,.....~•~~' ~..._._.~~ 0 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day 3 ~istogram of Monte Carlo Generated Charging Duration Variables 2500 >- 2000 " c: !g 1500 ~ u.. 1000 500 0 0.5 1.5 Charging Duralion 2.5 3.5 2500 Histogram of Monte Carlo Generated Plug-in Duration Variables 2000 >- g 1500 .. :J CT £1000 500 o '--_.-..___.-.~--.~---~--~--'~~~~ -0.5 0 0.5 1.5 Plug-in Duration 2.5 3 3.5 600 ~istogram of Monte Carlo Generated Charging Start SOC Variables 5000 ,._ 4000 " c: !g 3000 ~ u.. 2000 1000 o~~~~~~~~~~~~~~~~ 0 20 40 60 80 100 Battery SOC (%) Figure A.11 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for Industrial Midsize EV 232 USC Viterbi School of Engineering PhD Thesis 4000 ~~~~ 1M _u ~ s_ tri~ •l~ M_ id~ siz~ e~ EV~C ~M _ r~ gi~ ng~ (L ~ •- ••~ 12r )_ De~ m~ an_ d~ Fo ~ re ~ c~ as~ tB _ aT se~ d~ on~M ~ o~ nt~ •~ C•~ rlo ~S ~ im ~u_ la_ tior n~ (1~ 00~ 00 ~EV -r s,_ 10~ 0~ 00~ ti~ me ~ s~ )~~~ 3500 3000 1000 500 o O....i.....-ll!!l!!!~~L__1__J__L_J~L__.l__L_L_JL_J__....L_L-1~L-J__J___:r~~__J 0 10 11 12 13 1 4 15 16 17 18 19 20 21 22 23 nmeofDay Figure A.12 Monte Carlo Simulation Result for Industrial Midsize EV Demand Model Surf "'°'ol lnduwQI Mid1i~ EV ch•~ O.lllilld ForKnl S.Md on Mont• C.nti Simui.1'°'1 (10000 EV1, HlOOO times) 4(.'0J •••• '"' " -----·· 20 - ..... : ,5 .. ---::.:· -----·· 10 r11111o1o.y Figure A.13 Monte Carlo Simulation Result for Industrial Midsize EV Demand Model - Surf Plot in MATLAB 233 USC Viterbi School of Engineering PhD Thesis Surl Plot witt1 Spline of Industrial Midsii• EV Char" DtmWI fortcut Based on Monte tar1o Simulalion (10000 EVs, 10000 limn) 50() ••• 10000 ......... ~ _ ___ ,. ... --...:~ ---------·· '5 ------- 10 Figure A.14 Monte Carlo Simulation Result for Industrial Midsize EV Demand Model - Surf with Spline Plot in MATLAB Averaged EV demand forecast in 24 hour format of 10,000 Industrial Midsize EVs running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table A.3. 234 USC Viterbi School of Engineering PhD Thesis Table A.3 10,000 Industrial Mid size EV Charging Demand Estimation and Forecast Result 20.5 20.6 - 20.6 .===;;;;;;;;;:~===;;=====;; - 44.4 44.3 44.5 ----------- 100.5 100.9 .. 100.7 .===;;;;;;;;;: ~iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiipiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii - 839.8 838.4 841.2 - -=====~~== .............. ~==~ 3223.7 3221.0 3226.4 2952.5 2957.2 - 2954.8 .............................. ~~ ................................................................................. ~ 11111111 __ ~ 1_ 21_ 2~-~~~~~-p~~~~-- - 951.6 Mi!M 1218.8 ••• 1458.5 ••• 1632.1 MIM 1112.3 MIM 1677.8 •LW 1850 •• 2295.2 .,. 1580 Ml:M 871.1 •@• 599.4 MJM 238.6 MJW 134.4 MJM 67.2 Ycarh Demand 1210.9 950.8 1217.8 1457.4 1631.0 1711.2 1676.7 1848.5 2293.2 1578.4 870.1 598.7 392.7 238.3 134.2 67.l 235 1213.0 952.4 1219.7 1459.6 1633.3 1713.5 1679.1 1851.3 2297.0 1581.6 872.0 600.1 393.7 238.9 134.6 67.3 6.55 GWh USC Viterbi School of Engineering PhD Thesis B. Industrial Electric SUV Demand Forecast Model This section presents the summary and results of the Industrial Electric SUV model, with 4 LADWP assigned Electric SUVs' result and analysis period 05/20/2015 - 02/04/2017. • Training Set: 05/20/2015 - 08/20/2016 • Testing Set: 08/21/2016 - 02/04/2017 B.1 Charging Event The summary of 4 industrial Electric SUVs charging events are summarized below, Table B.1, Figure B.1. Table B.l Individual EV Charging Event oflndustrial Electric SUVs, 05/20/15 - 02/04/17 C' QI ... - 0 26 1 0 109 109 109 109 18.3% 8.0% 1.6% Charging Event of Industrial Electric SUVs (n=4) 05/20/15 - 02/04/17 CE=O CE=l CE=2 CE=3 #of Charging Event per Day • EVl • EV2 EV3 • EV4 . Average Figure B.1 Charging Event Distribution oflndustrial Electric SUVs, 05/20/15 - 02/04/17 236 USC Viterbi School of Engineering PhD Thesis B.2 Charging Start Time Distribution of the historical charging start time data on training and testing set are shown below, Figure B.2. 0.12 0.1 0.08 ~ ·~0.06 <lJ 0 0.04 0.02 0 Industrial Model (SUV}: Charging Start Time Distribution Training vs Testing 05/20/15 - 02/04/17 I ' 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day • train • test Figure B.2 Charging Start Time Distribution of Industrial Electric SUV s, 05/20/15-02/04/17 Using MATLAB curve fitting toolbox, the pdf model on CST training set is a combination of 3 Gaussian distributions, Figure B.3. _ .. _G-i ... ol'-..:Ca bl>'tll"J:l •ol'...C-C.~l.I· ol"-C.bnui"J:I ~,..,, ............ ~ .,. OJJ71QZICl.06llt,O.Om1) ~:: :~~~'~ .Z • o.otlto.Mlt.~ bl • U.!.1 11111712.J1l <l • G..Kll)tc.ZJIM.CU1'11 oJ• O.ClllWIO.OICll,O.OZIJJ) l>l• 1u.i111111n.c1 d• O'Cl.A1l,SM7) r.- ., ....... 111'<1Y'J:l•-••J·-·b)\ll<JJ•ll [Jc--- Figure B.3 CST Curve Fitting on Training Set for Industrial Electric SUVs The fitted pdf function for the training set is shown below: 237 ...... USC Viterbi School of Engineering PhD Thesis ( x-7.831) 2 (x-12.52) 2 (x-12.24) 2 f(x) = 0.07102e- 0.6206 + 0.061e- o.3083 + 0.01769e- 4.55 Validation on the testing set and residual plot can be found in Figure B.4. Charging Start Time Model (pdf) vs Actual Observation 0.15........, .......... ----..--=.---T--.----.---.---.-.,..-...-,......-'-;---T'---T----.---.--.----.---.---.-.,..-...-----. ~ 0.1 'iii c: Q) Cl 0.05 0.05 (ii ::J + -- CST Model + Actual 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day Residual Plot Charging Start Time ~ 0 I------.,...-.._,..-- ~ -0.05 0 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day Figure B.4 CST Validation on Testing Set for Industrial Electric SUVs B.3 Plug-in Duration Distribution of historical plug-in duration on training and testing set are shown below, Figure B.5. 238 0.25 0.2 z- 0.15 ·v; c ~ 0.1 0.05 0 USC Viterbi School of Engineering Industrial Model {SUV}: Plug-in Duration Distribution Training vs Testing 05/20/15 - 02/04/17 - - II I I II II II I I 0 0.5 1 1.5 2 2.5 3 3.5 4 Plug-in Duration (hr) • Train • Test PhD Thesis •• 4.5 Figure B.5 Plug-in Duration Distribution oflndustrial Electric SUVs, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a Gaussian distributions, Figure B.6. -- ~·.-.i"""*' IM• •r•o•1wwa:i c.ac-~w. .................. ol• •11•S ll1MUlt11 .,. 11"·1·1~ ti• 1.tn IHQI i.-i sw:oooeo s "-01'71 =~~011" '- or...:c-•1w1)•1J Ck-•- __ o ·-~ r: ... J ... L 0 __ o Figure B.6 Plug-in Duration Curve Fitting on Training Set for Industrial Electric SUVs ( x-1.378) 2 f(x) = 0.1805e- 1.693 Validation on the testing set and residual plot can be found in Figure B.7. 239 USC Viterbi School of Engineering Plug-In Duration Model (pdf) vs Actual Observation 0.2 ~-~~~--~--1--~--'-r--~-~-~-~ 0.15 ~ -~ 0.1 CD Cl 0.05 + + + -- DModel + Actual + + O'-------'------'---L----'-----'---J....._--'-----'-----=i 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) 0.05 Residual Plot Plug-in Duration ai ~ "O 0 "iii CD er:: -0.05 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) PhD Thesis Figure B.7 Plug-in Duration Validation on Testing Set for Industrial Electric SUVs B.4 Start SOC Distribution of historical charging start SOC on training and testing set are shown below, Figure B.8. 0.07 0.06 0.05 -~ 0.04 V1 c ~ 0.03 0.02 0.01 0 Industrial Model {SUV) Start SOC Distribution Training vs Testing 05/20/15 - 02/04/17 1111 ~ 11 11 0 3 6 9 121518212427303336394245485154576063666972757881848790939699 Start SOC Interval(%) • Train • Test Figure B.8 SSOC Distribution oflndustrial Electric SUVs, 05/20/15 - 02/04/17 240 USC Viterbi School of Engineering PhD Thesis Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a combination of 2 Gaussian distributions, Figure B.9. ·-- IW•o1'..C•1'1Jllll"Zl •.Z'-..< •Wll'<l)•Z1 ~~_,,._.,,_,,,._..,. ol• OJl:Olnlllo;ooaJ,OJ)JUIJ bl• 1'M(l'U17'7f) <I• ut.< !22"lutl) 0:-~1 .. fl).O'llQI,~ :;: ::.:::;:;:= -Ok ,,.._, •-UM• =--~':"'0-m) r.,- •l'..C.-t.•M11•l) • o:!'""f'(•l>Zli0t2!•Z1 D<-~- .. lnWYll_SSOC ___ J ___ : .. : .. : ·· .L.. - .. ,., Figure B.9 SSOC Curve Fitting on Training Set for Industrial Electric SUVs ( x-73.14) 2 (x-69.85) 2 f(x) = 0.02677e- 3.264 + 0.02164e- 22.34 Validation on the testing set and residual plot can be found in Figure B.10. 0.08 Ch•rging Start SOC Model (pdf) va Aclu•I Obaervation + 0.08 t~=Model l + + ~ + + ~ 0.04 + " + + D.02 + + + + o,.l+\-+1+++1-+++Jl+H--l+l!.H-1++!.\+4J=~~:Ep:!t'::L,.+!.J..:':.L.Li+]_LJ.!J__L_.LL_J_L.L_L_l---1.:':J.!.i_L_Ll_l_J...;:i=++.J 02488W12Mfflffl~~~uu~"MUU~~«~U~~MMM~~M99ron~nn~~M9~~~M9" Charging Start SOC(%) 0.01 ~ i -0.01 0: -0.02 -0.03 0 2 4 6 8 101214181820222426283032343638404244464850525458588062846868707274787880&28486889092949898 Charging Start SOC(%) Figure B.10 SSOC Validation on Testing Set for Industrial Electric SUVs B.5 Goodness of Fit Summary of the three goodness of fit indices can be found in Table B.2. 241 USC Viterbi School of Engineering PhD Thesis Table B.2 Goodness of Fit Indices Summary oflndustrial Electric SUV Model Goodness Train Test of fit Index R-square 89.80% 67.84% CST RMSE 0.0060 0.0110 (Gaussian 3) MAPE 0.0955 0.1100 R-square 83.40% 88.96% Duration RMSE 0.0314 0.0204 (Gaussian 1) MAPE 0.0785 0.0741 R-~uare 83.41% 68.11% Start SOC RMSE 0.0051 0.0075 (Gaussian 2) MAPE 0.0870 0.0907 B.6 Monte Carlo Simulation Result When number of Monte Carlo Simulation iterations is set to be 10,000 and number of Industrial Electric SUV number is set to be 10,000, the results of EV charging demand estimation and forecast are provided in the following figures, Figure B .11 to Figure B .14. sJliistogram of Monte Carlo Generated Charging Start Time Variables 800 Histogram of Monte Carlo Generated Plug-in Duration Variables 500 >.400 g ~ 300 ~ u.. 200 100 0 0 1 2 3 4 5 6 7 8 9 10, 1 1213 14 151617 1819 20 21 22 23 Time of Day 1 ~1stogram of Monte Carlo Generated Charging Duration Variables 800 :>. g 600 "' :> ~ 400 u.. 200 0 ·1 .~ 0 \'" 1 n . a 2 4 5 6 Charging Duration 600 :>. g "' :> 400 ~ u.. 200 o~~~~~~~~~~~~~~~~ -1 6 Plug-in Duration 2 s'(!lstogram of Monte Carlo Generated Charging Start SOC Variables 200 :>. g 150 "' :> <T ~ ~ 100 ·~"""•oo~~I u.. 50 0 0 20 40 60 80 100 Battery SOC(%) Figure B.11 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for Industrial Electric SUV 242 USC Viterbi School of Engineering PhD Thesis SOOO,---,--r~ ln, du_ st_ na, l_ SU_ V, E_ V, Ch_ ar~ g, ing~ (~ L•_ ve_ l~ 2) 0 De _ m, a_ nd_ F 0 or_ e~ _•, t_ Ba_ se, d_ o_ nM ,o_ n_ te_ Ca_ rl, o_ Si_ m_ ula_ ti_ onr (_ 10_ 00, 0_ EV, s,~ 1_ 00 0 00_ t_ im, •s~ l--.-~~~ 5000 ~ 4000 £ "O ~ ~ Cl 3000 g> . e> ~ .c u ~ 2000 1000 10 11 12 13 14 15 16 17 1 8 19 20 21 22 23 TimeofDay Figure B.12 Monte Carlo Simulation Result for Industrial Electric SUV Model Surf Plot of lndu1tri1I SUV EV Cl\lflling Demlnd f OtKHI e.sed Ofl M ont• CMlo Simulltiot1 (1 0000 EVl, 10000 limes) " Figure B.13 Monte Carlo Simulation Result for Industrial Electric SUV Demand Model - Surf Plot in MATLAB 243 USC Viterbi School of Engineering PhD Thesis Sul'f Plot wilh Spline of lndus.tn.I SUV EV Ch•ging DftMnd Forte.HI BHMI on Mont. Carlo Simuhtion (10000 EV1, 10000 limH) 5000 ·~ • .... ~ ... 1 -~ ! ·~ 2CXXI ~ ~ tCO'.l . , .. - ------ " 20 ... ---~·- " Trneolo.y Figure B.14 Monte Carlo Simulation Result for Industrial Electric SUV Demand Model - Surf with Spline Plot in MATLAB Averaged EV demand forecast in 24 hour format of 10,000 Industrial Electric SUVs running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table B. 3. 244 USC Viterbi School of Engineering PhD Thesis Table B.3 10,000 Industrial Electric SUV Charging Demand Estimation and Forecast Result .. - - - 1111 - - - - - MuM ••• ••• MFM MIM •• •• .,. Mi:M MQM M11M MJM M}W -· 4.8 3.1 7.7 21.8 54.1 119.7 272.3 1704.3 4937.6 3866.5 2117.5 1844.2 3500.2 4149.5 2597 1814.4 1407.9 1006.6 654.1 383.7 206.2 99.9 43.9 17.2 Ycarl~ Demand 4.8 3.1 7.7 21.8 54.0 119.5 272.1 1703.0 4933.1 3863.4 2115.8 1842.9 3497.8 4146.8 2595.2 1812.9 1406.7 1005.6 653.4 383.2 205.9 99.8 43.8 17.2 Upper (kWh) 4.8 3.1 7.7 21.8 54.2 119.9 272.5 1705.7 4941.7 3869.9 2119.1 1845.5 3502.8 4152.4 2599.0 1815.7 1409.0 1007.6 654.8 384.1 206.5 100.0 44.0 17.2 8.04 GWh C. Commercial Compact EV Demand Forecast Model This section presents the summary and results of the Commercial Compact EV model, with 24 LADWP carpool compact EVs' result and analysis period 05/20/2015 - 02/04/2017. • Training Set 05/20/2015 - 08/20/2016 245 USC Viterbi School of Engineering PhD Thesis • Testing Set: 08/21/2016 - 02/04/2017 C.1 Charging Event The summary of 24 commercial compact EVs charging events are summarized below, Table C. l, Figure C. l. Table C.1 Individual EV Charging Event of Commercial Compact EVs, 05/20/15 - 02/04/17 EVl EV2 EV3 EV4 EV5 EV6 EV7 EV8 EV9 EVlO EVll EV12 EV13 EV14 EV15 EV18 EV19 EV20 EV21 EV22 EV23 EV24 Average Average Percentage CE=O CE=l CE=2 CE=3 CE=4 CE=5 136 119 102 51 21 6 133 204 51 33 11 3 194 137 62 32 4 6 90 132 128 57 24 4 82 139 110 70 24 10 83 120 137 71 19 5 300 72 33 19 9 2 172 186 49 20 6 2 375 53 7 0 0 0 175 98 100 42 14 6 178 112 82 38 16 9 210 76 86 50 12 1 103 136 107 60 25 4 110 201 78 32 11 3 75 143 126 58 25 8 90 127 128 61 24 5 73 133 123 69 29 8 119 121 91 64 33 7 205 97 82 34 9 8 106 132 113 61 19 4 158 116 94 50 15 2 227 87 71 30 14 6 269 72 53 27 9 5 242 60 79 37 9 8 163 120 87 44 16 5 37.4% 27.5% 20.0% 10.2% 3.7% 1.2% 246 C" QJ ... .... USC Viterbi School of Engineering Charging Event of Commercial Compact EVs (n=24) 05/20/15 - 02/04/17 PhD Thesis - -l-11----w------- CE=O CE=l CE=2 CE=3 CE=4 CE=5 #of Charging Events per Day EVl • EV2 EV3 • EV4 EV5 EV6 • EV7 • EV8 • EV9 • EVlO • EVll • EV12 EV13 EV14 EV15 EV16 EV17 EV18 • EV19 EV20 EV21 Figure C. l Charging Event Distribution of Commercial Compact EV s, 05/20/15 - 02/04/17 C.2 Charging Start Time Distribution of the historical charging start time data on training and testing set are shown below, Figure C.2. 0.05 0.04 -~0.03 Vl c <lJ 0 0.02 0.01 0 Commercial Model (Compact): Charging Start Time Distribution Training vs Testing 05/20/15 - 02/04/17 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day • train • test Figure C.2 Charging Start Time Distribution of Commercial Compact EVs, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on CST training set is a combination of 2 Gaussian distributions, Figure C.3. 247 -·k $W_0-.-1 •-OtlJS :-"u!:OMH USC Viterbi School of Engineering •1"°""•~1t.ll:WZl•M" ..... _.~l) 0c---· . -' ' L . T L .. PhD Thesis Figure C.3 CST Curve Fitting on Training Set for Commercial Compact EVs The fitted pdf function for the training set is shown below: ( x-11.84) 2 (x-15.18) 2 f(x) = 0.03829e - 2.101 + 0.02459e- i.422 Validation on the testing set and residual plot can be found in Figure C.4. Charging Start Time Model (pdf} vs Actual Observation 0.06 ~~--------T-----,--~~--'-T----r------,---~--~ ~0.04 .iii c: Q) 0 0.02 -- CST Model + Actual 1 2 3 4 5 6 7 8 91011121314151617181920212223 Time of Day Residual Plot Charging Start Time 0. 02 ..--.---.r---.--,-----.-----r----r----.----r--r---.-----.---"'i---T---,--...--..--,---,r---.--,-----.-----r----, 0.01 Iii ::l ~ Of--------. ..... ~_.... ..... ._. ............. l""-.r"-'-----.-..--~~--l Q) c::: -0.01 -0.02 L--L.......l'--L--L--L--L.--'-_.__.__,__.__.__....__....___.__...__.__.__..__.__.__..__.___. 0 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day Figure C.4 CST Validation on Testing Set for Commercial Compact EVs 248 USC Viterbi School of Engineering PhD Thesis C.3 Plug-in Duration Distribution of historical plug-in duration on training and testing set are shown below, Figure C.5. 0.35 0.3 0.25 .£ 0.2 Vl c ~ 0.15 0.1 0.05 0 Commercial Model (Compact): Plug-in Duration Distribution Training vs Testing 05/20/15 - 02/04/17 I II II I I •• - · -- - - 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration {hr) • Train • Test Figure C.5 Plug-in Duration Distribution of Commercial Compact EVs, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a Gaussian distributions, Figure C.6. IW • •l'..pl-jt,. 1>111'<1!•.i> ~r----~ ol • oitU~Cll))AI lol • 0.ltol L ...... t..MJZI <1• 1111 O.l72-l.lt1J _ _, .. $$l.(lmllll1 ·-- ~~~·- ...... -- · I , I i r-·~-·J .... _ . Figure C.6 Plug-in Duration Curve Fitting on Training Set for Commercial Compact EVs ( x+0.1608) 2 f(x) = 0.2939e- 1.732 Validation on the testing set and residual plot can be found in Figure C. 7. 249 "iii ::I USC Viterbi School of Engineering Plug-in Duration Model (pdf} vs Actual Observation 0.3~-~-~-~--~-~-~-~~-~-~ .?;- 0.2 "iii c: Q) Cl 0.1 + - DModel + Actual Plug-in Duration (hr) Residual Plot Plug-in Duration 0.04 ~--~-~~-~-~=---,~~-~-~~-~ 0.02 ~ o~------ ~ -0.02 -0.04 .__ __ _._ _ _.____.. _ __._ _ _._____..____._ _ _.__..____,_ _ _, 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) PhD Thesis Figure C.7 Plug-in Duration Validation on Testing Set for Commercial Compact EVs C.4 Start SOC Distribution of historical charging start SOC on training and testing set are shown below, Figure C.8. 0.25 0.2 > 0.15 ...., ·v; c <lJ 0 0.1 0.05 0 Commercial Model (Compact} Start SOC Distribution Training vs Testing 05/20/15 - 02/04/17 0 3 6 9 12 1518 2124 27 30 33 36 39 42 45 48 5154 57 60 63 66 69 72 75 78 8184 87 90 93 96 99 Start SOC Interval{%) • Train • Test Figure C.8 SSOC Distribution of Commercial Compact EVs, 05/20/15 - 02/04/17 250 USC Viterbi School of Engineering PhD Thesis Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a Gaussian distributions, Figure C.9. ·-- ---· T T ~ •• 1·...c•"'""ll''l!> 1 ·=------ . • £--..~_-.. -._uoc ~"""',,,. ..................... .,. 0.00l'SIO.Ot1'9,o.ontl) - UCWS~MCC "' " IJIJ •7 1),11)<1) !"' ~···" cl• MZZISW.1'.~ 151.CJ:IOlltJ ·-~ ~o.!;"'°"m t '" !, ... ~ :·: a'" I 001 .. _._ssoc ~.1 1 .11~11 Figure C.9 SSOC Curve Fitting on Training Set for Commercial Compact EVs ( x-131.3) 2 f(x) = 0.0437Se- 56.22 Validation on the testing set and residual plot can be found in Figure C.10. o~s,r,--rrT"lrrT"lrrT"lrrT"lrc rh•~ tg1ng "T"r 11t11 0 rt 0 s_ o, c 0 M~ or "'r (p 0 dq~~ -Actu ,., •'- ob, u 0 rvat or io, •T"l--,T"l--,T"l--,l=l:::I:I:I:::r;1 - &BOC Model 0.2 + ._, Jl'0.15 ~ 0.1 0.05 + ol+-.++++44+44+.+-41-.++-l.~+44+.+-41-~#=l±!±±:!:±l±±±El±l:::l:'EE:t!:t:!::t!Il:!!In:!:l::EEE'..:EE +~:'.:::J 02468W12M~~~~~~u~~M~~~~«~~~~M~~~~MHMronNnn~~MHU~~M~~ Charging Start soc(%) 0.02 0.01 i ~~~~~~~~~~~~~---~~~.....-.~-----..._ ........ _._.11.1 ... 111111111 ........ -0.0t -0.02 0 2 4 6 8 10121418182022242628303234363840424446485052&456588062646668707274767880828488889092949698 Charging Start soc(%) Figure C.10 SSOC Validation on Testing Set for Commercial Compact EVs 251 USC Viterbi School of Engineering PhD Thesis C.5 Goodness of Fit Summary of the three goodness of fit indices can be found in Table C.2. Table C.2 Goodness of Fit Indices Summary of Commercial Compact EV Model Goodness Train Test of fit Index R- uare 98.35% 95.67% CST RMSE 0.0019 0.0030 (Gaussian 2) MAPE 0.1116 0.1148 R-square 98.98% 98.04% Duration RMSE 0.0123 0.0155 (Gaussian 1) MAPE 0.1132 0.1221 R-~uare 87.54% 69.68% Start SOC RMSE 0.0035 0.0159 (Gaussian 1) MAPE 0.0788 0.1275 C.6 Monte Carlo Simulation Result When number of Monte Carlo Simulation iterations is set to be 10,000 and number of Commercial Compact EV number is set to be 10,000, the results of EV charging demand estimation and forecast are provided in the following figures, Figure C.11 to Figure C.14. 252 USC Viterbi School of Engineering PhD Thesis Histogram of Monte Carlo Generated Charging Start Time Variables 1000 5000 Histogram of Monte Carlo Generated Plug-in Duration Variables 800 g 600 " " ~ 400 u. 4000 1000 200 I r1 0 ~..itl ............... ._.,._._............_._. ll , .........__. OI ~ O'--~_....._..-,.._.,.__.._...-.....-~-..~~--'-~~--' >. " c: 0 1 2 3 4 5 6 7 8 910 111213 14 151617181920212223 Time of Day ~lstogram of Monte Carlo Generated Charging Duration Variables 6000 ·1 Plug-in Duration so'(!lstogram of Monte Carlo Generated Charging Start SOC Variables 500 >. 400 " c: !!l 4000 ~ ~ 300 (ll u. 2000 O'--~_......_,._Lm....._,._.__...._~~~~--''--~--' ·1 Charging Duration u. 200 20 40 60 Battery SOC(%) 80 Figure C.11 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for Commercial Compact EV 4000 3500 3000 ~ .5 2500 "C ~ ~ 0 2000 "' c ·~ ~ .c () 1500 > w 1000 500 Commercial Compact EV Charging (Level 2) Demand Forecast Based on Monte Carlo Simulation (10000 EVs, 10000 times) 1 I I I r I I I 1 r I I I I oL-...J....__J~...J....~~"'""'!!!!!!!''.:___l__~~_l_~~_[__L~_[__L~~...L~~...L...:::'."'!l ... ..1....--L~_l__~~~---' 0 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day 100 Figure C.12 Monte Carlo Simulation Result for Commercial Compact EV Demand Model 253 4IXJO· ••• ~ · · ""' L .. ~ ~ 6 ,,., __ ~ USC Viterbi School of Engineering PhD Thesis Surf Plot ol Commerci1I Compact EV Ch1rv~ o.m.nd f orKHl S.Hd on M ont• Carlo Simulnion 110000 EV., 10000 limn) " .. ~.,_···· 15 ,· Figure C.13 Monte Carlo Simulation Result for Commercial Compact EV Demand Model - Surf Plot in MATLAB Surf Pio! with Spllrie of c-rc1 a1 Compact EV Chlf'lllng O emand Foreent 8Hed on Monte Carlo Slmulftkwl (1 0000 EV1, 10000 tlmn) " Norte carto 5"";Uaiion lrne Figure C.14 Monte Carlo Simulation Result for Commercial Compact EV Demand Model - Surf with Spline Plot in MATLAB 254 USC Viterbi School of Engineering PhD Thesis Averaged EV demand forecast in 24 hour format of 10,000 Commercial Compact EVs running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table C. 3. Table C.3 10,000 Commercial Compact EV Charging Demand Estimation and Forecast Result Average Hour I Demand (kWh) 95% Confidence Inten'al (Bootstrap) Lower (kWh) Upper (kWh) ' .. 0.0 ---~~==11-==--~-==~.--=-~--=-=- - 0.0 ~~~~:~~~-==~~=--~--=~ - 0.0 0.0 0.0 ---------------- - 0.0 0.0 0.0 ----------- 1111 0.8 0.8 0.8 ---------------- - 7.8 7.8 7.8 - ----------- 46 .4 46.4 46.4 194.7 195.1 - 194.9 "==~~==~~==~ - 613.4 612.8 613.9 1408.7 1411.1 - --~~-=::;=..~~--=~.--;;.._~___;-=1 1409.9 MuM 2383.9 ••• 3125.2 ••• 3350.2 MM 3204.o ••• 3168.0 •• 2944.1 •• 2015.1 .,. 831.7 Mi:M 1952 M@M 21.8 MJM 0.2 M}W o.o -· 0.0 Ycarl~ Demand 2381.9 3122.8 3347.9 3201.9 3165.9 2941.7 2013.3 830.9 195.0 27.8 2.8 0.2 0.0 0.0 255 2385.7 3127.6 3352.7 3206.2 3170.5 2946.2 2016.7 832.5 195.4 27.8 2.8 0.2 0.0 0.0 6.13 GWh USC Viterbi School of Engineering PhD Thesis D. Commercial Electric SUV Demand Forecast Model This section presents the summary and results of the Commercial Electric SUV model, with6 LADWPcarpool Electric SUVs' result and analysis period 05/20/2015 -02/04/2017. • Training Set: 05/20/2015 - 08/20/2016 • Testing Set: 08/21/2016 - 02/04/2017 D.1 Charging Event The summary of 6 commercial Electric SUVs charging events are summarized below, Table D.1, Figure D.1. Table D.l Individual EV Charging Event of Commercial Electric SUVs, 05/20/15- 02/04/17 CE=O A\'erage Percentage 165 182 256 170 39.1% 177 116 109 155 35,6% 73 15 5 78 39 20 54 15 1 78 24 8 18.0% 5,6% 1.7% Charging Event of Commertcial Electric SUVs (n=6} 05/20/16 - 02/04/17 CE=l CE=2 CE=3 CE=4 #of Charging Event per Day • EVl EV3 EV4 EVS EV6 . Average Figure D.1 Charging Event Distribution of Commercial Electric SUVs, 05/20/15 - 02/04/17 256 USC Viterbi School of Engineering PhD Thesis D.2 Charging Start Time Distribution of the historical charging start time data on training and testing set are shown below, Figure D.2. 0.1 0.08 ?: 0.06 ·v; c QJ 0 0.04 0.02 0 Commercial Model (SUV): Charging Start Time Distribution Training vs Testing 05/20/15 - 02/04/17 ............. 11 .. 11111 lljll I 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time of Day • train • test Figure D.2 Charging Start Time Distribution of Commercial Electric SUVs, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on CST training set is a combination of 2 Gaussian distributions, Figure D. 3. - --- lllf••l'...C•b1lo'<WU•.r...cr.Wlil<W'll '1f~~~: ~= ::.i:~:-::~ -·~ "'"'_,, . ·-·-· ="~'""' ooof ! O<M~ i',., I'"' ,_ ., ....... , ...... ,.ll•o2"""""'1oMl!"ll D<--- r· ! .t l l~ rl jf . I j .• I " 1 11 'II ' . T . .L_ " Figure D.3 CST Curve Fitting on Training Set for Commercial Electric SUVs The fitted pdf function for the training set is shown below: 257 USC Viterbi School of Engineering ( x-15.41) 2 (x-14.54) 2 f(x) = 0.01905e- o.7237 + 0.03111e- 4.os2 Validation on the testing set and residual plot can be found in Figure D.4. Charging Start Time Model (pdf) vs Actual Observation 0.1....--.----..--...--=r---T----..--....--.----..--...---.-----.--''r---i----.----..--...---.---.--..--~---...-~ + -- CST Model ~ + + Actual ·~ 0.05 CD c iii ::I "O & 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day Residual Plot Charging Start Time 0. 02 ....--.---...---.---...---.---.---.--..--.---.....--.----..--=;---r---,---.--.....--...--....--.---...---.---...-~ 0.01 -0.01 -0.02 ~~~~~~~~~~~~~~~~~~~~~~ 0 1 2 3 4 5 6 7 8 9 1011121314151617181920212223 Time of Day PhD Thesis Figure D.4 CST Validation on Testing Set for Commercial Electric SUVs D.3 Plug-in Duration Distribution of historical plug-in duration on training and testing set are shown below, Figure D.5. 258 0.35 0.3 0.25 -~ 0.2 Vl c ~ 0.15 0.1 0.05 0 0 USC Viterbi School of Engineering PhD Thesis Commercial Model (SUV}: Plug-in Duration Distribution Training vs Testing 05/20/15 - 02/04/17 ·~ II 1 1 II 11 •• ·- 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) • Train • Test Figure D.5 Plug-in Duration Distribution of Commercial Electric SUV s, 05/20/15 - 02/04/17 Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a Gaussian distributions, Figure D .6. ~;:::::.:.:..-. .,. O.lSZll. fl)ttU.o..100)) -· o 0.Hll ( 1 Its Otuf cl• l.O*CUU!Mtl j " ~ ... I " 1.-_ o1'..C , -1~11'1' D<-~- L ' ....... __ o __ o Figure D.6 Plug-in Duration Curve Fitting on Training Set for Commercial Electric SUVs ( x+0.1412) 2 f(x) = 0.2528e- 2.046 Validation on the testing set and residual plot can be found in Figure D.7. 259 USC Viterbi School of Engineering Plug-in Duration Model (pdf) vs Actual Observation 0.4 ~--~--~--.----=~----~---- 0.3 ~ 'iii 5i 0.2 c 0.1 + + - DModel + Actual OL_ _ _L_ _ ____._ _ _J_ __ L_ _ _L_ _ _r==::::=$==~'--~~ 0 0.5 1.5 2 2.5 3 3.5 4 4.5 Plug-in Duration (hr) Residual Plot Plug-In Duration 0.05 ~--~-~~-~-~'----.---~~-~- m o~--- :::J "O ~ -0.05 0 0.5 1~ 2 2.5 3 ~5 4 4~ Plug-in Duration (hr) PhD Thesis Figure D.7 Plug-in Duration Validation on Testing Set for Commercial Electric SUVs D.4 Start SOC Distribution of historical charging start SOC on training and testing set are shown below, Figure D.8. 0.045 0.04 0.035 0.03 > -w.025 c ~ 0.02 0.015 0.01 0.005 0 Commercial Model (SUV} Start SOC Distribution Training vs Testing 05/20/15 - 02/04/17 0 3 6 9 121518212427303336394245485154576063666972757881848790939699 Start SOC Interval(%) • Train • Test Figure D.8 SSOC Distribution of Commercial Electric SUV s, 05/20/15 - 02/04/17 260 USC Viterbi School of Engineering PhD Thesis Using MATLAB curve fitting toolbox, the pdf model on plug-in duration training set is a Gaussian distributions, Figure D. 9. ~-.1 ·-.i1 .. .,1~1)"'Z) ~-,,·---.. o1 • OOlHJ!O-OlMol,O.O.H.ol b1• 11•2(J"2,"-Q <I • 2"1QO.M.~ '· " !, • .,, ~ .. , !""" I 001 .. ~•SSOC ~··- ... . .. ti ,,,1ii,1 ,1 ii .. ( i I i ,J;~,jjl_, , : I 1::t.----~~~, - 11 ·-''lll.,-.,,,,.1 T''-'rr'-T' .~-!, ., ~,,1,,.,.,----,111 ~ I ~ .. ~ .._...._ssoc Figure D.9 SSOC Curve Fitting on Training Set for Commercial Electric SUVs ( x-75.12) 2 f(x) = 0.02553e - 23.31 Validation on the testing set and residual plot can be found in Figure D .10. 0.04 1> 0.03 !! ~ 0.02 0.01 + + + + 1~:uo; Model I + + + + + ++ + + + + +++ + + + 04+-H-+1+44+,H-+1+44+,-l.±+ ....... 4'R~+~= ++ t;rr::r~l.I±L.L...L.J_LL.L_L.J_LL...L..L.l_L_L...L..L.l_L_L++.-+.+J 0 2 4 8 8 10 12 14 18 18 2D 22 24 26 28 30 32 34 36 38 40 42 44 48 48 50 52 54 56 58 60 62 6C 68 88 70 72 74 76 78 80 82 84 86 88 90 92 9' 96 98 Charging start SOC(%) 0.01 ~ of--~~~~~~~~~~...--~-.,,....._,._. ..... .._...._..__......l,...llLl,,.DI,_._,.. ~ ~ ..(J.01 0 2 4 15 8 1012141818202224262830323436384042444648~52545e586062646668707274767880826486889092949698 Charging start soc(%) Figure D.10 SSOC Validation on Testing Set for Commercial Electric SUVs D.5 Goodness of Fit Summary of the three goodness of fit indices can be found in Table D.2. 261 USC Viterbi School of Engineering PhD Thesis Table D.2 Goodness of Fit Indices Summary of Commercial Electric SUV Model Goodness of fit Index Train Test R-square 96.41% 88.10% CST RMSE 0.0026 0.0058 (Gaussian 2) MAPE 0.0925 0.1278 R-square 96.50% 93.51 % Duration RMSE 0.0201 0.0279 (Gaussian 1) MAPE 0.0851 0.0853 R-~uare 84.05% 72.25% Start SOC RMSE 0.0041 0.0066 (Gaussian 1) MAPE 0.1218 0.0776 D.6 Monte Carlo Simulation Result When number of Monte Carlo Simulation iterations is set to be 10,000 and number of Commercial Electric SUV number is set to be 10,000, the results of EV charging demand estimation and forecast are provided in the following figures, Figure D .11 to Figure D .14. 1 oJ6istogram of Monte Carlo Generated Charging Start Time Variables 800 >- g 600 Q) :i ~ 400 u.. 200 0 0 1 2 3 4 5 6 7 8 9 10111213 14151617181920212223 Time ofDay 40 otUstogram of Monte Carlo Generated Charging Duration Variables 3000 i)' c: ~ 2000 ~ u.. 1000 O'--__......., .......... _._..-_.-...._.__~-'--~-'-~~ ·1 Charging Duration 4000 Histogram of Monte Carlo Generated Plug-in Duration Variables 3000 i)' c Q) :i 2000 ~ u.. 1000 O '----'......_-......-....c-.~~~...__~...._~--'-~~ ·1 0 3 Plug-in Duration 6 40 ~1stogram of Monte Carlo Generated Charging Start SOC Variables 300 >- 0 II c: ~ 200 _ .... ...n11~I~ 0- ~ u.. 100 0 i 0 20 40 60 80 100 Battery SOC (%) Figure D.11 Generated CST, Plug-in Duration, Charging Duration and SSOC Random Variables for Commercial Electric SUV 262 USC Viterbi School of Engineering PhD Thesis 9000 ~~~- Co _m ~m _ e_ ~_ ia_ I S~ U~ V_ EV_ C_ h_ arT gin~ g~ (L~ ev _ e_ l 2r ) D _ e~ m~ an_ d_ Fo_ re_ ca~ s_ tB_ •~ ••_ d_ on_ M_ o~ nre _C _ a_ rlo_S_ im ~ u_ la_ tio~ n~ (1_ 00_ 00_ E~ V_ s ,_ 10~ 00_ 0_ tim ~•- •l~~~~ 8000 7000 6000 5000 4000 3000 2000 1000 ol...~-1~~~--lool!!!!!!!!""~:__~~_JL_~_J_~l_~_J~~_l_~L__~_J~~---=~L_~_J 0 to 11 12 13 14 1s 1s 11 1 s 19 20 21 22 23 Time of Day Figure D.12 Monte Carlo Simulation Result for Commercial Electric SUV Demand Model Surf Plot ofComrn.rdal SUVEVCllflrg91g Demand Foncat BaMClon Monie CarloSimuAatlon 110000 EYs, 1 00001ime•) ---- ·- " Figure D.13 Monte Carlo Simulation Result for Commercial Electric SUV Demand Model - Surf Plot in MATLAB 263 7000 3000 2000 "'" .... .... USC Viterbi School of Engineering S~Plot with Spline of Com!MrcMI SUV EV Charging Demand F~t BneC on Mottt C.lo SW!Uation (10000 EVt, 10000 limes) -------------"'·~· . . ---------~;-----------.......:..··,§ 0- 0 TmeolDay PhD Thesis . ----=.::---------- ,. Figure D.14 Monte Carlo Simulation Result for Commercial Electric SUV Demand Model - Surf with Spline Plot in MATLAB Averaged EV demand forecast in 24 hour format of 10,000 Commercial Electric SUVs running Monte Carlo Simulation for 10,000 times is displayed below, and Bootstrap 95% confidence interval result can also be found below, Table D.3. 264 USC Viterbi School of Engineering PhD Thesis Table D.3 10,000 Commercial Electric SUV Charging Demand Estimation and Forecast Result Average Hour I Demand (kWh) .. - - - .. - - - - - MllM MIM MfW MFM MIM MM MfM MfM Mi:M Mp+ Wl1M WJM WJW •• 33.6 4.3 0.9 2.8 10.2 33.2 96.8 249.7 568.0 1144.1 2033.1 3191.3 4434.4 5492.6 6340.8 7848.1 7349.3 5338.1 3797.0 2536.7 1516.6 802.6 377.6 151.1 Ycarh Demand 95% Confidence Inten•al (Bootstrap) Lower (kWh) Upper (kWh) 33.5 33.6 4.3 4.3 0.9 0.9 2.8 2.8 10.2 10.2 33.1 33.2 96.7 96.9 249.5 249.9 567.5 568.5 1143.1 1145.0 - 2031.4 2034.6 - 3188.9 3193.7 - 4431.4 4437.7 - 5489.0 5496.4 - 6336.7 6345.7 - 7841.7 7853.6 - 7342.9 7355.1 - 5332.7 5343.5 - 3792.6 3800.9 - 2533.8 - 2539.7 - 1514.7 1518.5 801.6 803.6 377.1 378.2 150.9 151.3 13.91 GWh 265 USC Viterbi School of Engineering E. MATLAB Code for EV Demand Forecast Model E.1 Model for Industrial Area %% Industrial - Generate Random Number U, Ul, U2, U3 follows three pdf prompt_EV _num = 'Please enter the number oflndustrial EVs'; nl = input(prompt_EV _num); % nl = # of total Industrial EVs to run Monte Carlo prompt_IN_Comp = 'Please enter the percentage oflndustrial Compact EV>>'; pl = input(prompt_IN_Comp); prompt_IN_Mid = 'Please enter the percentage oflndustrial Midsize EV>>'; p2 = input(prompt_ IN_ Mid); prompt_ IN _SUV= 'Please enter the percentage oflndustrial Electric SUV>> '; p3 = input(prompt_IN_SUV); prompt_MC_num = 'Please enter Monte Carlo Simulation times>> '; n2 = input(prompt_MC_num); L2_P _IN_Comp = 3.3; % Level 2 charging power = 3.3 kW bat_IN_Comp = 12; % battery size of compact EV = 12 kWh rng('shufile'); rvl_CST_IN_Comp = ones(round(nl *pl),n2)*-l; rvl_D_IN_Comp = ones(round(nl *pl),n2)*-l; rvl_SSOC_IN_Comp = ones(round(nl *pl),n2)*-l; rv2_CST_IN_Comp = ones(round(nl *pl),n2)*-l; rv2 _ D _IN_ Comp= ones(round(nl *pl ),n2)*-l; rv2_SSOC_IN_Comp = ones(round(nl *pl),n2)*-l; rv3_CST_IN_Comp = ones(round(nl *pl),n2)*-l; rv3_D_IN_Comp = ones(round(nl *pl),n2)*-l; rv3_SSOC_IN_Comp = ones(round(nl *pl),n2)*-l; rv4_CST_IN_Comp = ones(round(nl *pl),n2)*-l; rv4_D _IN_ Comp= ones(round(nl *pl),n2)*-l; rv4_SSOC_IN_Comp = ones(round(nl *pl ),n2)*-l; wl_CE_IN_Comp = 0.6098; w2 _CE _IN_ Comp= 0.2697; w3_CE_IN_Comp = 0.0989; w4_CE_IN_Comp = 0.0169; w5_CE_IN_Comp = 0.0047; CST_ al _IN_ Comp= CST_bl_IN_Comp = CST_cl_IN_Comp = CST_a2_IN_Comp = CST_b2_IN_Comp = CST_ c2 _IN_ Comp= CST_ a3 _IN_ Comp= CST_b3_IN_Comp = CST_ c3 _IN_ Comp = 0.04742;% (0.04314, 0.05171) 6.647;% (6.612, 6.681) 0.4718;% (0.4211, 0.5225) 0.02051;% (0.01717, 0.02385) 12.39;% (12.15, 12.63) 3.443;% (2.877, 4.009) 0.004491 ;% (0.001795, 0.007186) 16.97;% (4.826, 29.12) 16.33;% (-3.42, 36.08) mul_CST_IN_Comp = CST_bl_IN_Comp; % meanl_CST = bl in Gaussian pdf mu2_CST_IN_Comp = CST_b2_IN_Comp; mu3_CST_IN_Comp = CST_b3_IN_Comp; PhD Thesis stdl_CST_IN_Comp = CST_cl_IN_Comp/sqrt(2); % standard deviation 1 CST= cl/sqrt(2) in Gaussian pdf std2_CST_IN_Comp = CST_c2_IN_Comp/sqrt(2); 266 USC Viterbi School of Engineering PhD Thesis std3_CST_IN_Comp = CST_c3_IN_Comp/sqrt(2); coefl_CST_IN_Comp = stdl_CST_IN_Comp*sqrt(2*pi)*CST_al_IN_Comp; % coefficient! of CST = std*sqrt(2*pi)*al coef2 _CST _IN_ Comp = std2 _CST _IN_ Comp*sqrt(2*pi)*CST _ a2 _IN_ Comp; coef3 _CST _IN_ Comp= std3 _CST _IN_ Comp*sqrt(2*pi)*CST _ a3 _IN_ Comp; wl_CST_IN_Comp = coefl_CST_IN_Comp/(coefl_CST_IN_Comp + coef2_CST_IN_Comp + coef3 _CST _IN_ Comp); % weightl of CST = coefl/sum( coefl +coef2+coef3); w2_CST_IN_Comp = coef2_CST_IN_Comp/(coefl_CST_IN_Comp + coef2_CST_IN_Comp + coef3 _CST _IN_ Comp); w3_CST_IN_Comp = coef3_CST_IN_Comp/(coefl_CST_IN_Comp + coef2_CST_IN_Comp + coef3 _CST _IN_ Comp); D_al_IN_Comp = 0.1331;% (0.06614, 0.2001) D_bl_IN_Comp = 2.075;% (1.832, 2.318) D_cl_IN_Comp = 0.5632;% (0.239, 0.8874) D _a2_IN_Comp = 0.1503;% (0.1166, 0.1839) D_b2_IN_Comp = 0.4716;% (0.2029, 0.7402) D _c2_IN_Comp = 1.091;% (0.1503, 2.032) D_a3_IN_Comp= 0.1268;% (0.0922, 0.1613) D_b3_IN_Comp = 3.588;% (3.449, 3.728) D _c3_IN_Comp = 0.5581;% (0.3633, 0.7529) mul_D_IN_Comp = D_bl_IN_Comp; o/omean_D =bl in Gaussian pdf mu2_D_IN_Comp = D_b2_IN_Comp; mu3_D_IN_Comp = D_b3_IN_Comp; stdl_D_IN_Comp = D_cl_IN_Comp/sqrt(2); % standard deviation 1 D = cl/sqrt(2) in Gaussian pdf std2_D_IN_Comp = D_c2_IN_Comp/sqrt(2); std3_D_IN_Comp = D_c3_IN_Comp/sqrt(2); coefl_D_IN_Comp = stdl_D_IN_Comp*sqrt(2*pi)*D_al_IN_Comp; % coefficient! of D std*sqrt(2*pi)*al coef2_D _IN_ Comp= std2_D _IN_ Comp*sqrt(2*pi)*D _a2_IN_ Comp; coef3 _ D _IN_ Comp = std3 _ D _IN_ Comp*sqrt(2 *pi )*D _ a3 _IN_ Comp; wl_D_IN_Comp = coefl_D_IN_Comp/(coefl_D_IN_Comp + coef2_D_IN_Comp + coef3_D_IN_Comp); % weightl ofD = coefl/sum(coefl:coef3); w2_D_IN_Comp = coef2_D_IN_Comp/(coefl_D_IN_Comp + coef2_D_IN_Comp + coef3_D_IN_Comp); w3_D_IN_Comp = coef3_D_IN_Comp/(coefl_D_IN_Comp + coef2_D_IN_Comp + coef3_D_IN_Comp); SSOC_a_IN_Comp = 0.3172;% (0.3098, 0.3245) SSOC_b_IN_Comp = -2.279;% (-2.503, -2.056) SSOC_c_IN_Comp = 0.002259;% (0.00143, 0.003088) SSOC_d_IN_Comp = 0.01843;% (0.01373, 0.02313) mul_SSOC_IN_Comp = -1/SSOC_b_IN_Comp; % mean_SSOC =-lib mu2_SSOC_IN_Comp = -1/SSOC_d_IN_Comp; coefl_SSOC_IN_Comp = SSOC_a_IN_Comp * mul_SSOC_IN_Comp; % coefficient! ofSSOC = a*mean coef2_SSOC_IN_Comp = SSOC_c_IN_Comp * mu2_SSOC_IN_Comp; w l _ SSOC _IN_ Comp coefl _ SSOC _IN_ Comp/( abs( coefl _ SSOC _IN_ Comp) + abs( coef2_SSOC _IN_ Comp)); w2_SSOC_IN_Comp = 1 - wl_SSOC_IN_Comp; for j=l:n2 for i=l:nl *pl ul = rand; % ul =random variable to check how many times the EV is charging today u2 = rand; % u2=random variable follows CST pdf u3 = rand; % u3=random variable follows D pdf u4 = rand; % u4=random variable follows SSOC pdf iful > wl_CE_IN_Comp % iful > probability ofO charging event today, (or CE >= 1), generate CSTl, Dl, SSOCl while rvl_CST_IN_Comp(i,j) <= 0 II rvl_CST_IN_Comp(i,j) > 23.5 267 USC Viterbi School of Engineering PhD Thesis if u2 <= wl_CST_IN_Comp rvl_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; else if u2<= wl _CST_IN_Comp + w2_CST_IN_Comp rvl_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; end end else rv l _CST _IN_ Comp(i,j) = std3 _CST _IN_ Comp*randn + mu3 _CST _IN_ Comp; end while rv l _ D _IN_ Comp(i,j) <= 0 if u3 <= wl_D_IN_Comp end rvl_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; else if u3 <= wl_D_IN_Comp + w2_D_IN_Comp end rv l _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; else rvl_D_IN_Comp(i,j) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; end while rvl_SSOC_IN_Comp(i,j) <= 0 II rvl_SSOC_IN_Comp(i,j) > 100 if u4 <= wl_SSOC_IN_Comp rvl_SSOC_IN_ Comp(i,j) = exprnd(mul_SSOC_IN_ Comp); else rvl_SSOC_IN_Comp(i,j) = 3.3*log(rand+l)/SSOC_d_IN_Comp; % mverse transform of the second exponential term ofSSOC pdf end end end iful > wl_CE_IN_Comp + w2_CE_IN_Comp % iful > probabilityofl charging event today, (or CE >= 2), generate CST2, D2, SSOC2 while rv2_CST_IN_Comp(i,j) <= 0 II rv2_CST_IN_Comp(i,j) > 23.5 if u2 <= wl_CST_IN_Comp end rv2_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; else ifu2<= wl_CST_IN_Comp + w2_CST_IN_Comp rv2_CST_IN_Comp(i,j) = std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; else rv2_CST_IN_Comp(i,j) = std3_CST_IN_Comp*randn + mu3_CST_IN_Comp; end end while rv2 _ D _IN_ Comp(i,j) <= 0 if u3 <= wl_D_IN_Comp end rv2_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; else if u3 <= wl_D_IN_Comp + w2_D_IN_Comp end rv2 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; else rv2_D_IN_Comp(ij) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; end while rv2_SSOC_IN_Comp(i,j) <= 0 II rv2_SSOC_IN_Comp(i,j) > 100 if u4 <= wl_SSOC_IN_Comp rv2 _SSOC _IN_ Comp(i,j) = exprnd(mul _SSOC _IN_ Comp); else rv2_SSOC_IN_Comp(i,j) = 3.3*log(rand+l)/SSOC_d_IN_Comp; % mverse 268 USC Viterbi School of Engineering PhD Thesis transform of the second exponential term ofSSOC pdf end end end if ul > wl_CE_IN_Comp + w2_CE_IN_Comp + w3_CE_IN_Comp % iful > probability of2 charging event today, (or CE >= 3), generate CST3, D3, SSOC3 while rv3_CST_IN_Comp(i,j) <= 0 II rv3_CST_IN_Comp(i,j) > 23.5 if u2 <= wl_CST_IN_Comp end rv3_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; else ifu2<= wl_CST_IN_Comp + w2_CST_IN_Comp rv3_CST_IN_Comp(i,j) = std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; else rv3_CST_IN_Comp(i,j) = std3_CST_IN_Comp*randn + mu3_CST_IN_Comp; end end while rv3_D_IN_Comp(i,j) <= 0 if u3 <= wl_D_IN_Comp end rv3_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; else if u3 <= wl_D_IN_Comp + w2_D_IN_Comp end rv3 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; else rv3_D_IN_Comp(i,j) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; end while rv3_SSOC_IN_Comp(i,j) <= 0 II rv3_SSOC_IN_Comp(i,j) > 100 if u4 <= wl_SSOC_IN_Comp rv3_SSOC_IN_ Comp(i,j) = exprnd(mul_SSOC_IN_ Comp); else rv3 _ SSOC _IN_ Comp(i,j) = 3.3*log(rand+ 1 )/SSOC _ d _IN_ Comp; % mverse transform of the second exponential term ofSSOC pdf end end end if ul > wl_CE_IN_Comp + w2_CE_IN_Comp + w3_CE_IN_Comp + w4_CE_IN_Comp % if ul > probability of 3 charging event today, (or CE = 4 ), generate CST 4, D4, SSOC4 while rv4_CST_IN_Comp(i,j) <= 0 II rv4_CST_IN_Comp(i,j) > 23.5 if u2 <= wl_CST_IN_Comp end rv4_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; else ifu2<= wl_CST_IN_Comp + w2_CST_IN_Comp rv4_CST_IN_Comp(i,j) = std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; else rv4_CST_IN_Comp(i,j) = std3_CST_IN_Comp*randn + mu3_CST_IN_Comp; end end while rv4 _ D _IN_ Comp(i,j) <= 0 if u3 <= wl_D_IN_Comp rv4_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; else if u3 <= wl_D_IN_Comp + w2_D_IN_Comp rv4 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; else rv4_D_IN_Comp(i,j) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; end 269 USC Viterbi School of Engineering end end while rv4_SSOC_IN_Comp(i,j) <= 0 II rv4_SSOC_IN_Comp(i,j) > 100 if u4 <= wl_SSOC_IN_Comp rv4_SSOC_IN_ Comp(i,j) = exprnd(mul_SSOC_IN_ Comp); PhD Thesis else rv4_SSOC_IN_Comp(i,j) = 3.3*log(rand+l)/SSOC_d_IN_Comp; % inverse transform of the second exponential term ofSSOC pdf end end end end end % determine if there is any overlap between charging event 2 and charging % event 1. If there is, re-generate CST2, D2 and re-determine for j=l:n2 for i=l:nl *pl if rv2_ CST _IN_ Comp(i,j)>O while rv2_CST_IN_Comp(i,j) + rv2_D_IN_Comp(i,j) > rvl_CST_IN_Comp(i,j) && rv2_CST_IN_Comp(i,j) < rvl_CST_IN_Comp(i,j) + rvl_D_IN_Comp(i,j) u22 =rand; u33 =rand; if u22 <= wl_CST_IN_Comp rv2_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; while rv2_CST_IN_Comp(i,j)<=O II rv2_CST_IN_Comp(i,j)>23.5 rv2_CST_IN_Comp(i,j) stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; end else ifu22<= wl_CST_IN_Comp + w2_CST_IN_Comp rv2_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; while rv2_CST_IN_Comp(i,j)<=O II rv2_CST_IN_Comp(i,j)>23.5 rv2_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; end else rv2_ CST_IN_ Comp(i,j) std3 _CST_ IN_ Comp*randn + mu3_CST_IN_Comp; while rv2_CST_IN_Comp(i,j)<=O II rv2_CST_IN_Comp(i,j)>23.5 rv2_CST_IN_Comp(i,j) std3_CST_IN_Comp*randn + mu3_CST_IN_Comp; end end end if u33 <= wl_D_IN_Comp rv2_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; while rv2_D_IN_Comp(i,j) <= 0 rv2 _ D _IN_ Comp(i,j) = stdl _ D _IN_ Comp*randn + mul _ D _IN_ Comp; end else ifu33 <= wl_D_IN_Comp + w2_D_IN_Comp rv2 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; while rv2_D_IN_Comp(i,j) <= 0 rv2 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; end 270 end end end end end else end USC Viterbi School of Engineering PhD Thesis rv2_D_IN_Comp(i,j) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; while rv2_D_IN_Comp(i,j) <= 0 rv2 _ D _IN_ Comp(i,j) = std3 _ D _IN_ Comp*randn + mu3 _ D _IN_ Comp; end % determine ifthere is any overlap between charging event 3 and charging % event 1, 2. If there is, re-generate CST3, D3 and re-determine for j=l:n2 for i=l:nl *pl if rv3 _CST_ IN_ Comp(i,j)>O while (rv3_CST_IN_Comp(i,j) + rv3_D_IN_Comp(i,j) > rvl_CST_IN_Comp(i,j) && rv3_CST_IN_Comp(i,j) < rvl_CST_IN_Comp(i,j) + rvl_D_IN_Comp(i,j)) II (rv3_CST_IN_Comp(i,j) + rv3_D_IN_Comp(i,j) > 1v2_CST_IN_Comp(i,j) && rv3_CST_IN_Comp(i,j) < rv2_CST_IN_Comp(i,j) + rv2 _ D _IN_ Comp(i,j)) u22 =rand; u33 =rand; if u22 <= wl_CST_IN_Comp rv3_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; while rv3_CST_IN_Comp(i,j)<=O II rv3_CST_IN_Comp(i,j)>23.5 rv3_CST_IN_Comp(i,j) stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; end else ifu22<= wl_CST_IN_Comp + w2_CST_IN_Comp rv3_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; while rv3_CST_IN_Comp(i,j)<=O II rv3_CST_IN_Comp(i,j)>23.5 rv3_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; end else rv3 _ CST_IN_ Comp(i,j) std3 _CST_ IN_ Comp*randn + mu3_CST_IN_Comp; while rv3_CST_IN_Comp(i,j)<=O II rv3_CST_IN_Comp(i,j )>23.5 rv3 _CST _IN_ Comp(i,j) std3 _CST _IN_ Comp*randn + mu3_CST_IN_Comp; end end end if u33 <= wl_D_IN_Comp rv3_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; while rv3_D_IN_Comp(i,j) <= 0 rv3 _ D _IN_ Comp(i,j) = stdl _ D _IN_ Comp*randn + mul _ D _IN_ Comp; end else ifu33 <= wl_D_IN_Comp + w2_D_IN_Comp rv3 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; while rv3_D_IN_Comp(i,j) <= 0 rv3 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; end 271 end end end end end else end USC Viterbi School of Engineering PhD Thesis rv3_D_IN_Comp(i,j) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; while rv3_D_IN_Comp(i,j) <= 0 rv3 _ D _IN_ Comp(i,j) = std3 _ D _IN_ Comp*randn + mu3 _ D _IN_ Comp; end % determine ifthere is any overlap between charging event 4 and charging % event 1, 2 and 3. If there is, re-generate CST4, D4 and re-determine for j=l:n2 for i=l:nl *pl if rv4 _CST_ IN_ Comp(i,j)>O while (rv4_CST_IN_Comp(i,j) + rv4_D_IN_Comp(i,j) > rvl_CST_IN_Comp(i,j) && rv4_CST_IN_Comp(i,j) < rvl_CST_IN_Comp(i,j) + rvl_D_IN_Comp(i,j)) II (rv4_CST_IN_Comp(i,j) + rv4_D_IN_Comp(i,j) > 1v2_CST_IN_Comp(i,j) && rv4_CST_IN_Comp(i,j) < rv2_CST_IN_Comp(i,j) + rv2_D_IN_Comp(i,j)) II (rv4_CST_IN_Comp(i,j) + rv4_D_IN_Comp(i,j) > rv3_CST_IN_Comp(i,j) && rv4_CST_IN_Comp(i,j) < rv3_CST_IN_Comp(i,j) + rv3_D_IN_Comp(i,j)) u22 =rand; u33 = rand; if u22 <= wl_CST_IN_Comp rv4_CST_IN_Comp(i,j) = stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; while rv4_CST_IN_Comp(i,j)<=O II rv4_CST_IN_Comp(i,j)>23.5 rv4_CST_IN_Comp(i,j) stdl_CST_IN_Comp*randn + mul_CST_IN_Comp; end else ifu22<= wl_CST_IN_Comp + w2_CST_IN_Comp rv4_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; while rv4_CST_IN_Comp(i,j)<=O II rv4_CST_IN_Comp(i,j)>23.5 rv4_CST_IN_Comp(i,j) std2_CST_IN_Comp*randn + mu2_CST_IN_Comp; end else rv4_ CST_IN_ Comp(i,j) std3 _CST_ IN_ Comp*randn + mu3_CST_IN_Comp; while rv4_CST_IN_Comp(i,j)<=O II rv4_CST_IN_Comp(i,j )>23.5 rv4 _CST _IN_ Comp(i,j) std3 _CST_ IN_ Comp*randn + mu3_CST_IN_Comp; end end end if u33 <= wl_D_IN_Comp rv4_D_IN_Comp(i,j) = stdl_D_IN_Comp*randn + mul_D_IN_Comp; while rv4_D_IN_Comp(i,j ) <= 0 rv4 _ D _IN_ Comp(i,j) = stdl _ D _IN_ Comp*randn + mul _ D _IN_ Comp; end else ifu33 <= wl_D_IN_Comp + w2_D_IN_Comp rv4 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; while rv4_D_IN_Comp(i,j) <= 0 rv4 _ D _IN_ Comp(i,j) = std2 _ D _IN_ Comp*randn + mu2 _ D _IN_ Comp; 272 USC Viterbi School of Engineering PhD Thesis end else rv4_D_IN_Comp(i,j) = std3_D_IN_Comp*randn + mu3_D_IN_Comp; while rv4_D_IN_Comp(i,j) <= 0 rv4 _ D _IN_ Comp(i,j) = std3 _ D _IN_ Comp*randn + mu3 _ D _IN_ Comp; end end end end end end end Dl_actual_IN_Comp = ones(round(nl *pl),n2)*-l; EVl_demand_IN_Comp = zeros(round(nl *pl),n2); D2_actual_IN_Comp = ones(round(nl *pl),n2)*-l; EV2_demand_IN_Comp = zeros(round(nl *pl),n2); D3_actual_IN_Comp = ones(round(nl *pl),n2)*-l; EV3 _demand_ IN_ Comp = zeros( round( nl *p 1 ),n2); D4_actual_IN_Comp = ones(round(nl *pl),n2)*-l; EV4_demand_IN_Comp = zeros(round(nl *pl),n2); for j=l:n2 for i=l:nl *pl if rvl_D _IN_ Comp(i,j)>O && rv l_ SSOC _IN_ Comp(i,j)< lOO EVl_demand_IN_Comp(i,j) = min(min(rvl_D_IN_Comp(i,j)*L2_P _IN_Comp, (100- rvl_SSOC_IN_ Comp(i,j))/l OO*bat_IN _Comp), bat_ IN_ Comp); Dl_actual_IN_Comp(i,j) = EVl_demand_IN_Comp(i,j)/L2_P _IN_Comp; end if rv2_D_IN_Comp(i,j)>O && rv2_SSOC_IN_Comp(i,j)<l00 EV2_demand_IN_Comp(i,j) = min(min(rv2_D_IN_Comp(i,j)*L2_P _IN_Comp, (100- rv2_SSOC_IN_ Comp(i,j))/l OO*bat_IN _Comp), bat_ IN_ Comp); D2_actual_IN_Comp(i,j) = EV2_demand_IN_Comp(i,j)/L2_P _IN_Comp; end if rv3_D_IN_Comp(i,j)>O && rv3_SSOC_IN_Comp(i,j)<l00 EV3_demand_IN_Comp(i,j) = min(min(rv3_D_IN_Comp(i,j)*L2_P _IN_Comp, (100- rv3_SSOC_IN_ Comp(i,j))/l OO*bat_IN _Comp), bat_ IN_ Comp); D3 _actual _IN_ Comp(i,j) = EV3 _demand _IN_ Comp(i,j )/L2 _ P _IN_ Comp; end if rv4 _D _IN_ Comp(i,j)>O && rv4_SSOC _IN_ Comp(i,j)< lOO EV4_demand_IN_Comp(i,j) = min(min(rv4_D_IN_Comp(i,j)*L2_P _IN_Comp, (100- rv4_SSOC_IN_ Comp(i,j))/l OO*bat_IN _Comp), bat_ IN_ Comp); D4_actual_IN_Comp(i,j) = EV4_demand_IN_Comp(i,j)/L2_P _IN_Comp; end end end rv_CST_IN_Comp [rvl_CST_IN_Comp; rv2_CST_IN_Comp; rv3_CST_IN_Comp; rv4_CST_IN_Comp]; rv_D_IN_Comp = [rvl_D_IN_Comp; rv2_D_IN_Comp; rv3_D_IN_Comp; rv4_D_IN_Comp]; rv_SSOC_IN_Comp [rvl_SSOC_IN_Comp; rv2_SSOC_IN_Comp; rv3_SSOC_IN_Comp; rv4_SSOC_IN_Comp]; D _actual_ IN_ Comp [D l _actual_ IN_ Comp; D2 _actual_ IN_ Comp; D3 _actual _IN_ Comp; D4_actual_IN_ Comp]; 273 USC Viterbi School of Engineering L2_P _IN_Mid = 3.3; % Level 2 charging power= 3.3 kW bat_IN_Mid = 6.2; % battery size ofmidsize EV= 6.2 kWh rvl_CST_IN_Mid = ones(round(nl *p2),n2)*-l; rvl_D_IN_Mid = ones(round(nl *p2),n2)*-l; rvl_SSOC_IN_Mid = ones(round(nl *p2),n2)*-l; rv2_CST_IN_Mid = ones(round(nl *p2),n2)*-l; rv2_D_IN_Mid = ones(round(nl *p2),n2)*-l; rv2_SSOC_IN_Mid = ones(round(nl *p2),n2)*-l; rv3_CST_IN_Mid = ones(round(nl *p2),n2)*-l; rv3_D_IN_Mid = ones(round(nl *p2),n2)*-l; rv3_SSOC_IN_Mid = ones(round(nl *p2),n2)*-l; rv4_CST_IN_Mid = ones(round(nl *p2),n2)*-l; rv4_D_IN_Mid = ones(round(nl *p2),n2)*-l; rv4_SSOC_IN_Mid = ones(round(nl *p2),n2)*-l; wl_CE_IN_Mid = 0.5877; w2_CE_IN_Mid = 0.2677; w3_CE_IN_Mid = 0.1046; w4_CE_IN_Mid = 0.0346; w5_CE_IN_Mid = 0.0054; CST_al_IN_Mid = 0.06624;% (0.05631, 0.07616) CST_bl_IN_Mid = 6.166;% (6.074, 6.257) CST_cl_IN_Mid = 0.7674;% (0.6284, 0.9064) CST_a2_IN_Mid = 0.02274;% (0.01061, 0.03488) CST_b2_IN_Mid = 15.93;% (15.72, 16.15) CST_c2_IN_Mid = 0.5137;% (0.1816, 0.8458) CST_a3_IN_Mid = 0.01515;% (0.0111, 0.0192) CST_b3_IN_Mid = 13.16;% (11.89, 14.42) CST_c3_IN_Mid = 5.307;% (3.583, 7.032) mul_CST_IN_Mid = CST_bl_IN_Mid; % meanl *p2_CST = bl in Gaussian pdf mu2_CST_IN_Mid = CST_b2_IN_Mid; mu3_CST_IN_Mid = CST_b3_IN_Mid; PhD Thesis stdl_CST_IN_Mid = CST_cl_IN_Mid/sqrt(2); % standard deviation 1 CST= cl/sqrt(2) in Gaussian pdf std2_CST_IN_Mid = CST_c2_IN_Mid/sqrt(2); std3_CST_IN_Mid = CST_c3_IN_Mid/sqrt(2); coefl_CST_IN_Mid = stdl_CST_IN_Mid*sqrt(2*pi)*CST_al_IN_Mid; % coefficient! of CST std*sqrt(2*pi)*al coef2 _CST _IN_ Mid= std2 _CST_ IN_ Mid*sqrt(2*pi)*CST _ a2 _IN_ Mid; coef3_ CST _IN_Mid = std3 _ CST_IN_Mid*sqrt(2*pi)*CST _a3 _IN_Mid; wl_CST IN Mid coefl_CST_IN_Mid/(coefl_CST_IN_Mid + coef2 CST IN Mid + coef3_CST_IN_Mid); % weightl of CST = coefl/sum(coefl +coef2+coef3); w2 _CST IN Mid coef2 _CST _IN_ Mid/( coefl _CST _IN_ Mid + coef2 CST IN Mid + coef3 _CST _IN_ Mid); w3 _CST _IN_ Mid coef3 _CST_ IN_ Mid/( coefl _CST _IN_ Mid + coef2 CST IN Mid + coef3 _CST _IN_ Mid); D_al_IN_Mid = 1.509;% (-1.546e+07, 1.546e+07) D _bl_IN_Mid = 2.17;% (-3.278e+05, 3.278e+05) D_cl_IN_Mid = 0.1492;% (-3.04e+05, 3.04e+05) D_a2_IN_Mid = 0.2227;% (0.1398, 0.3056) D _b2_IN_Mid = 0.8723;% (0.462, 1.283) D_c2_IN_Mid = 0.6463;% (0.0136, 1.279) mul_D_IN_Mid = D_bl_IN_Mid; %mean_D =bl in Gaussian pdf mu2 _ D _IN_ Mid = D _ b2 _IN_ Mid; stdl_D_IN_Mid = D_cl_IN_Mid/sqrt(2); % standard deviation 1 D = cl/sqrt(2) in Gaussian pdf 274 USC Viterbi School of Engineering PhD Thesis std2_D_IN_Mid = D_c2_IN_Mid/sqrt(2); coefl_D_IN_Mid = stdl_D_IN_Mid*sqrt(2*pi)*D_al_IN_Mid; % coefficient! ofD = std*sqrt(2*pi)*al coef2_D_IN_Mid = std2_D_IN_Mid*sqrt(2*pi)*D_a2_IN_Mid; wl_D_IN_Mid = coefl_D_IN_Mid/(coefl_D_IN_Mid + coef2_D_IN_Mid); % weightl of D = coefl/sum( coefl :coef3 ); w2 _ D _IN_ Mid = coef2 _ D _IN_ Mid/( coefl _ D _IN_ Mid + coef2 _ D _IN_ Mid); SSOC_a_IN_Mid = 0.6136;% (0.6083, 0.6189) SSOC_b_IN_Mid = -2.784;% (-2.925, -2.644) SSOC_c_IN_Mid = 0.002848;% (0.001868, 0.003828) SSOC_d_IN_Mid = 0.003818;% (-0.001633, 0.009269) mul_SSOC_IN_Mid = -1/SSOC_b_IN_Mid; % mean_SSOC = -1/b, where b = -lambda mu2_SSOC_IN_Mid = -1/SSOC_d_IN_Mid; coefl_SSOC_IN_Mid = SSOC_a_IN_Mid * mul_SSOC_IN_Mid; % coefficient! ofSSOC = a*mean coef2_SSOC_IN_Mid = SSOC_c_IN_Mid * mu2_SSOC_IN_Mid; % wl_SSOC = coefl_SSOC/(abs(coefl_SSOC) + abs(coef2_SSOC)); % w2_SSOC = 1 - wl_SSOC; wl_SSOC_IN_Mid = 0.95; w2_SSOC_IN_Mid = 1 - wl_SSOC_IN_Mid; for j=l:n2 for i=l:nl *p2 ul = rand; % ul =random variable to check how many times the EV is charging today u2 = rand; % u2=random variable follows CST pdf u3 = rand; % u3=random variable follows D pdf u4 = rand; % u4=random variable follows SSOC pdf if ul > w l _CE_ IN_ Mid % if ul > probability of 0 charging event today, (or CE >= 1 ), generate CSTl, Dl, SSOCl while rvl _CST_ IN_ Mid(i,j ) <= 0 1 1 rvl _CST _IN_ Mid(i,j) > 23.5 if u2 <= wl CST IN Mid end end - - - rvl_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; else if u2<= wl CST IN Mid + w2 CST IN Mid - - - - - - rvl_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; else rvl_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end while rvl_D_IN_Mid(i,j) <= 0 if u3 <= wl D IN Mid end rvl_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; else rvl_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end while rvl_SSOC_IN_Mid(i,j) <= 0 II rvl_SSOC_IN_Mid(i,j) > 100 if u4 <= wl SSOC IN Mid - - - rvl_SSOC_IN_Mid(i,j) = exprnd(mul_SSOC_IN_Mid); else rvl_SSOC_IN_Mid(i,j) = log(rand+l)/(3.9*SSOC_d_IN_Mid); transform of the second exponential term ofSSOC pdf % mverse end end end if ul > wl_CE_IN_Mid + w2_CE_IN_Mid % if ul > probability of 1 charging event today, (or CE >= 2), generate CST2, D2, SSOC2 while rv2_CST_IN_Mid(i,j) <= 0 II rv2_CST_IN_Mid(i,j) > 23.5 275 USC Viterbi School of Engineering PhD Thesis if u2 <= wl CST IN Mid - - - rv2_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; end else ifu2<= wl CST IN Mid+ w2 CST IN Mid - - - - - - rv2_CST_IN_Mid(ij ) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; else rv2_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end end while rv2_D_IN_Mid(i,j) <= 0 if u3 <= wl D IN Mid end rv2_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; else rv2_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end while rv2_SSOC_IN_Mid(i,j) <= 0 II rv2_SSOC_IN_Mid(i,j) > 100 if u4 <= wl SSOC IN Mid - - - rv2_SSOC_IN_Mid(i,j) = exprnd(mul_SSOC_IN_Mid); else rv2_SSOC_IN_Mid(i,j) = log(rand+l)/(3.9*SSOC_d_IN_Mid); transform of the second exponential term ofSSOC pdf % mverse end end end iful > wl_CE_IN_Mid + w2_CE_IN_Mid + w3_CE_IN_Mid % iful > probability of2 charging event today, (or CE >= 3 ), generate CST3, D3, SSOC3 while rv3_CST_IN_Mid(i,j) <= 0 II rv3_CST_IN_Mid(i,j) > 23.5 if u2 <= wl CST IN Mid - - - rv3_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; else ifu2<= wl CST IN Mid+ w2 CST IN Mid - - - - - - rv3_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; else rv3_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end end end while rv3 _ D _IN_ Mid(i,j) <= 0 if u3 <= wl D IN Mid end rv3_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; else rv3_D_IN_Mid(ij) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end while rv3_SSOC_IN_Mid(i,j) <= 0 II rv3_SSOC_IN_Mid(i,j) > 100 if u4 <= wl SSOC IN Mid - - - rv3 _ SSOC _IN _Mid(i,j) = exprnd(mul_ SSOC _IN _Mid); else rv3_SSOC_IN_Mid(i,j) = log(rand+l )/(3.9*SSOC_d_IN_Mid); transform of the second exponential term ofSSOC pdf end end end if ul > wl CE IN Mid+ w2 CE IN Mid+ w3 CE IN Mid + w4 CE IN Mid - - - probability of3 charging event today, (or CE= 4), generate CST4, D4, SSOC4 276 % mverse % iful > USC Viterbi School of Engineering while rv4_CST_IN_Mid(i,j) <= 0 II rv4_CST_IN_Mid(i,j) > 23.5 if u2 <= wl CST IN Mid - - - PhD Thesis rv4_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; end else ifu2<= wl CST IN Mid+ w2 CST IN Mid - - - - - - rv4_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; else rv4_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end end while rv4_D_IN_Mid(i,j) <= 0 if u3 <= wl D IN Mid end rv4_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; else rv4_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end while rv4_SSOC_IN_Mid(i,j) <= 0 II rv4_SSOC_IN_Mid(i,j) > 100 if u4 <= wl SSOC IN Mid - - - rv4_SSOC_IN_Mid(i,j) = exprnd(mul_SSOC_IN_Mid); else rv4_SSOC_IN_Mid(i,j) = log(rand+l)/(3.9*SSOC_d_IN_Mid); transform of the second exponential term ofSSOC pdf % mverse end end end end end % determine if there is any overlap between charging event 2 and charging % event 1. If there is, re-generate CST2, D2 and re-determine for j=l:n2 for i=l:nl *p2 if rv2 _CST_ IN_ Mid(i,j)>O while rv2_CST_IN_Mid(i,j) + rv2_D_IN_Mid(i,j) > rvl_CST_IN_Mid(i,j) && rv2_CST_IN_Mid(ij) < rvl_CST_IN_Mid(i,j) + rvl_D_IN_Mid(i,j) u22 = rand; u33 = rand; if u22 <= w 1 CST IN Mid rv2_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; while rv2_CST_IN_Mid(i,j)<=O II rv2_CST_IN_Mid(i,j)>23.5 rv2_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; end else ifu22<= wl CST IN Mid+ w2 CST IN Mid else end end - - - - - - rv2_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; while rv2_CST_IN_Mid(i,j)<=O II rv2_CST_IN_Mid(ij)>23.5 rv2_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; end rv2_CST_IN_Mid(i,j ) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; while rv2_CST_IN_Mid(i,j)<=O II rv2_CST_IN_Mid(i,j)>23.5 rv2_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end if u33 <= wl D IN Mid 277 end end end end USC Viterbi School of Engineering rv2_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; while rv2 _ D _IN_ Mid(i,j) <= 0 PhD Thesis rv2_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; end else rv2_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; while rv2 _ D _IN_ Mid(i,j) <= 0 end rv2_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end % determine ifthere is any overlap between charging event 3 and charging % event 1, 2. If there is, re-generate CST3, D3 and re-determine for j=l:n2 for i=l:nl *p2 if rv3 _CST _IN _Mid(i,j)>O while (rv3_CST_IN_Mid(i,j) + rv3_D_IN_Mid(i,j) > rvl_CST_IN_Mid(i,j) && rv3_CST_IN_Mid(i,j) < rvl_CST_IN_Mid(i,j) + rvl_D_IN_Mid(i,j)) II (rv3_CST_IN_Mid(i,j) + rv3_D_IN_Mid(i,j) > rv2_CST_IN_Mid(i,j) && rv3_CST_IN_Mid(i,j) < rv2_CST_IN_Mid(i,j) + rv2_D _IN_Mid(i,j)) end end end u22 = rand; u33 = rand; if u22 <= w 1 CST IN Mid rv3_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; while rv3_CST_IN_Mid(i,j)<=O II rv3_CST_IN_Mid(i,j)>23.5 rv3_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; end else ifu22<= wl CST IN Mid+ w2 CST IN Mid else end end - - - - - - rv3_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; while rv3_CST_IN_Mid(i,j)<=O II rv3_CST_IN_Mid(i,j )>23.5 rv3_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; end rv3_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; while rv3_CST_IN_Mid(i,j)<=O II rv3_CST_IN_Mid(i,j )>23.5 rv3_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end if u33 <= wl D IN Mid - - - rv3_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; while rv3 _ D _IN_ Mid(i,j) <= 0 rv3_D_IN_Mid(i,j ) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; end else rv3_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; while rv3_D_IN_Mid(i,j) <= 0 end rv3_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end 278 USC Viterbi School of Engineering PhD Thesis end % determine ifthere is any overlap between charging event 4 and charging % event 1, 2 and 3. If there is, re-generate CST4, D4 and re-determine for j=l:n2 for i=l:nl *p2 if rv4 _CST_ IN_ Mid(i,j)>O while (rv4_CST_IN_Mid(i,j) + rv4_D_IN_Mid(i,j) > rvl_CST_IN_Mid(i,j) && rv4_CST_IN_Mid(ij) < rvl_CST_IN_Mid(i,j) + rvl_D_IN_Mid(i,j)) II (rv4_CST_IN_Mid(i,j) + rv4_D_IN_Mid(i,j) > rv2_CST_IN_Mid(i,j) && rv4_CST_IN_Mid(i,j) < rv2_CST_IN_Mid(i,j) + rv2_D_IN_Mid(i,j)) II (rv4_CST_IN_Mid(i,j) + rv4_D_IN_Mid(i,j) > rv3_CST_IN_Mid(i,j) && rv4_CST_IN_Mid(i,j) < rv3_CST_IN_Mid(i,j) + rv3_D_IN_Mid(i,j)) end end end end u22 =rand; u33 =rand; if u22 <= wl CST IN Mid - - - rv4_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; while rv4_CST_IN_Mid(i,j)<=O II rv4_CST_IN_Mid(i,j)>23.5 rv4_CST_IN_Mid(i,j) = stdl_CST_IN_Mid*randn + mul_CST_IN_Mid; end else ifu22<= wl CST IN Mid+ w2 CST IN Mid else end end - - - - - - rv4_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; while rv4_CST_IN_Mid(i,j)<=O II rv4_CST_IN_Mid(i,j)>23.5 rv4_CST_IN_Mid(i,j) = std2_CST_IN_Mid*randn + mu2_CST_IN_Mid; end rv4_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; while rv4_CST_IN_Mid(i,j)<=O II rv4_CST_IN_Mid(i,j )>23.5 rv4_CST_IN_Mid(i,j) = std3_CST_IN_Mid*randn + mu3_CST_IN_Mid; end if u33 <= wl D IN Mid - - - rv4_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; while rv4 _ D _IN_ Mid(i,j) <= 0 rv4_D_IN_Mid(i,j) = stdl_D_IN_Mid*randn + mul_D_IN_Mid; end else rv4_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; while rv4 _ D _IN_ Mid(i,j) <= 0 end rv4_D_IN_Mid(i,j) = std2_D_IN_Mid*randn + mu2_D_IN_Mid; end Dl_actual_IN_Mid = ones(round(nl *p2),n2)*-l; EVl_demand_IN_Mid = zeros(round(nl *p2),n2); D2_actual_IN_Mid = ones(round(nl *p2),n2)*-l; EV2_demand_IN_Mid = zeros(round(nl *p2),n2); D3_actual_IN_Mid = ones(round(nl *p2),n2)*-l ; EV3_demand_IN_Mid = zeros(round(nl *p2),n2); D4_actual_IN_Mid = ones(round(nl *p2),n2)*-l; EV4_demand_IN_Mid = zeros(round(nl *p2),n2); 279 USC Viterbi School of Engineering PhD Thesis for j=l:n2 for i=l:nl *p2 ifrvl_D_IN_Mid(i,j)>O && rvl_SSOC_IN_Mid(i,j)<lOO EVl_demand_IN_Mid(i,j) min(min(rvl_D_IN_Mid(i,j)*L2_P _IN_Mid, (100- rvl_SSOC_IN_Mid(i,j))/lOO*bat_IN_Mid), bat_IN_Mid); Dl_actual_IN_Mid(i,j) = EVl_demand_IN_Mid(i,j)/L2_P _IN_Mid; end if rv2_D_IN_Mid(i,j )>O && rv2_SSOC_IN_Mid(i,j)<l00 EV2_demand_IN_Mid(i,j) min(min(rv2_D_IN_Mid(i,j)*L2_P _IN_Mid, (100- rv2_SSOC_IN_Mid(i,j))/100*bat_IN_Mid), bat_IN_Mid); D2_actual_IN_Mid(i,j) = EV2_demand_IN_Mid(i,j)/L2_P _IN_Mid; end if rv3_D_IN_Mid(i,j )>O && rv3_SSOC_IN_Mid(i,j)<l00 EV3_demand_IN_Mid(i,j) min(min(rv3_D_IN_Mid(i,j)*L2_P _IN_Mid, (100- rv3_SSOC_IN_Mid(i,j))/100*bat_IN_Mid), bat_IN_Mid); D3_actual_IN_Mid(i,j) = EV3 _demand_IN_Mid(i,j)/L2_P _IN_Mid; end if rv4_D_IN_Mid(i,j)>O && rv4_SSOC_IN_Mid(i,j)<l00 EV4_demand_IN_Mid(i,j) min(min(rv4_D_IN_Mid(i,j)*L2_P _IN_Mid, (100- rv4_SSOC_IN_Mid(i,j))/100*bat_IN_Mid), bat_IN_Mid); D4_actual_IN_Mid(i,j) = EV4_demand_IN_Mid(i,j)/L2_P _IN_Mid; end end end rv_CST_IN_Mid = [rvl_CST_IN_Mid; rv2_CST_IN_Mid; rv3_CST_IN_Mid; rv4_CST_IN_Mid]; rv_D_IN_Mid = [rvl_D_IN_Mid; rv2_D_IN_Mid; rv3_D_IN_Mid; rv4_D_IN_Mid]; rv_SSOC_IN_Mid = [rvl_SSOC_IN_Mid; rv2_SSOC_IN_Mid; rv3_SSOC_IN_Mid; rv4_SSOC_IN_Mid]; D_actual_IN_Mid = [Dl_actual_IN_Mid; D2_actual_IN_Mid; D3_actual_IN_Mid; D4_actual_IN_Mid]; L2_P _IN_SUV = 6.6; % Level 2 charging power = 6.6 kW bat_IN _SUV= 42; % battery size of Electric SUV = 42 kWh rvl_CST_IN_SUV = ones(round(nl *p3),n2)*-l ; rvl_D_IN_SUV = ones(round(nl *p3),n2)*-l; rvl_SSOC_IN_SUV = ones(round(nl *p3),n2)*-l; rv2_CST_IN_SUV = ones(round(nl *p3),n2)*-l ; rv2_D_IN_SUV = ones(round(nl *p3),n2)*-l; rv2_SSOC_IN_SUV = ones(round(nl *p3),n2)*-l; rv3_CST_IN_SUV = ones(round(nl *p3),n2)*-l; rv3_D_IN_SUV = ones(round(nl *p3),n2)*-l; rv3_SSOC_IN_SUV = ones(round(nl *p3),n2)*-l; wl_CE_IN_SUV = 0.7218; w2_CE_IN_SUV = 0.1828; w3_CE_IN_SUV = 0.0799; w4_CE_IN_SUV = 0.0155; CST al IN SUV= - - - CST bl IN SUV= - - - CST cl IN SUV = - - - CST a2 IN SUV = - - - CST b2 IN SUV = - - - CST c2 IN SUV= - - - CST a3 IN SUV= 0.071 02;% (0.06226, 0.07977) 7.831;% (7.772, 7.89) 0.6206;% (0.5258, 0.7154) 0.061;% (0.04919, 0.0728) 12.52;% (12.47, 12.57) 0.3083;% (0.2386, 0.378) 0.01769;% (0.01406, 0.02133) 280 USC Viterbi School of Engineering CST_b3_IN_SUV = 12.24;% (11.47, 13.01) CST_c3_IN_SUV = 4.55;% (3.413, 5.687) mul_CST_IN_SUV= CST_bl_IN_SUV; % meanl *p3_CST =bl in Gaussian pdf mu2 _CST _IN_ SUV = CST_ b2 _IN_ SUV; mu3_CST_IN_SUV = CST_b3_IN_SUV; PhD Thesis stdl_CST_IN_SUV = CST_cl_IN_SUV/sqrt(2); % standard deviation 1 CST = cl/sqrt(2) in Gaussian pdf std2 _CST _IN_ SUV= CST_ c2 _IN_ SUV/sqrt(2); std3_CST_IN_SUV = CST_c3_IN_SUV/sqrt(2); coefl_CST_IN_SUV = stdl_CST_IN_SUV*sqrt(2*pi)*CST_al_IN_SUV; % coefficientl of CST = std*sqrt(2*pi)*al coef2 _CST _IN_ SUV = std2 _CST _IN_ SUV*sqrt(2*pi)*CST _ a2 _IN_ SUV; coef3 _CST _IN_ SUV= std3 _CST _IN_ SUV*sqrt(2*pi)*CST _ a3 _IN_ SUV; wl_CST IN SUV coefl_CST_IN_SUV/(coefl_CST_IN_SUV + coef2 CST IN SUV + coef3_CST_IN_SUV); % weightl of CST= coefl/sum(coefl+coef2+coef3); w2_CST_IN_SUV coef2_CST_IN_SUV/(coefl_CST_IN_SUV + coef2 CST IN SUV + coef3 _CST _IN_ SUV); w3_CST_IN_SUV coef3_CST_IN_SUV/(coefl_CST_IN_SUV + coef2 CST IN SUV + coef3 _CST _IN_ SUV); D_al_IN_SUV = 0.1805;% (0.1346, 0.2263) D_bl_IN_SUV = 1.378;% (0.9962, 1.76) D_cl_IN_SUV= 1.693;% (1.086,2.3) mul_D_IN_SUV = D_bl_IN_SUV; o/omean_D =bl in Gaussian pdf stdl_D_IN_SUV = D_cl_IN_SUV/sqrt(2); % standard deviation 1 D = cl/sqrt(2) in Gaussian pdf coefl_D_IN_SUV = stdl_D_IN_SUV*sqrt(2*pi)*D_al_IN_SUV; % coefficientl ofD = std*sqrt(2*pi)*al wl_D_IN_SUV = coefl_D_IN_SUV/coefl_D_IN_SUV; % weightl ofD = coefl/sum(coefl:coef3); SSOC_al_IN_SUV = 0.02677;% (0.02003, 0.03351) SSOC_bl_IN_SUV = 73.14;% (72.51, 73.76) SSOC_cl_IN_SUV = 3.264;% (2.236, 4.291) SSOC_a2_IN_SUV = 0.02164;% (0.01803, 0.02525) SSOC_b2_IN_SUV = 69.85;% (67.73, 71.96) SSOC_c2_IN_SUV = 22.34;% (18.85, 25.83) mul_SSOC_IN_SUV = SSOC_bl_IN_SUV; % meanl*p3_SSOC =bl mu2 _ SSOC _IN_ SUV= SSOC _ b2 _IN_ SUV; stdl_SSOC_IN_SUV = SSOC_cl_IN_SUV/sqrt(2); % standard deviation 1 SSOC = cl/sqrt(2) in Gaussian pdf std2_SSOC_IN_SUV = SSOC_c2_IN_SUV/sqrt(2); coefl_SSOC_IN_SUV = stdl_SSOC_IN_SUV*sqrt(2*pi)*SSOC_al_IN_SUV; % coefficientl ofSSOC = std*sqrt(2*pi)*al coef2 _ SSOC _IN _SUV= std2 _ SSOC _IN_ SUV*sqrt(2*pi)*SSOC _ a2 _IN_ SUV; wl_SSOC _IN _SUV= coefl_ SSOC _IN_ SUV/(abs(coefl_SSOC _IN_ SUV) + abs( coef2_ SSOC _IN _SUV)); w2_SSOC_IN_SUV = 1 - wl_SSOC_IN_SUV; for j=l:n2 for i=l:nl *p3 ul = rand; % ul =random variable to check how many times the EV is charging today u2 = rand; % u2=random variable follows CST pdf u3 = rand; % u3=random variable follows D pdf u4 = rand; % u4=random variable follows SSOC pdf iful > wl_ CE_IN_SUV % iful > probability ofO charging event today, (or CE >= 1), generate CSTl, Dl, SSOCl while rvl_CST_IN_SUV(i,j) <= 0 II rvl_CST_IN_SUV(i,j) > 23.5 if u2 <= wl CST IN SUV - - - rvl_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; else if u2<= wl CST IN SUV+ w2 CST IN SUV - - - - - - rvl_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; else rvl_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; 281 end USC Viterbi end end end while rvl_D_IN_SUV(i,j) <= 0 if u3 <= wl D IN SUV - - - School of Engineering PhD Thesis rvl_D_IN_SUV(i,j) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; end end while rvl_SSOC_IN_SUV(i,j) <= 0 II rvl_SSOC_IN_SUV(i,j) > 100 if u4 <= wl SSOC IN SUV end - - - rvl_SSOC_IN_SUV(i,j) = stdl_SSOC_IN_SUV*randn + mul_SSOC_IN_SUV; else rvl_SSOC_IN_SUV(i,j) = std2_SSOC_IN_SUV*randn + mu2_SSOC_IN_SUV; end if ul > wl_CE_IN_SUV + w2_CE_IN_SUV % iful > probability of 1 charging event today, (or CE >= 2), generate CST2, D2, SSOC2 end while rv2 _CST_ IN_ SUV(i,j) <= 0 11 rv2 _CST_ IN_ SUV(i,j) > 23 .5 if u2 <= wl CST IN SUV end - - - rv2_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; else ifu2<= wl CST IN SUV + w2 CST IN SUV - - - - - - rv2_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; else rv2_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; end end while rv2_D_IN_SUV(i,j) <= 0 if u3 <= wl D IN SUV - - - rv2_D_IN_SUV(i,j) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; end end while rv2_SSOC_IN_SUV(i,j) <= 0 II rv2_SSOC_IN_SUV(ij) > 100 if u4 <= wl SSOC IN SUV end - - - rv2_SSOC_IN_SUV(ij) = stdl_SSOC_IN_SUV*randn + mul_SSOC_IN_SUV; else rv2_SSOC_IN_SUV(i,j) = std2_SSOC_IN_SUV*randn + mu2_SSOC_IN_SUV; end if ul > wl_CE_IN_SUV + w2_CE_IN_SUV + w3_CE_IN_SUV % if ul > probability of 2 charging event today, (or CE >= 3), generate CST3, D3, SSOC3 while rv3 _CST_ IN_ SUV(i,j) <= 0 11 rv3 _CST_ IN_ SUV(i,j) > 23 .5 if u2 <= wl CST IN SUV end - - - rv3_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; else ifu2<= wl CST IN SUV + w2 CST IN SUV - - - - - - rv3_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; else rv3_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; end end 282 end end end USC Viterbi while rv3_D_IN_SUV(i,j) <= 0 if u3 <= wl D IN SUV - - - School of Engineering PhD Thesis rv3_D_IN_SUV(i,j) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; end end while rv3_SSOC_IN_SUV(i,j) <= 0 II rv3_SSOC_IN_SUV(i,j) > 100 if u4 <= wl SSOC IN SUV end - - - rv3_SSOC_IN_SUV(i,j) = stdl_SSOC_IN_SUV*randn + mul_SSOC_IN_SUV; else rv3_SSOC_IN_SUV(i,j) = std2_SSOC_IN_SUV*randn + mu2_SSOC_IN_SUV; end % determine if there is any overlap between charging event 2 and charging % event 1. If there is, re-generate CST2, D2 and re-determine for j=l:n2 for i=l:nl *p3 if rv2 _CST_ IN_ SUV (i,j )>O while rv2_CST_IN_SUV(i,j) + rv2_D_IN_SUV(i,j) > rvl_CST_IN_SUV(i,j) && rv2_ CST _IN_ SUV(i,j) < rv l _CST_ IN_ SUV(i,j) + rv l _ D _IN_ SUV(i,j) u22 = rand; end end end end u33 = rand; if u22 <= wl CST IN SUV rv2_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; while rv2_CST_IN_SUV(i,j)<=O II rv2_CST_IN_SUV(i,j)>23.5 rv2_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; end else ifu22<= wl CST IN SUV + w2 CST IN SUV else end end - - - - - - rv2_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; while rv2_CST_IN_SUV(i,j)<=O II rv2_CST_IN_SUV(i,j )>23.5 rv2_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; end rv2_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; while rv2_CST_IN_SUV(ij)<=O II rv2_CST_IN_SUV(i,j )>23.5 rv2_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; end if u33 <= wl D IN SUV - - - rv2_D_IN_SUV(i,j ) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; while rv2_D_IN_SUV(i,j) <= 0 rv2_D_IN_SUV(i,j) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; end end % determine if there is any overlap between charging event 3 and charging 283 USC Viterbi School of Engineering PhD Thesis % event 1, 2. If there is, re-generate CST3, D3 and re-determine for j=l:n2 for i=l:nl *p3 if rv3 _CST_ IN_ SUV(i,j )>O while (rv3_CST_IN_SUV(i,j) + rv3_D_IN_SUV(i,j) > rvl_CST_IN_SUV(i,j) && rv3_CST_IN_SUV(i,j) < rvl_CST_IN_SUV(i,j) + rvl_D_IN_SUV(i,j)) II (rv3_CST_IN_SUV(i,j) + rv3_D_IN_SUV(i,j) > rv2_CST_IN_SUV(i,j) && rv3_CST_IN_SUV(i,j) < rv2_CST_IN_SUV(i,j) + rv2 _ D _IN_ SUV(i,j )) end end end end u22 = rand; u33 =rand; if u22 <= wl CST IN SUV rv3_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; while rv3_CST_IN_SUV(i,j)<=O II rv3_CST_IN_SUV(i,j)>23.5 rv3_CST_IN_SUV(i,j) = stdl_CST_IN_SUV*randn + mul_CST_IN_SUV; end else ifu22<= wl CST IN SUV + w2 CST IN SUV else end end - - - - - - rv3_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; while rv3_CST_IN_SUV(i,j)<=O II rv3_CST_IN_SUV(i,j )>23.5 rv3_CST_IN_SUV(i,j) = std2_CST_IN_SUV*randn + mu2_CST_IN_SUV; end rv3_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; while rv3_CST_IN_SUV(i,j)<=O II rv3_CST_IN_SUV(i,j )>23.5 rv3_CST_IN_SUV(i,j) = std3_CST_IN_SUV*randn + mu3_CST_IN_SUV; end if u33 <= wl D IN SUV - - - rv3_D_IN_SUV(i,j) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; while rv3_D_IN_SUV(i,j) <= 0 rv3_D_IN_SUV(i,j) = stdl_D_IN_SUV*randn + mul_D_IN_SUV; end end Dl_actual_IN_SUV = ones(round(nl *p3),n2)*-l; EVl_demand_IN_SUV = zeros(round(nl *p3),n2); D2_actual_IN_SUV = ones(round(nl *p3),n2)*-l; EV2_demand_IN_SUV = zeros(round(nl *p3),n2); D3_actual_IN_SUV = ones(round(nl *p3),n2)*-l ; EV3 _demand_ IN_ SUV = zeros( round( nl *p3 ),n2 ); for j=l:n2 for i=l:nl *p3 if rvl _ D _IN _SUV(i,j)>O && rvl _SSOC _IN_ SUV(i,j)<lOO EVl_demand_IN_SUV(i,j) min(min(rvl_D _IN_SUV(i,j)*L2_P _IN_SUV, (100- rvl_SSOC_IN_SUV(i,j))/lOO*bat_IN _SUV), bat_IN_SUV); Dl_actual_IN_SUV(i,j) = EVl_demand_IN_SUV(i,j )/L2_P _IN_SUV; end if rv2_D_IN_SUV(i,j )>O && rv2_SSOC_IN_SUV(i,j)<l00 EV2_demand_IN_SUV(i,j) min(min(rv2_D _IN_SUV(i,j)*L2_P _IN_SUV, (100- rv2_SSOC_IN_SUV(i,j))/100*bat_IN _SUV), bat_IN_SUV); 284 USC Viterbi School of Engineering PhD Thesis D2 _ actual_IN _SUV(i,j) = EV2 _demand_ IN_ SUV(i,j )/L2 _p _IN_ SUV; end if rv3_D_IN_SUV(i,j)>O && rv3_SSOC_IN_SUV(i,j)<l00 EV3_demand_IN_SUV(i,j) min(min(rv3 _D _IN_SUV(i,j)*L2_P _IN_SUV, (100- rv3_SSOC_IN_SUV(i,j))/100*bat_IN_SUV), bat_IN_SUV); D3_actual_IN_SUV(i,j) = EV3_demand_IN_SUV(i,j)/L2_P _IN_SUV; end end end rv_CST_IN_SUV = [rvl_CST_IN_SUV; rv2_CST_IN_SUV; rv3_CST_IN_SUV]; rv_D_IN_SUV = [rvl_D_IN_SUV; rv2_D_IN_SUV; rv3_D_IN_SUV]; rv_SSOC_IN_SUV= [rvl_SSOC_IN_SUV; rv2_SSOC_IN_SUV; rv3_SSOC_IN_SUV]; D_actual_IN_SUV = [Dl_actual_IN_SUV; D2_actual_IN_SUV; D3_actual_IN_SUV]; rv_CST_IN = [rv_CST_IN_Comp; rv_CST_IN_Mid; rv_CST_IN_SUV]; rv_D_IN = [rv_D_IN_Comp; rv_D_IN_Mid; rv_D_IN_SUV]; rv_SSOC_IN = [rv_SSOC_IN_Comp; rv_SSOC_IN_Mid; rv_SSOC_IN_SUV]; D_actual_IN = [D_actual_IN_Comp; D_actual_IN_Mid; D_actual_IN_SUV]; rv_D_IN_freq = zeros(l2,n2); rv _D _IN_Hr = zeros(l2,l); for i=0:0.5:5.5 rv _D_IN_Hr(i*2+ l)=i; end for j=l:n2 for i=l:length(rv _ D _IN) for k=1:12 if rv _D _IN(i,j )>rv _D _IN_Hr(k) && rv _D _IN(i,j)<=rv _D _IN_Hr(k)+0.5 rv _D _IN_freq(k,j) = rv _D _IN_freq(k,j )+ l ; end end end end D_actual_IN_freq = zeros(12,n2); for j=l:n2 for i=l:length(D_actual_IN) for k=1:12 if D _ actual_IN(i,j )>rv _ D _IN_ Hr(k) && D _actual_ IN(i,j )<=rv _ D _IN _Hr(k)+0.5 D_actual_IN_freq(k,j) = D_actual_IN_freq(k,j)+l; end end end end figure subplot(2,2, 1) hist(rv _CST_ IN,48); set(gca,'xtick',0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start Time Variables') 285 USC Viterbi subplot(2,2,2) % hist(rv_D_IN); % set(gca,'xtick',0:7); % xlim([O 7]); bar(rv _ D _IN_ Hr,rv _ D _IN_ freq); xlabel('Plug-in Duration'); ylabel('Frequency'); School of Engineering title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) % hist(D_actual_IN); % set(gca,'xtick',0:6); % xlim([O.l 6]); bar(rv _ D _IN_ Hr,D _actual_ IN_ freq); xlabel('Charging Duration' ); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv _SSOC _IN, 100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') figure subplot(2,2, 1) hist(rv _CST_ IN_ Comp,48); set(gca, 'xtick', 0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(2,2,2) hist(rv _ D _IN_ Comp); set(gca,'xtick',0:6); xlim([O 5]); xlabel('Plug-in Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) hist(D _actual_IN_ Comp); set(gca,'xtick',0: 5); xlim( [O .1 4 ]); xlabel('Charging Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv _SSOC_IN_ Comp,100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') figure subplot(2,2, 1) hist(rv _CST_ IN_ Mid,48); 286 PhD Thesis set(gca,'xtick',0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); USC Viterbi School of Engineering title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(2,2,2) hist(rv _ D _IN_ Mid); set(gca,'xtick',0:6); xlim([O 4]); xlabel('Plug-in Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) hist(D_actual_IN_Mid); set(gca,'xtick',0: 5); xlim([O.l 3]); xlabel('Charging Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv _SSOC_IN_Mid,100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') figure subplot(2,2, 1) hist(rv _CST _IN _SUV,48); set(gca,'xtick',0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(2,2,2) hist(rv _ D _IN_ SUV); set(gca,'xtick',0:7); xlim([O 7]); xlabel('Plug-in Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) hist(D _actual_ IN_ SUV); set(gca,'xtick',0:6); xlim([O.l 6]); xlabel('Charging Duration' ); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv _ SSOC _IN_ SUV, 100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') 287 PhD Thesis USC Viterbi School of Engineering PhD Thesis % Convert EV Demand Forecast into 24-hr Format and Plot, then compare with Actual Weekday EV Demand (05/20115-02/04/17) EV _Demand_24hr_IN_Comp = zeros(n2,24); EVl_Demand_72hr_IN_Comp = zeros(n2,72); EV2_Demand_72hr_IN_Comp = zeros(n2,72); EV3_Demand_72hr_IN_Comp = zeros(n2,72); EV4_Demand_72hr_IN_Comp = zeros(n2,72); EV_ Demand_ 24 hr_ avg_ IN_ Comp = zeros( 1,24 ); EV _Demand_24hr_IN = zeros(n2,24); EV _Demand_24hr_avg_IN = zeros(l,24); for k=l :n2 for j=l:nl *pl for i=l :72 % Charging Event = 1 if floor(rvl _CST_IN_Comp(j,k)) floor(rvl_CST_IN_Comp(j,k) + Dl_actual_IN_Comp(j,k)) && floor(rvl_CST_IN_Comp(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr= i-1 EVl_Demand_72hr_IN_Comp(k,i) = EVl_Demand_72hr_IN_Comp(k,i) + EV!_ demand _IN_ Comp(j,k); % assign all EV demand to this hour slot else floor(rvl _CST_ IN_ Comp(j,k)+Dl _ actual_IN _ Comp(j,k))>floor(rvl _CST _IN_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rvl_CST_IN_Comp(j,k)) % if CS hr = (i-1 )th hour if 1- EVl_Demand_72hr_IN_Comp(k,i) = EVl_Demand_72hr_IN_Comp(k,i) + ( ceil(rvl _CST _IN_ Comp(j,k))- rvl _CST _IN_ Comp(j,k))*EVl _demand_ IN_ Comp(j,k)/Dl _actual _IN_ Comp(j,k); % EV demand at this hour= ( ceil(CST) - CST)*Demand/Duration end if floor(1vl _CST_ IN_ Comp(j,k)+Dl _ actual_IN _ Comp(j,k)) % if CEhr = (i-1 )th hour 1- EVl_Demand_72hr_IN_Comp(k,i) = EVl_Demand_72hr_IN_Comp(k,i) + (rvl _CST _IN_ Comp(j,k)+Dl_actual_IN _ Comp(j,k)- floor(rvl_ CST _IN_ Comp(j,k)+Dl_ actual_IN _ Comp(j,k)))*EVl_ demand _IN_ Comp(j,k)/Dl_ actual_IN _ C omp(j,k); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rvl _CST_IN_Comp(j,k)) && i-1 < floor(rvl_CST_IN_Comp(j,k)+Dl_actual_IN_Comp(j,k)) % ifCShr < (i-l)th hour < CEhr EVl_Demand_72hr_IN_Comp(k,i) = EVl_Demand_72hr_IN_Comp(k,i) + EVl_demand_IN_Comp(j,k)IDl_actual_IN_Comp(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 2 if floor(rv2_CST_IN_Comp(j,k)) floor(rv2_CST_IN_Comp(j ,k) + D2_actual_IN_Comp(j,k)) && floor(rv2_CST_IN_Comp(j,k)) == i-1 % if Charging Start Hr= Charging End Hr && Charging Start Hr= i-1 EV2 _Demand_ 72hr _IN_ Comp(k,i) = EV2 _Demand_ 72hr _IN_ Comp(k,i) + EV2_demand_IN_Comp(j,k); % assign all EV demand to this hour slot else if floor(rv2 _CST_ IN_ Comp(j,k)+D2 _actual_ IN_ Comp(j,k))>floor(rv2 _CST _IN_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv2_CST_IN_Comp(j,k)) 1- 288 USC Viterbi School of Engineering PhD Thesis % ifCShr = (i-l)th hour EV2_Demand_72hr_IN_Comp(k,i) = EV2_Demand_72hr_IN_Comp(k,i) + ( ceil(rv2_ CST _IN_ Comp(j,k))- rv2 _CST _IN_ Comp(j,k))*EV2 _demand_ IN_ Comp(j,k)/D2 _ actual_IN _ Comp(j,k); % EV demand at this hour= ( ceil(CST) - CST)*Demand/Duration end if floor(rv2 _CST_ IN_ Comp(j,k)+D2 _ actual_IN _ Comp(j,k)) % if CEhr = (i-1 )th hour 1- EV2_Demand_72hr_IN_Comp(k,i) = EV2_Demand_72hr_IN_Comp(k,i) + (rv2 _CST _IN_ Comp(j,k)+D2 _ actual_IN _ Comp(j,k)- floor(rv2 _CST_ IN_ Comp(j,k)+D2 _ actual_IN _ Comp(j,k)))*EV2 _demand_ IN_ Comp(j,k)/D2 _ actual_IN _ C omp(j,k); % EV demand at this hour= (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv2_CST_IN_Comp(j,k)) && i-1 < floor(rv2 _CST_ IN_ Comp(j,k)+D2 _ actual_IN _ Comp(j,k)) % if CShr < (i-1 )th hour < CEhr EV2_Demand_72hr_IN_Comp(k,i) = EV2_Demand_72hr_IN_Comp(k,i) + EV2 _demand _IN_ Comp(j,k)/D2 _ actual_IN _ Comp(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 3 if floor(rv3_CST_IN_Comp(j,k)) floor(rv3_CST_IN_Comp(j,k) + D3_actual_IN_Comp(j,k)) && floor(rv3_CST_IN_Comp(j ,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV3 _Demand_ 72hr _IN_ Comp(k,i) = EV3 _Demand_ 72hr _IN_ Comp(k,i) + EV3_demand_IN_Comp(j,k); % assign all EV demand to this hour slot else floor(rv3 _CST _IN_ Comp(j,k)+D3 _actual_IN _ Comp(j,k))>floor(rv3 _CST _IN_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor( rv3 _CST_ IN_ Comp(j ,k)) % ifCShr = (i-l)th hour if 1- EV3 _Demand_ 72hr _IN_ Comp(k,i) = E V3 _Demand_ 72hr_ IN_ Comp(k,i) + ( ceil(rv3 _CST _IN_ Comp(j,k))- rv3 _CST _IN_ Comp(j,k))*EV3 _demand _IN_ Comp(j,k)/D3 _ actual_IN _ Comp(j,k); % EV demand at this hour = ( ceil(CST) - CST)*Demand/Duration end if floor(rv3 _CST_ IN_ Comp(j,k)+D3 _ actual_IN _ Comp(j,k)) % if CEhr = (i-1 )th hour 1- EV3_Demand_72hr_IN_Comp(k,i) = EV3_Demand_72hr_IN_Comp(k,i) + (rv3 _CST _IN_ Comp(j,k)+D3 _ actual_IN _ Comp(j,k)- floor(rv3 _CST_ IN_ Comp(j,k)+D3 _ actual_IN _ Comp(j,k)))*EV3 _demand_ IN_ Comp(j,k)/D3 _ actual_IN _ C omp(j,k); % EV demand at this hour= (CST+D-CEhr)*Demand/Duration end if i-1 > floor(rv3_CST_IN_Comp(j,k)) && i-1 < floor(rv3 _CST_ IN_ Comp(j,k)+D3 _ actual_IN _ Comp(j,k)) % if CShr < (i-1 )th hour < CEhr EV3_Demand_72hr_IN_Comp(k,i) = EV3_Demand_72hr_IN_Comp(k,i) + EV3 _demand _IN_ Comp(j,k)/D3 _ actual_IN _ Comp(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 4 if floor(rv4 _CST_ IN_ Comp(j,k)) floor(rv4 _CST_ IN_ Comp(j,k) + D4_actual_IN_Comp(j,k)) && floor(rv4_CST_IN_Comp(j,k)) == i-1 % if Charging Start Hr= Charging 289 USC Viterbi School of Engineering PhD Thesis End Hr && Charging Start Hr= i-1 EV4_Demand_72hr_IN_Comp(k,i) = EV4_Demand_72hr_IN_Comp(k,i) + EV4_demand_IN_Comp(j,k); % assign all EV demand to this hour slot else floor(rv4 _CST_ IN_ Comp(j,k)+D4 _ actual_IN _ Comp(j,k))>floor(rv4_ CST _IN_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor( rv4 _CST_ IN_ Comp(j ,k)) % ifCShr = (i-l)th hour if I- EV4_Demand_72hr_IN_Comp(k,i) = EV4_Demand_72hr_IN_Comp(k,i) + ( ceil(rv4 _CST _IN_ Comp(j,k))- rv4_ CST _IN_ Comp(j,k))*EV4_ demand_ IN_ Comp(j,k)/D4_ actual_IN _ Comp(j,k); % EV demand at this hour= ( ceil(CST) - CST)*Demand/Duration end if floor(rv4_ CST _IN_ Comp(j,k)+D4_actual_IN _ Comp(j,k)) % if CEhr = (i-1 )th hour I- EV4_Demand_72hr_IN_Comp(k,i) = EV4_Demand_72hr_IN_Comp(k,i) + (rv4 _CST _IN_ Comp(j,k)+D4_ actual_IN _ Comp(j,k)- floor(rv4 _CST _IN_ Comp(j,k)+D4 _actual_IN _ Comp(j,k)))*EV4_ demand_IN _ Comp(j,k)/D4_actual_IN_ C omp(j,k); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv4_CST_IN_Comp(j,k)) && i-1 < floor(rv4_CST_IN_Comp(j,k)+D4_actual_IN_Comp(j,k)) % ifCShr < (i-l)th hour < CEhr EV4_Demand_72hr_IN_Comp(k,i) = EV4_Demand_72hr_IN_Comp(k,i) + EV4_demand_IN_Comp(j,k)ID4_actual_IN_Comp(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end end end end for k=l:n2 for i=1:24 EV_ Demand_ 24hr _IN_ Comp(k,i) EVl_Demand _72hr _IN_ Comp(k,i+24) EV2_Demand_72hr_IN_Comp(k,i) EVl_Demand_72hr_IN_Comp(k,i) + + EVl_Demand_72hr_IN_Comp(k,i+48) + + EV2 _Demand_ 72hr _IN_ Comp(k,i+24) + + EV3 _Demand_ 72hr _IN_ Comp(k,i) + + EV3_Demand_72hr_IN_Comp(k,i+48) + EV2 _Demand_ 72hr _IN_ Comp(k,i+48) EV3 _Demand_ 72hr _IN_ Comp(k,i+24) EV4_Demand_72hr_IN_Comp(k,i) EV4 _Demand _72hr_IN _ Comp(k,i+48); hour format + EV4_Demand_72hr_IN_Comp(k,i+24) + % convert charging start at late night and ends tomorrow into 24 end end for i=1:24 EV _Demand_24hr_avg_IN_Comp(l,i) = mean(EV _Demand_24hr_IN_Comp(:,i)); end EV_ Demand_ 24hr _IN_ Mid = zeros(n2,24 ); EVl_Demand_72hr_IN_Mid = zeros(n2,72); EV2_Demand_72hr_IN_Mid = zeros(n2,72); EV3_Demand_72hr_IN_Mid = zeros(n2,72); EV4_Demand_72hr_IN_Mid = zeros(n2,72); EV _Demand_24hr_avg_IN_Mid = zeros(l,24); 290 for k=l :n2 for j=l:nl *p2 for i=l :72 USC Viterbi School of Engineering % Charging Event = 1 PhD Thesis if floor(rvl_CST_IN_Mid(j,k)) == floor(rvl_CST_IN_Mid(j,k) + Dl_actual_IN_Mid(j,k)) && floor(1vl_CST_IN_Mid(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr= i-1 EVl _Demand_ 72hr _IN_ Mid(k,i) EVl _Demand_ 72hr _IN_ Mid(k,i) + EVl_demand_IN_Mid(j,k); % assign all EV demand to this hour slot else floor(rvl _CST_ IN_ Mid(j,k)+Dl _actual_ IN_ Mid(j,k))>floor(rvl _CST _IN_ Mid(j,k)) % else if CE hr (charging ending hr) > CS hr (charging start hr) if floor(rvl_CST_IN_Mid(j,k)) if I- % if CS hr= (i-1 )th hour EVl_Demand_72hr_IN_Mid(k,i) = EVl_Demand_72hr_IN_Mid(k,i) + ( ceil(rvl _CST _IN_ Mid(j,k))-rvl _CST _IN_ Mid(j,k))*EVl _demand_ IN_ Mid(j,k)/Dl _ actual_IN _ Mid(j,k); % EV demand at this hour = ( ceil(CST) - CST)*Demand/Duration end if floor(rvl_ CST _IN _Mid(j,k)+Dl_actual_IN _Mid(j,k)) % if CEhr = (i-1 )th hour I- EVl_Demand_72hr_IN_Mid(k,i) = EVl_Demand_72hr_IN_Mid(k,i) + (rvl _CST _IN_ Mid(j,k)+Dl _actual_ IN_ Mid(j,k)- floor(rvl _ CST _IN_Mid(j,k)+Dl_actual_IN _Mid(j,k)))*EVl _ demand_IN _Mid(j,k)/Dl _actual_IN_Mid(j,k ); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rvl_CST_IN_Mid(j,k)) && i-1 < floor(rvl _ CST_IN_Mid(j,k)+Dl_actual_IN_Mid(j,k)) % if CShr < (i-l)th hour < CEhr EVl_Demand_72hr_IN_Mid(k,i) = EVl_Demand_72hr_IN_Mid(k,i) + EVl_demand_IN_Mid(j,k)/Dl_actual_IN_Mid(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 2 if floor(rv2_CST_IN_Mid(j,k)) == floor(rv2_CST_IN_Mid(j,k) + D2_actual_IN_Mid(j,k)) && floor(rv2_CST_IN_Mid(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV2 _Demand_ 72hr _IN_ Mid(k,i) EV2 _Demand_ 72hr _IN_ Mid(k,i) + EV2_demand_IN_Mid(j,k); % assign all EV demand to this hour slot else floor(rv2 _CST _IN _Mid(j,k)+D2 _ actual_IN _Mid(j,k))>floor(rv2 _CST_ IN _Mid(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv2 _CST _IN_ Mid(j,k)) % ifCShr = (i-l)th hour if I- EV2 _Demand_ 72hr _IN_ Mid(k,i) = EV2 _Demand_ 72hr _IN_ Mid(k,i) + ( ceil(rv2 _CST _IN _Mid(j,k))-rv2 _CST _IN _Mid(j,k))*EV2 _demand_ IN _Mid(j,k)/D2 _ actual_IN _ Mid(j,k); % EV demand at this hour = ( ceil(CST) - CST)*Demand/Duration end if floor(rv2 _CST _IN_ Mid(j,k)+D2 _ actual_IN _ Mid(j,k)) % if CEhr = (i-1 )th hour I- EV2_Demand_72hr_IN_Mid(k,i) = EV2_Demand_72hr_IN_Mid(k,i) + (rv2 _CST _IN_ Mid(j ,k)+D2 _ actual_IN _ Mid(j,k)- floor(rv2 _CST_ IN_ Mid(j,k)+D2 _actual_ IN_ Mid(j,k)))*EV2 _demand _IN_ Mid(j,k)/D2 _actual_ IN_ Mid(j,k ); % EV demand at this hour= (CST +D-CEhr)*Demand/Duration end 291 USC Viterbi School of Engineering PhD Thesis if i-1 > floor(rv2_CST_IN_Mid(j,k)) && i-1 < floor(rv2_ CST_IN_Mid(j,k)+D2_actual_IN_Mid(j,k)) % if CShr < (i-l)th hour < CEhr EV2_Demand_72hr_IN_Mid(k,i) = EV2_Demand_72hr_IN_Mid(k,i) + EV2_demand_IN_Mid(j,k)/D2_actual_IN_Mid(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event = 3 if floor(rv3_CST_IN_Mid(j,k)) == floor(rv3_CST_IN_Mid(j,k) + D3_actual_IN_Mid(j,k)) && floor(1v3_CST_IN_Mid(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr= i-1 EV3 _Demand_ 72hr _IN_ Mid(k,i) EV3 _Demand_ 72hr _IN_ Mid(k,i) + EV3_demand_IN_Mid(j,k); % assign all EV demand to this hour slot else floor(rv3 _CST_ IN_ Mid(j,k)+D3 _actual_ IN_ Mid(j,k))>floor(rv3 _CST _IN_ Mid(j,k)) % else if CE hr (charging ending hr) > CS hr (charging start hr) if floor(rv3_CST_IN_Mid(j ,k)) if I- % if CS hr = (i-1 )th hour EV3_Demand_72hr_IN_Mid(k,i) = EV3_Demand_72hr_IN_Mid(k,i) + ( ceil(rv3 _CST _IN_ Mid(j,k))-rv3 _CST _IN_ Mid(j,k))*EV3 _demand_ IN_ Mid(j,k)/D3 _ actual_IN _ Mid(j,k); % EV demand at this hour= ( ceil(CST) - CST)*Demand/Duration end if floor(rv3 _CST _IN _Mid(j,k)+ D3 _actual_IN _Mid(j,k)) % if CEhr = (i-1 )th hour I- EV3_Demand_72hr_IN_Mid(k,i) = EV3_Demand_72hr_IN_Mid(k,i) + (rv3 _CST _IN_ Mid(j,k)+D3 _actual_ IN_ Mid(j,k)- floor(rv3 _CST_ IN_ Mid(j,k)+D3 _actual_ IN_ Mid(j,k)))*EV3 _demand _IN_ Mid(j,k)/D3 _actual_ IN_ Mid(j,k ); % EV demand at this hour= (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv3_CST_IN_Mid(j ,k)) && i-1 < floor(rv3 _CST_ IN_ Mid(j,k)+D3 _actual_ IN_ Mid(j,k)) % if CS hr < (i-1 )th hour < CEhr EV3_Demand_72hr_IN_Mid(k,i) = EV3_Demand_72hr_IN_Mid(k,i) + EV3_demand_IN_Mid(j,k)/D3_actual_IN_Mid(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 4 if floor(rv4_CST_IN_Mid(j ,k)) == floor(rv4_CST_IN_Mid(j,k) + D4_actual_IN_Mid(j,k)) && floor(rv4_CST_IN_Mid(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV4_Demand_72hr_IN_Mid(k,i) EV4_Demand_72hr_IN_Mid(k,i) + EV4_demand_IN_Mid(j,k); % assign all EV demand to this hour slot else floor(rv4 _CST_ IN_ Mid(j,k)+D4 _actual_ IN_ Mid(j,k))>floor(rv4_ CST _IN _Mid(j,k)) % else if CEhr (charging ending hr) > CS hr (charging start hr) if floor(rv4_CST_IN_Mid(j,k)) if I- % if CS hr= (i-1 )th hour EV4_Demand_72hr_IN_Mid(k,i) = EV4_Demand_72hr_IN_Mid(k,i) + ( ceil(rv4_ CST _IN _Mid(j,k))-rv4 _CST _IN _Mid(j,k))*EV4_demand_IN _Mid(j,k)/D4 _actual_IN _Mid(j,k); % EV demand at this hour = ( ceil(CST) - CST)*Demand/Duration end if floor(rv4 _CST _IN_ Mid(j,k)+ D4_ actual_IN _ Mid(j,k)) % if CEhr = (i-1 )th hour 292 I- USC Viterbi School of Engineering PhD Thesis EV4_Demand_72hr_IN_Mid(k,i) = EV4_Demand_72hr_IN_Mid(k,i) + (rv4 _CST _IN _Mid(j,k)+D4_actual_IN _Mid(j,k)- floor(rv4_ CST _IN _Mid(j,k)+D4 _actual_IN _Mid(j,k)))*EV4 _demand _IN _Mid(j,k)/D4_actual_IN _Mid(j,k ); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv4_CST_IN_Mid(j,k)) && i-1 < floor(rv4_ CST_IN_Mid(j,k)+D4_actual_IN _Mid(j,k)) % if CShr < (i-l)th hour < CEhr EV4_Demand_72hr_IN_Mid(k,i) = EV4_Demand_72hr_IN_Mid(k,i) + EV4_demand_IN_Mid(j,k)/D4_actual_IN_Mid(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end end end end for k=l:n2 for i=1:24 EVl_Demand_72hr_IN_Mid(k,i) + + EVl_Demand_72hr_IN_Mid(k,i+48) + + EV2_Demand_72hr_IN_Mid(k,i+24) + + EV3_Demand_72hr_IN_Mid(k,i) + EV _Demand_24hr_IN_Mid(k,i) EV1_Demand_72hr _IN_ Mid(k,i+24) EV2 _Demand_ 72hr _IN_ Mid(k,i) EV2_Demand_72hr_IN_Mid(k,i+48) EV3 _Demand_ 72hr _IN_ Mid(k,i+ 24) + EV3_Demand_72hr_IN_Mid(k,i+48) + EV 4 _Demand_ 72hr _IN_ Mid(k,i) EV4_Demand_72hr_IN_Mid(k,i+48); hour format + EV4_Demand_72hr_IN_Mid(k,i+24) + % convert charging start at late night and ends tomorrow into 24 end end for i=1:24 EV _Demand_24hr_avg_IN_Mid(l ,i) = mean(EV _Demand_24hr_IN_Mid(:,i)); end EV _Demand_24hr_IN_SUV = zeros(n2,24); EVl_Demand_72hr_IN_SUV = zeros(n2,72); EV2_Demand_72hr_IN_SUV = zeros(n2,72); EV3_Demand_72hr_IN_SUV = zeros(n2,72); EV _Demand_24hr_avg_IN_SUV = zeros(l ,24); for k=l:n2 for j=l:nl *p3 for i=l :72 % Charging Event= 1 if floor(rvl_CST_IN_SUV(j,k)) == floor(rvl_CST_IN_SUV(j,k) + Dl_actual_IN_SUV(j,k)) && floor(rvl_CST_IN_SUV(j,k)) == i-1 % if Charging Start Hr= Charging End Hr && Charging Start Hr = i-1 EVl_Demand_72hr_IN_SUV(k,i) EVl_Demand_72hr_IN_SUV(k,i) + EVl_demand_IN_SUV(j,k); % assign all EV demand to this hour slot else if floor(rvl _CST_ IN_ SUV(j,k)+Dl _ actual_IN _ SUV(j,k))>floor(rvl _CST _IN_ SUV(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rvl_CST_IN_SUV(j,k)) 1- % ifCShr = (i-l)th hour EVl_Demand_72hr_IN_SUV(k,i) = EVl_Demand_72hr_IN_SUV(k,i) + ( ceil(rvl _CST _IN _SUV(j,k))- 293 USC Viterbi School of Engineering PhD Thesis rvl_CST_IN_SUV(j,k))*EVl_demand_IN_SUV(j,k)/Dl_actual_IN_SUV(j,k); % EV demand at this hour = ( ceil(CST) - CST)*Demand/Duration end if floor(rvl_ CST _IN_ SUV(j,k)+Dl_ actual_IN _ SUV(j,k)) % if CEhr = (i-1 )th hour 1- EV1_Demand_72hr_IN_SUV(k,i) = EV1_Demand_72hr_IN_SUV(k,i) + (rvl _CST _IN_ SUV(j,k)+Dl _ actual_IN _ SUV(j,k)- floor(rvl_ CST_ IN_ SUV(j,k)+ Dl _actual _IN_ SUV(j,k)))*EVl _demand_ IN_ SUV(j,k)/Dl _ actual_IN _SUV (j,k); % EV demand at this hour = (CST+ D-CEhr)*Demand/Duration end if i-1 > floor(rvl _CST_IN_SUV(j,k)) && i-1 < floor(rvl_CST_IN_SUV(j,k)+Dl_actual_IN_SUV(j,k)) % ifCShr < (i-l)th hour < CEhr EV1_Demand_72hr_IN_SUV(k,i) = EV1_Demand_72hr_IN_SUV(k,i) + EVl_demand_IN_SUV(j,k)/Dl_actual_IN_SUV(j,k); % EV demand at this hour= Demand/Duration it means this whole hour will be fully occupied end end end % Charging Event = 2 iffloor(rv2_ CST _IN_SUV(j ,k)) == floor(rv2_ CST _IN_SUV(j,k) + D2_actual_IN_SUV(j,k)) && floor(rv2_CST_IN_SUV(j,k)) == i-1 % if Charging Start Hr= Charging End Hr && Charging Start Hr= i-1 EV2_Demand_72hr_IN_SUV(k,i) EV2_Demand_72hr_IN_SUV(k,i) + EV2_demand_IN_SUV(j,k); % assign all EV demand to this hour slot else floor(rv2 _CST_ IN_ SUV(j,k)+D2 _ actual_IN _ SUV(j,k))>floor(rv2 _CST _IN_ SUV(j,k)) % else if CEhr (charging ending hr) > CS hr (charging start hr) if floor(rv2 _CST _IN_ SUV(j,k)) % ifCShr = (i-l)th hour if 1- EV2_Demand_72hr_IN_SUV(k,i) = EV2_Demand_72hr_IN_SUV(k,i) + ( ceil(rv2 _CST _IN _SUV(j,k))- rv2 _CST _IN_ SUV(j,k))*EV2 _demand_ IN_ SUV(j,k)/D2 _actual_ IN_ SUV(j,k); % EV demand at this hour = (ceil(CST) - CST)*Demand/Duration end if floor(rv2 _CST _IN_ SUV(j,k)+D2_ actual_IN _ SUV(j,k)) % if CEhr = (i-1 )th hour 1- EV2_Demand_72hr_IN_SUV(k,i) = EV2_Demand_72hr_IN_SUV(k,i) + (rv2 _CST _IN_ SUV(j,k)+D2 _ actual_IN _ SUV(j,k)- floor(rv2 _CST_ IN_ SUV(j,k)+D2 _ actual_IN _ SUV(j,k)))*EV2 _demand _IN _SUV(j,k)/D2 _ actual_IN _SUV (j,k); % EV demand at this hour= (CST+ D-CEhr)*Demand/Duration end if i-1 > floor(rv2_CST_IN_SUV(j ,k)) && i-1 < floor(rv2_CST_IN_SUV(j,k)+D2_actual_IN_SUV(j,k)) % ifCShr < (i-l)th hour < CEhr EV2_Demand_72hr_IN_SUV(k,i) = EV2_Demand_72hr_IN_SUV(k,i) + EV2_demand_IN_SUV(j,k)/D2_actual_IN_SUV(j,k); % EV demand at this hour= Demand/Duration it means this whole hour will be fully occupied end end end % Charging Event= 3 iffloor(rv3 _CST _IN_SUV(j,k)) == floor(rv3 _CST _IN_SUV(j,k) + D3 _actual_IN_SUV(j,k)) && floor(rv3_CST_IN_SUV(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr= i-1 EV3_Demand_72hr_IN_SUV(k,i) EV3_Demand_72hr_IN_SUV(k,i) + EV3_demand_IN_SUV(j,k); % assign all EV demand to this hour slot 294 USC Viterbi School of Engineering else floor(rv3 _CST _IN_ SUV(j,k)+D3 _actual_IN _ SUV(j,k))>floor(rv3 _CST _IN_ SUV(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv3_CST_IN_SUV(j,k)) % ifCShr = (i-l)th hour PhD Thesis if 1- EV3_Demand_72hr_IN_SUV(k,i) = EV3_Demand_72hr_IN_SUV(k,i) + ( ceil(rv3 _CST _IN _SUV(j,k))- rv3 _CST _IN_ SUV(j,k))*EV3 _demand _IN _SUV(j,k)/D3 _actual_IN _ SUV(j,k); % EV demand at this hour = (ceil(CST) - CST)*Demand/Duration end if floor(rv3 _CST _IN_ SUV(j,k)+D3 _actual_ IN_ SUV(j,k)) % if CEhr = (i-1 )th hour 1- EV3_Demand_72hr_IN_SUV(k,i) = EV3_Demand_72hr_IN_SUV(k,i) + (rv3 _CST _IN _SUV(j,k)+D3 _actual_IN _ SUV(j,k)- floor(rv3 _CST_ IN_ SUV(j,k)+D3 _ actual_IN _ SUV(j,k)))*EV3 _demand_ IN_ SUV(j,k)/D3 _actual_ IN_ SUV (j,k); % EV demand at this hour= (CST+ D-CEhr)*Demand/Duration end if i-1 > floor(rv3_CST_IN_SUV(j,k)) && i-1 < floor(rv3_CST_IN_SUV(j,k)+D3_actual_IN_SUV(j,k)) % ifCShr < (i-l)th hour < CEhr EV3_Demand_72hr_IN_SUV(k,i) = EV3_Demand_72hr_IN_SUV(k,i) + EV3_demand_IN_SUV(j,k)/D3_actual_IN_SUV(j,k); % EV demand at this hour= Demand/Duration it means this whole hour will be fully occupied end end end end end end for k=l:n2 for i=1:24 EV _Demand_24hr_IN_SUV(k,i) EVl_Demand_72hr_IN_SUV(k,i) + EVl_Demand_72hr_IN_SUV(k,i+24) + EVl_Demand_72hr_IN_SUV(k,i+48) + EV2_Demand_72hr_IN_SUV(k,i) + EV2_Demand_72hr_IN_SUV(k,i+24) + EV2_Demand_72hr_IN_SUV(k,i+48) + EV3_Demand_72hr_IN_SUV(k,i) + EV3_Demand_72hr_IN_SUV(k,i+24) + EV3_Demand_72hr_IN_SUV(k,i+48); % convert charging start at late night and ends tomorrow into 24 hour format end end for i=1:24 EV _Demand_24hr_avg_IN_SUV(l,i) = mean(EV _Demand_24hr_IN_SUV(:,i)); end for k=l:n2 for i=1:24 EV _Demand_24hr_IN(k,i) = EV _Demand_24hr_IN_Comp(k,i) +EV _Demand_24hr_IN_Mid(k,i) +EV _Demand_24hr_IN_SUV(k,i); end end for i=1:24 EV _Demand_24hr_avg_IN(l,i) = mean(EV _Demand_24hr_IN(:,i)); end 295 USC Viterbi figure plot(O:l:23, EV_Demand_24hr_IN); xlabel('Time of Day'); ylabel('EV Charging Demand in kW'); School of Engineering PhD Thesis strl = sprintf ('Industrial EV Charging (Level 2) Demand Forecast Based on Monte Carlo Simulation (%d EVs: o/od %% Compact+ %d %% Midsize + %d %% SUV, %d times)',nl,pl *100,p2*100,p3*100,n2); strl 1 =textwrap( { strl },85); title( str 11 ); set(gca,'xtick', 0:23 ); figure % plot surf of EV_ Demand_ 24hr matrix hr= [O: 1 :23]'; MC_num = [l:l:n2]'; surf(hr, MC_ num, EV_ Demand_ 24hr _IN); color bar str3 = sprintf('Surf Plot oflndustrial EV Charging Demand Forecast Based on Monte Carlo Simulation (o/od EVs: o/od %% Compact + %d %% Midsize + %d %% SUV, %d times)',nl,pl *100,p2*100,p3*100,n2); str33=textwrap( {str3 },85); title( str33 ); xlabel('Time of Day'); ylabel('Monte Carlo Simulation Time'); zlabel('EV Charging Demand in kW'); hrq = 0:0.1 :23; MC_numq = l:l:n2; [hrq, MC_ numq] = meshgrid(hrq, MC_ numq); EV _Demand_24hrq = interp2(hr, MC_num, EV _Demand_24hr_IN, hrq, MC_numq, 'spline'); figure surf(hrq, MC_numq, EV _Demand_24hrq); color bar str4 = sprintf ('Surf Plot with Spline oflndustrial Compact EV Charging Demand Forecast Based on Monte Carlo Simulation (%d EVs: %d %% Compact + %d %% Midsize + %d %% SUV , %d times)',nl ,pl *100,p2*100,p3*100,n2); str44=textwrap( { str4} ,90); title( str44 ); xlabel('Time of Day'); ylabel('Monte Carlo Simulation Time'); zlabel('EV Charging Demand in kW'); 296 USC Viterbi School of Engineering E.2 Model for Commercial Area %% Commercial - Generate Random Number U, Ul, U2, U3 follows three pdf prompt_EV _num = 'Please enter the number ofEVs'; nl = input(prompt_EV _num); % nl =# of Total Commercial EVs to run Monte Carlo prompt_ COM_ Comp= 'Please enter the percentage of Commercial Compact EV>> '; p2 = input(prompt_ COM_ Comp); prompt_ COM_ SUV = 'Please enter the percentage of Commercial Electric SUV>> '; p3 = input(prompt_COM_SUV); prompt_MC_num = 'Please enter Monte Carlo Simulation times>> '; n2 = input(prompt_MC_num); L2_P _COM_Comp = 3.3; % Level 2 charging power= 3.3 kW bat_COM_Comp = 24; % battery size of Compact EV = 24 kWh rng('shuffle'); rvl_CST_COM_Comp = ones(round(nl *p2),n2)*-l; rvl_D_COM_Comp = ones(round(nl *p2),n2)*-l; rvl_SSOC_COM_Comp = ones(round(nl *p2),n2)*-l; rv2_CST_COM_Comp = ones(round(nl *p2),n2)*-l; rv2_D_COM_Comp = ones(round(nl *p2),n2)*-l; rv2_SSOC_COM_Comp = ones(round(nl *p2),n2)*-l; rv3_CST_COM_Comp = ones(round(nl *p2),n2)*-l; rv3_D_COM_Comp = ones(round(nl *p2),n2)*-l ; rv3_SSOC_COM_Comp = ones(round(nl *p2),n2)*-l ; rv4_CST_COM_Comp = ones(round(nl *p2),n2)*-l; rv4_D_COM_Comp = ones(round(nl *p2),n2)*-l; rv4_SSOC_COM_Comp = ones(round(nl *p2),n2)*-l; rv5_CST_COM_Comp = ones(round(nl *p2),n2)*-l; rv5_D_COM_Comp = ones(round(nl *p2),n2)*-l ; rv5_SSOC_COM_Comp = ones(round(nl *p2),n2)*-l; wl_CE_COM_Comp = 0.374; w2_CE_COM_Comp = 0.275; w3_CE_COM_Comp = 0.200; w4_CE_COM_Comp = 0.102; w5_CE_COM_Comp = 0.037; w6_CE_COM_Comp = 0.012; CST_al_COM_Comp = CST_bl_COM_Comp = CST_cl_COM_Comp = CST_a2_COM_Comp = CST_b2_COM_Comp = CST_c2_COM_Comp = 0.03829;% (0.03693, 0.03964) 11.84;% (11.64, 12.04) 2.701;% (2.464, 2.938) 0.02459;% (0.02085, 0.02834) 15.18;% (15.04, 15.32) 1.422;% (1.232, 1.612) mul_CST_COM_Comp = CST_bl_COM_Comp; % meanl_CST =bl in Gaussian pdf mu2_CST_COM_Comp = CST_b2_COM_Comp; PhD Thesis stdl_CST_COM_Comp = CST_cl_COM_Comp/sqrt(2); % standard deviation 1 CST= cl/sqrt(2) in Gaussian pdf std2_CST_COM_Comp = CST_c2_COM_Comp/sqrt(2); coefl_CST_COM_Comp = stdl_CST_COM_Comp*sqrt(2*pi)*CST_al_COM_Comp; % coefficient! of CST= std*sqrt(2*pi)*al coef2 _CST_ COM_ Comp= std2 _CST_ COM_ Comp*sqrt(2*pi)*CST _ a2 _COM_ Comp; wl_CST_COM_Comp = coefl_CST_COM_Comp/(coefl_CST_COM_Comp + coef2_CST_COM_Comp); % weightl of CST = coefl/sum(coefl +coef2+coef3); 297 USC Viterbi School of Engineering w2 _CST_ COM_ Comp= coef2 _CST_ COM_ Comp/( coefl _CST_ COM_ Comp+ coef2 _CST_ COM_ Comp); D_al_COM_Comp = 0.2939;% (0.2543, 0.3336) D_bl_COM_Comp = -0.1608;% (-0.6848, 0.3632) D_cl_COM_Comp = 1.732;% (1.272, 2.191) mul_D_COM_Comp = D_bl_COM_Comp; %mean_D = bl in Gaussian pdf PhD Thesis stdl_D_COM_Comp = D_cl_COM_Comp/sqrt(2); % standard deviation 1 D = cl/sqrt(2) in Gaussian pdf coefl_D_COM_Comp = stdl_D_COM_Comp*sqrt(2*pi)*D_al_COM_Comp; % coefficientl ofD = std*sqrt(2*pi)*al wl_D_COM_Comp = coefl_D_COM_Comp/coefl_D_COM_Comp; % weightl ofD = coefl/sum( coefl :coef3 ); SSOC_al_COM_Comp = 0.04375;% (0.01159, 0.07591) SSOC_bl_COM_Comp= 131.3;% (87.13, 175.4) SSOC_cl_COM_Comp = 56.22;% (33.87, 78.58) mul_SSOC_COM_Comp = SSOC_bl_COM_Comp; % meanl_SSOC = bl in Gaussian pdf stdl_SSOC_COM_Comp = SSOC_cl_COM_Comp/sqrt(2); % standard deviation 1 SSOC = cl/sqrt(2) in Gaussian pdf coefl_SSOC_COM_Comp = stdl_SSOC_COM_Comp*sqrt(2*pi)*SSOC_al_COM_Comp; % coefficient! of SSOC = std*sqrt(2*pi)*al wl_SSOC_COM_Comp = coefl_SSOC_COM_Comp/coefl_SSOC_COM_Comp; % weightl ofSSOC = coefl/sum( coefl +coef2+coef3 ); for j=l:n2 for i=l:nl *p2 ul = rand; % ul =random variable to check how many times the EV is charging today u2 = rand; % u2=random variable follows CST pdf u3 = rand; % u3=random variable follows D pdf u4 = rand; % u4=random variable follows SSOC pdf iful > wl_CE_COM_Comp % iful > probability ofO charging event today, (or CE >= 1), generate CSTl, Dl, SSOCl while rvl_CST_COM_Comp(i,j) <= 0 II 1vl_CST_COM_Comp(i,j) > 23.5 if u2 <= wl_CST_COM_Comp rvl_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; else rv l_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end while rvl_D_COM_Comp(i,j) <= 0 if u3 <= wl_D_COM_Comp rvl_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; end end while rvl_SSOC_COM_Comp(i,j) <= 0 II rvl_SSOC_COM_Comp(i,j) > 100 if u4 <= wl_SSOC_COM_Comp rvl_SSOC_COM_Comp(i,j) = stdl_SSOC_COM_Comp*randn + mul _ SSOC _COM_ Comp; end end end 298 USC Viterbi School of Engineering PhD Thesis iful > wl_CE_COM_Comp + w2_CE_COM_Comp % iful > probability ofl charging event today, (or CE >= 2 ), generate CS T2, D2, S SOC2 while rv2_CST_COM_Comp(i,j) <= 0 II rv2_CST_COM_Comp(i,j) > 23.5 if u2 <= wl_CST_COM_Comp rv2_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; else rv2_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end while rv2_D_COM_Comp(i,j) <= 0 if u3 <= wl_D_COM_Comp rv2_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; end end while rv2_SSOC_COM_Comp(i,j) <= 0 II rv2_SSOC_COM_Comp(i,j) > 100 if u4 <= wl_SSOC_COM_Comp rv2_SSOC_COM_Comp(i,j) = stdl_SSOC_COM_Comp*randn + mul _ SSOC _COM_ Comp; end end end iful > wl_CE_COM_Comp + w2_CE_COM_Comp + w3_CE_COM_Comp % iful > probability of2 charging event today, (or CE >= 3), generate CST3, D3, SSOC3 while rv3_CST_COM_Comp(i,j) <= 0 II 1v3_CST_COM_Comp(i,j) > 23.5 if u2 <= wl_CST_COM_Comp rv3_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; else rv3_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end while rv3_D_COM_Comp(i,j) <= 0 if u3 <= wl_D_COM_Comp rv3_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; end end while rv3_SSOC_COM_Comp(i,j) <= 0 II rv3_SSOC_COM_Comp(i,j) > 100 if u4 <= wl_SSOC_COM_Comp rv3_SSOC_COM_Comp(i,j) = stdl_SSOC_COM_Comp*randn + mul_SSOC_COM_Comp; end end end if ul > wl_CE_COM_Comp + w2_CE_COM_Comp + w3_CE_COM_Comp + w4_CE_COM_Comp % iful > probability of3 charging event today, (or CE >= 4), generate CST4, D4, SSOC4 while rv4_CST_COM_Comp(i,j) <= 0 II rv4_CST_COM_Comp(i,j) > 23.5 if u2 <= wl_CST_COM_Comp 299 USC Viterbi School of Engineering rv4_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; PhD Thesis else rv4_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end while rv4_D_COM_Comp(i,j) <= 0 if u3 <= wl_D_COM_Comp rv4_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; end end while rv4_SSOC_COM_Comp(i,j) <= 0 II rv4_SSOC_COM_Comp(i,j) > 100 if u4 <= wl_SSOC_COM_Comp rv4_SSOC_COM_Comp(i,j) = stdl_SSOC_COM_Comp*randn + mul _ SSOC _COM_ Comp; end end end iful > wl_CE_COM_Comp + w2_CE_COM_Comp + w3_CE_COM_Comp + w4_CE_COM_Comp + w5_CE_COM_Comp % iful >probability of 4 charging event today, (or CE = 5), generate CST5, D5, SSOC5 while rv5_CST_COM_Comp(i,j) <= 0 II rv5_CST_COM_Comp(i,j) > 23.5 if u2 <= wl_CST_COM_Comp rv5_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; else rv5_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end while rv5_D_COM_Comp(i,j) <= 0 if u3 <= wl_D_COM_Comp rv5_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; end end while rv5_SSOC_COM_Comp(i,j) <= 0 II rv5_SSOC_COM_Comp(i,j) > 100 if u4 <= wl_SSOC_COM_Comp rv5_SSOC_COM_Comp(i,j) = stdl_SSOC_COM_Comp*randn + mul _ SSOC _COM_ Comp; end end end end end % determine ifthere is any overlap between charging event 2 and charging % event 1. If there is, re-generate CST2, D2 and re-determine for j=l:n2 for i=l:nl *p2 if rv2 _CST_ COM_ Comp(i,j)>O 300 USC Viterbi School of Engineering PhD Thesis while rv2_CST_COM_Comp(i,j) + rv2_D_COM_Comp(i,j) > rvl_CST_COM_Comp(i,j) && rv2_CST_COM_Comp(i,j) < rvl_CST_COM_Comp(i,j) + rvl_D_COM_Comp(i,j) u22 =rand; u33 = rand; if u22 <= wl_CST_COM_Comp rv2_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; while rv2_CST_COM_Comp(i,j)<=O II rv2_CST_COM_Comp(i,j)>23.5 rv2_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; end else rv2_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; while rv2_CST_COM_Comp(i,j)<=O II rv2_CST_COM_Comp(i,j)>23.5 rv2_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end if u33 <= wl_D_COM_Comp rv2_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; while rv2_D_COM_Comp(i,j) <= 0 rv2_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; end end end end end end % determine if there is any overlap between charging event 3 and charging % event 1, 2. If there is, re-generate CST3, D3 and re-determine for j=l:n2 for i=l:nl *p2 if rv3 _CST_ COM_ Comp(i,j)>O while (rv3_CST_COM_Comp(i,j) + rv3_D_COM_Comp(i,j) > rvl_CST_COM_Comp(i,j) && rv3_CST_COM_Comp(i,j) < rvl_CST_COM_Comp(i,j) + rvl_D_COM_Comp(i,j)) II (rv3_CST_COM_Comp(i,j) + rv3_D_COM_Comp(i,j) > rv2_CST_COM_Comp(i,j) && rv3_CST_COM_Comp(i,j) < rv2_CST_COM_Comp(ij) + rv2_D_COM_Comp(i,j)) u22 =rand; u33 = rand; if u22 <= wl_CST_COM_Comp rv3_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; while rv3_CST_COM_Comp(i,j)<=O II rv3_CST_COM_Comp(i,j)>23.5 rv3_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; end else rv3_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; while rv3_CST_COM_Comp(i,j)<=O II rv3_CST_COM_Comp(i,j)>23.5 rv3_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end 301 USC Viterbi School of Engineering PhD Thesis if u33 <= wl_D_COM_Comp rv3_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; while rv3_D_COM_Comp(i,j) <= 0 rv3_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul _ D _COM_ Comp; end end end end end end % determine if there is any overlap between charging event 4 and charging % event 1, 2 and 3. If there is, re-generate CST4, D4 and re-determine for j=l:n2 for i=l:nl *p2 if rv4_CST_COM_Comp(i,j)>O while (rv4_CST_COM_Comp(i,j) + rv4_D_COM_Comp(i,j) > rvl_CST_COM_Comp(i,j) && rv4_CST_COM_Comp(i,j) < rvl_CST_COM_Comp(i,j) + rvl_D_COM_Comp(i,j)) II (rv4_CST_COM_Comp(i,j) + rv4_D_COM_Comp(i,j) > rv2_CST_COM_Comp(i,j) && rv4_CST_COM_Comp(i,j) < rv2_CST_COM_Comp(i,j) + rv2_D_COM_Comp(i,j)) II (rv4_CST_COM_Comp(i,j) + rv4_D_COM_Comp(i,j) > rv3_CST_COM_Comp(i,j) && rv4_CST_COM_Comp(i,j) < rv3_CST_COM_Comp(i,j) + rv3_D_COM_Comp(i,j)) u22 = rand; u33 = rand; if u22 <= wl_CST_COM_Comp rv4_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; while rv4_CST_COM_Comp(i,j)<=O II rv4_CST_COM_Comp(i,j)>23.5 rv4_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; end else rv4_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; while rv4_CST_COM_Comp(i,j)<=O II rv4_CST_COM_Comp(i,j )>23.5 rv4_CST_COM_Comp(i,j ) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end if u33 <= wl_D_COM_Comp mul _ D _COM_ Comp; end end end end end rv4_D_COM_Comp(i,j ) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; while rv4_D_COM_Comp(i,j) <= 0 rv4_D_COM_Comp(i,j ) = stdl_D_COM_Comp*randn + end % determine if there is any overlap between charging event 5 and charging % event 1, 2,3 and 4. If there is, re-generate CST5, D5 and re-determine for j=l:n2 for i=l:nl *p2 302 USC Viterbi School of Engineering PhD Thesis if rv5 _CST_ COM_ Comp(i,j)>O while (rv5_CST_COM_Comp(i,j) + rv5_D_COM_Comp(i,j) > rvl_CST_COM_Comp(i,j) && rv5_CST_COM_Comp(i,j) < rvl _CST_COM_Comp(i,j) + rvl_D_COM_Comp(i,j)) II (rv5_CST_COM_Comp(i,j) + rv5_D_COM_Comp(i,j) > rv2_CST_COM_Comp(i,j) && rv5_CST_COM_Comp(i,j) < rv2_CST_COM_Comp(i,j) + rv2_D_COM_Comp(i,j)) II (rv5_CST_COM_Comp(i,j) + rv5_D_COM_Comp(i,j) > rv3_CST_COM_Comp(i,j) && rv5_CST_COM_Comp(i,j) < rv3_CST_COM_Comp(i,j) + rv3_D_COM_Comp(i,j)) II (rv5_CST_COM_Comp(i,j) + rv5_D_COM_Comp(i,j) > rv4_CST_COM_Comp(i,j) && rv5_CST_COM_Comp(i,j) < rv4_CST_COM_Comp(i,j) + rv4_D_COM_Comp(i,j)) u22 =rand; u33 = rand; if u22 <= wl_CST_COM_Comp rv5_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; while rv5_CST_COM_Comp(i,j)<=O II rv5_CST_COM_Comp(i,j)>23.5 rv5_CST_COM_Comp(i,j) = stdl_CST_COM_Comp*randn + mul_CST_COM_Comp; end else rv5 _CST_ COM_ Comp(i,j) = std2_ CST_ COM_ Comp*randn + mu2_CST_COM_Comp; while rv5_CST_COM_Comp(i,j)<=O II rv5_CST_COM_Comp(i,j)>23.5 rv5_CST_COM_Comp(i,j) = std2_CST_COM_Comp*randn + mu2_CST_COM_Comp; end end if u33 <= wl_D_COM_Comp mul_D_COM_Comp; end end end end end rv5_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + mul_D_COM_Comp; while rv5_D_COM_Comp(i,j) <= 0 rv5_D_COM_Comp(i,j) = stdl_D_COM_Comp*randn + end Dl_actual_COM_Comp = ones(round(nl *p2),n2)*-l ; EV!_ demand_ COM_ Comp= zeros(round(nl *p2),n2); D2_actual_COM_Comp = ones(round(nl *p2),n2)*-l ; EV2 _demand_ COM_ Comp = zeros(round(nl *p2),n2); D3_actual_COM_Comp = ones(round(nl *p2),n2)*-l; EV3_demand_COM_Comp = zeros(round(nl *p2),n2); D4_actual_COM_Comp = ones(round(nl *p2),n2)*-l ; EV4_demand_COM_Comp = zeros(round(nl *p2),n2); D5_actual_COM_Comp = ones(round(nl *p2),n2)*-l; EV5 _demand_ COM_ Comp = zeros(round(nl *p2),n2); for j=l:n2 for i=l:nl *p2 if rvl_D_COM_Comp(i,j)>O && rvl_SSOC_COM_Comp(i,j)< lOO EVl_demand_COM_Comp(i,j) = min(min(rvl_D_COM_Comp(i,j)*L2_P _COM_Comp, (l00-1vl_SSOC_COM_Comp(i,j))/lOO*bat_COM_Comp), bat_COM_Comp); Dl _actual_ COM_ Comp(i,j) = EVl _demand_ COM_ Comp(i,j )/L2 _P _COM_ Comp; end 303 USC Viterbi School of Engineering PhD Thesis if rv2_D_COM_Comp(i,j)>O && rv2_SSOC_COM_Comp(i,j)<l00 EV2_demand_COM_Comp(i,j) = min(min(rv2_D_COM_Comp(i,j)*L2_P _COM_Comp, (100-rv2_SSOC_COM_Comp(i,j))/100*bat_COM_Comp), bat_COM_Comp); D2_actual_COM_Comp(i,j) = EV2_demand_COM_Comp(i,j)/L2_P _COM_Comp; end if rv3_D_COM_Comp(i,j)>O && rv3_SSOC_COM_Comp(i,j)<l00 EV3_demand_COM_Comp(i,j) = min(min(rv3_D_COM_Comp(i,j)*L2_P _COM_Comp, (100-rv3_SSOC_COM_Comp(i,j))/100*bat_COM_Comp), bat_COM_Comp); D3 _actual_ COM_ Comp(i,j) = EV3 _demand_ COM_ Comp(i,j )/L2 _P _COM_ Comp; end if rv4_D_COM_Comp(i,j)>O && rv4_SSOC_COM_Comp(i,j)<l00 EV4_demand_COM_Comp(i,j) = min(min(rv4_D_COM_Comp(i,j)*L2_P _COM_Comp, (100-rv4_SSOC_COM_Comp(i,j))/100*bat_COM_Comp), bat_COM_Comp); D4_actual_COM_Comp(i,j) = EV4_demand_COM_Comp(i,j)/L2_P _COM_Comp; end if rv5_D_COM_Comp(i,j)>O && rv5_SSOC_COM_Comp(i,j)<l00 EV5_demand_COM_Comp(i,j) = min(min(rv5_D_COM_Comp(i,j)*L2_P _COM_Comp, (100-rv5_SSOC_COM_Comp(i,j))/100*bat_COM_Comp), bat_COM_Comp); D5_actual_COM_Comp(i,j) = EV5_demand_COM_Comp(i,j)/L2_P _COM_Comp; end end end rv_CST_COM_Comp = [rvl_CST_COM_Comp; rv2_CST_COM_Comp; rv3_CST_COM_Comp; rv4_CST_COM_Comp; rv5_CST_COM_Comp]; rv_D_COM_Comp = [rvl_D_COM_Comp; rv2_D_COM_Comp; rv3_D_COM_Comp; rv4_D_COM_Comp; rv5_D_COM_Comp]; rv_SSOC_COM_Comp = [rvl_SSOC_COM_Comp; rv2_SSOC_COM_Comp; rv3_SSOC_COM_Comp; rv4_SSOC_COM_Comp; rv5_SSOC_COM_Comp]; D_actual_COM_Comp = [Dl_actual_COM_Comp; D2_actual_COM_Comp; D3_actual_COM_Comp; D4_actual_COM_Comp; D5_actual_COM_Comp]; L2_P _COM_SUV = 6.6; % Level 2 charging power= 6.6 kW bat_ COM_ SUV = 48; % battery size of Electric SUV = 48 kWh mg(' shuffle'); rvl_CST_COM_SUV = ones(round(nl *p3),n2)*-l; rvl_D_COM_SUV = ones(round(nl *p3),n2)*-l; rvl_SSOC_COM_SUV = ones(round(nl *p3),n2)*-l; rv2_CST_COM_SUV = ones(round(nl *p3),n2)*-l ; rv2_D_COM_SUV = ones(round(nl *p3),n2)*-l; rv2_SSOC_COM_SUV = ones(round(nl *p3),n2)*-l; rv3_CST_COM_SUV = ones(round(nl *p3),n2)*-l; rv3_D_COM_SUV = ones(round(nl *p3),n2)*-l; rv3_SSOC_COM_SUV = ones(round(nl *p3),n2)*-l; rv4_CST_COM_SUV = ones(round(nl *p3),n2)*-l ; rv4_D_COM_SUV = ones(round(nl *p3),n2)*-l; rv4_SSOC_COM_SUV = ones(round(nl *p3),n2)*-l; wl_CE_COM_SUV = 0.391; w2_CE_COM_SUV = 0.356; w3_CE_COM_SUV = 0.180; w4_CE_COM_SUV = 0.056; w5_CE_COM_SUV = 0.017; CST al COM SUV = 0.01905;% (0.0153, 0.0228) 304 CST bl COM SUV = - - - CST cl COM SUV= - - - CST a2 COM SUV = - - - CST b2 COM SUV = - - - CST c2 COM SUV = USC Viterbi School of Engineering 15.41;% (15.31, 15.52) 0.7237;% (0.5447, 0.9026) 0.03111;% (0.02887, 0.03336) 14.54;% (14.37, 14.7) 4.082;% (3.836, 4.329) mul_CST_COM_SUV = CST_bl_COM_SUV; % meanl_CST =bl in Gaussian pdf mu2_CST_COM_SUV = CST_b2_COM_SUV; PhD Thesis stdl_CST_COM_SUV = CST_cl_COM_SUV/sqrt(2); % standard deviation 1 CST = cl/sqrt(2) in Gaussian pdf std2 _CST_ COM_ SUV = CST_ c2 _COM_ SUV/sqrt(2); coefl_CST_COM_SUV = stdl_CST_COM_SUV*sqrt(2*pi)*CST_al_COM_SUV; % coefficientl of CST = std*sqrt(2*pi)*al coef2_CST_COM_SUV = std2_CST_COM_SUV*sqrt(2*pi)*CST_a2_COM_SUV; wl_CST_COM_SUV = coefl_CST_COM_SUV/(coefl_CST_COM_SUV + coef2_CST_COM_SUV); % weightl of CST= coefl/sum(coefl +coef2+coef3); w2_CST_COM_SUV = coef2_CST_COM_SUV/(coefl_CST_COM_SUV + coef2_CST_COM_SUV); D_al_COM_SUV= 0.2528;% (0.1953, 0.3103) D_bl_COM_SUV = -0.1412;% (-1.185, 0.9024) D cl COM_SUV = 2.046;% (1.123, 2.969) mul_D_COM_SUV = D_bl_COM_SUV; %mean_D =bl in Gaussian pdf stdl_D_COM_SUV = D_cl_COM_SUV/sqrt(2); % standard deviation 1 D = cl/sqrt(2) in Gaussian pdf coefl_D_COM_SUV = stdl_D_COM_SUV*sqrt(2*pi)*D_al_COM_SUV; % coefficientl ofD = std*sqrt(2*pi)*al wl_D_COM_SUV = coefl_D_COM_SUV/coefl_D_COM_SUV; % weightl ofD = coefl/sum( coefl :coef3 ); SSOC_al_COM_SUV = 0.02553;% (0.02362, 0.02744) SSOC_bl_COM_SUV = 75.12;% (73.62, 76.63) SSOC_cl_COM_SUV = 23.31;% (20.96, 25.66) mul_SSOC_COM_SUV = SSOC bl COM_SUV; % meanl_CST = bl in Gaussian pdf stdl_SSOC_COM_SUV = SSOC_cl_COM_SUV/sqrt(2); % standard deviation 1 CST = cl/sqrt(2) in Gaussian pdf coefl_SSOC_COM_SUV = stdl_SSOC_COM_SUV*sqrt(2*pi)*SSOC_al_COM_SUV; % coefficientl of CST = std*sqrt(2*pi)*al wl_SSOC_COM_SUV = coefl_SSOC_COM_SUV/coefl_SSOC_COM_SUV; % weightl of CST = coefl/coefl; for j=l:n2 for i=l:nl *p3 ul = rand; % ul =random variable to check how many times the EV is charging today u2 = rand; % u2=random variable follows CST pdf u3 = rand; % u3=random variable follows D pdf u4 = rand; % u4=random variable follows SSOC pdf iful > wl_CE_COM_SUV % iful > probability ofO charging event today, (or CE >= 1), generate CSTl, Dl, SSOCl while rvl_CST_COM_SUV(i,j) <= 0 II rvl_CST_COM_SUV(i,j) > 23.5 if u2 <= wl CST COM SUV - - - rvl_CST_COM_SUV(i,j ) = stdl_CST_COM_SUV*randn + mul CST_COM_SUV; else rvl_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; end end 305 USC Viterbi while rvl_D_COM_SUV(i,j) <= 0 if u3 <= wl D COM SUV - - - School of Engineering PhD Thesis rv l _ D _COM_ SUV(i,j) = stdl _ D _COM_ SUV*randn + mul _ D _COM_ SUV; end end while rvl_SSOC_COM_SUV(i,j) <= 0 I I rvl_SSOC_COM_SUV(i,j) > 100 if u4 <= wl SSOC COM SUV - - - rvl_SSOC_COM_SUV(i,j) = stdl_SSOC_COM_SUV*randn + mul SSOC_COM_SUV; end end end iful > wl_CE_COM_SUV + w2_CE_COM_SUV % iful > probability ofl charging event today, (or CE >= 2), generate CST2, D2, SSOC2 while rv2_CST_COM_SUV(i,j) <= 0 II rv2_CST_COM_SUV(i,j) > 23.5 if u2 <= wl CST COM SUV - - - rv2_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + mul_CST_COM_SUV; else rv2_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; end end while rv2_D_COM_SUV(i,j) <= 0 if u3 <= wl D COM SUV - - - rv2_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; end end while rv2_SSOC_COM_SUV(i,j) <= 0 I I rv2_SSOC_COM_SUV(i,j) > 100 if u4 <= wl SSOC COM SUV - - - rv2_SSOC_COM_SUV(i,j) = stdl_SSOC_COM_SUV*randn + mul SSOC_COM_SUV; end end end if ul > wl_CE_COM_SUV + w2_CE_COM_SUV + w3_CE_COM_SUV % iful > probability of2 charging event today, (or CE >= 3), generate CST3, D3, SSOC3 while rv3_CST_COM_SUV(i,j) <= 0 II rv3_CST_COM_SUV(i,j) > 23.5 if u2 <= wl CST COM SUV - - - rv3_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + mul CST_COM_SUV; else rv3_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; end end while rv3_D_COM_SUV(i,j) <= 0 if u3 <= wl D COM SUV - - - rv3_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; end end 306 USC Viterbi School of Engineering while rv3_SSOC_COM_SUV(i,j) <= 0 I I rv3_SSOC_COM_SUV(i,j) > 100 if u4 <= wl SSOC COM SUV - - - rv3_SSOC_COM_SUV(i,j) = stdl_SSOC_COM_SUV*randn + mul SSOC_COM_SUV; end end end iful > wl CE COM SUV+ w2 CE COM SUV+ w3 CE COM SUV+ - PhD Thesis w4_CE_COM_SUV % iful > probability of3 charging event today, (or CE = 4), generate CST4, D4, SSOC4 while rv4_CST_COM_SUV(i,j) <= 0 II rv4_CST_COM_SUV(i,j) > 23.5 if u2 <= wl CST COM SUV - - - rv4_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + mul_CST_COM_SUV; else rv4_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; end end while rv4_D_COM_SUV(i,j) <= 0 if u3 <= wl D COM SUV - - - rv4_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; end end while rv4_SSOC_COM_SUV(i,j) <= 0 II rv4_SSOC_COM_SUV(i,j) > 100 if u4 <= wl SSOC COM SUV - - - rv4_SSOC_COM_SUV(i,j) = stdl_SSOC_COM_SUV*randn + mul SSOC_COM_SUV; end end end end end % determine if there is any overlap between charging event 2 and charging % event 1. If there is, re-generate CST2, D2 and re-determine for j=l:n2 for i=l:nl *p3 if rv2 _CST_ COM_ SUV(i,j)>O while rv2_CST_COM_SUV(i,j) + rv2_D_COM_SUV(i,j) > rvl_CST_COM_SUV(i,j) && rv2_CST_COM_SUV(i,j) < rvl_CST_COM_SUV(i,j) + rvl_D_COM_SUV(i,j) u22 =rand; u33 =rand; if u22 <= wl CST COM SUV mul_CST_COM_SUV; mul_CST_COM_SUV; - - rv2_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + while rv2_CST_COM_SUV(i,j)<=O II rv2_CST_COM_SUV(i,j)>23.5 rv2_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + end else rv2_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; 307 USC Viterbi School of Engineering PhD Thesis while rv2_CST_COM_SUV(i,j)<=O II rv2_CST_COM_SUV(i,j)>23.5 rv2_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; end end end end end end if u33 <= wl D COM SUV end rv2_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; while rv2_D_COM_SUV(i,j) <= 0 rv2_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; end % determine if there is any overlap between charging event 3 and charging % event 1, 2. If there is, re-generate CST3, D3 and re-determine for j=l:n2 for i=l:nl *p3 if rv3 _CST_ COM_ SUV(i,j)>O while (1v3_CST_COM_SUV(i,j) + rv3_D_COM_SUV(i,j) > rvl_CST_COM_SUV(i,j) && rv3_CST_COM_SUV(i,j) < rvl_CST_COM_SUV(i,j) + rvl_D_COM_SUV(i,j)) II (rv3_CST_COM_SUV(i,j) + rv3_D_COM_SUV(i,j) > rv2_CST_COM_SUV(i,j) && rv3_CST_COM_SUV(i,j) < rv2_CST_COM_SUV(i,j) + rv2_D_COM_SUV(i,j)) u22 = rand; u33 =rand; if u22 <= wl CST COM SUV - - rv3_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + mul_CST_COM_SUV; mul_CST_COM_SUV; while rv3_CST_COM_SUV(i,j)<=O II rv3_CST_COM_SUV(i,j )>23.5 rv3_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + end else rv3_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; mu2_CST_COM_SUV; end while rv3_CST_COM_SUV(i,j)<=O II 1v3_CST_COM_SUV(i,j)>23.5 rv3_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + end if u33 <= wl D COM SUV end end end end end rv3_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; while rv3_D_COM_SUV(i,j) <= 0 rv3_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; end % determine if there is any overlap between charging event 4 and charging % event 1, 2 and 3. If there is, re-generate CST4, D4 and re-determine for j=l:n2 308 USC Viterbi School of Engineering PhD Thesis for i=l:nl *p3 if rv4 _CST_ COM_SUV(i,j)>O while (rv4_CST_COM_SUV(i,j) + rv4_D_COM_SUV(i,j) > rvl_CST_COM_SUV(i,j) && rv4 _CST_ COM_ SUV(i,j) < rv 1 _CST_ COM_ SUV(i,j) + rv 1 _ D _COM_ SUV(i,j)) 11 (rv4_CST_COM_SUV(i,j) + rv4_D_COM_SUV(i,j) > rv2_CST_COM_SUV(i,j) && rv4_CST_COM_SUV(i,j) < rv2_CST_COM_SUV(i,j) + rv2_D_COM_SUV(i,j)) II (rv4_CST_COM_SUV(i,j) + rv4_D_COM_SUV(i,j) > rv3_CST_COM_SUV(i,j) && rv4_CST_COM_SUV(i,j) < rv3_CST_COM_SUV(i,j) + rv3_D_COM_SUV(i,j)) u22 = rand; u33 =rand; if u22 <= wl CST COM SUV - - - rv4_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + mul_CST_COM_SUV; while rv4_CST_COM_SUV(i,j)<=O II rv4_CST_COM_SUV(i,j)>23.5 rv4_CST_COM_SUV(i,j) = stdl_CST_COM_SUV*randn + mul_CST_COM_SUV; end else rv4_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; while rv4_CST_COM_SUV(i,j)<=O II rv4_CST_COM_SUV(i,j)>23.5 rv4_CST_COM_SUV(i,j) = std2_CST_COM_SUV*randn + mu2_CST_COM_SUV; end end end end end end if u33 <= wl D COM SUV end rv4_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; while rv4_D_COM_SUV(i,j) <= 0 rv4_D_COM_SUV(i,j) = stdl_D_COM_SUV*randn + mul_D_COM_SUV; end Dl_actual_COM_SUV = ones(round(nl *p3),n2)*-1; EVl _demand_ COM _SUV = zeros(round(nl *p3),n2); D2_actual_COM_SUV = ones(round(nl *p3),n2)*-l; EV2_demand_COM_SUV = zeros(round(nl *p3),n2); D3_actual_COM_SUV = ones(round(nl *p3),n2)*-1; EV3_demand_COM_SUV = zeros(round(nl *p3),n2); D4_actual_COM_SUV = ones(round(nl *p3),n2)*-l; EV4_demand_COM_SUV = zeros(round(nl *p3),n2); for j=l:n2 for i=l:nl *p3 ifrvl_D_COM_SUV(ij)>O && rvl_SSOC_COM_SUV(i,j)< lOO EVl_demand_COM_SUV(i,j) = min(min(rvl_D_COM_SUV(i,j)*L2_P _COM_SUV, (100- rvl_SSOC_COM_SUV(i,j))/lOO*bat_COM_SUV), bat_COM_SUV); Dl_actual_COM_SUV(i,j) = EVl_demand_COM_SUV(i,j)/L2_P _COM_SUV; end if rv2_D_COM_SUV(i,j)>O && rv2_SSOC_COM_SUV(i,j)<100 EV2_demand_COM_SUV(i,j) = min(min(rv2_D_COM_SUV(i,j)*L2_P _COM_SUV, (100- rv2_SSOC_COM_SUV(i,j))/100*bat_COM_SUV), bat_COM_SUV); D2_actual_COM_SUV(i,j) = EV2_demand_COM_SUV(i,j)/L2_P _COM_SUV; 309 USC Viterbi School of Engineering PhD Thesis end ifrv3_D_COM_SUV(i,j)>O && rv3_SSOC_COM_SUV(i,j)<l00 EV3_demand_COM_SUV(i,j) = min(min(rv3_D_COM_SUV(i,j)*L2_P _COM_SUV, (100- rv3_SSOC_COM_SUV(ij))/100*bat_COM_SUV), bat_COM_SUV); D3_actual_COM_SUV(i,j) = EV3_demand_COM_SUV(i,j)/L2_P _COM_SUV; end ifrv4_D_COM_SUV(i,j)>O && rv4_SSOC_COM_SUV(i,j)<l00 EV4_demand_COM_SUV(i,j) = min(min(rv4_D_COM_SUV(i,j)*L2_P _COM_SUV, (100- rv4_SSOC_COM_SUV(ij))/100*bat_COM_SUV), bat_COM_SUV); D4_actual_COM_SUV(i,j) = EV4_demand_COM_SUV(i,j)/L2_P _COM_SUV; end end end rv_CST_COM_SUV = [rvl_CST_COM_SUV; rv2_CST_COM_SUV; rv3_CST_COM_SUV; rv4_CST_COM_SUV]; rv_D_COM_SUV = [rvl_D_COM_SUV; rv2_D_COM_SUV; rv3_D_COM_SUV; rv4_D_COM_SUV]; rv_SSOC_COM_SUV = [rvl_SSOC_COM_SUV; rv2_SSOC_COM_SUV; rv3_SSOC_COM_SUV; rv4_SSOC_COM_SUV]; D_actual_COM_SUV = [Dl_actual_COM_SUV; D2_actual_COM_SUV; D3_actual_COM_SUV; D4 _actual_ COM_ SUV]; rv_CST_COM = [rv_CST_COM_Comp; rv_CST_COM_SUV]; rv_D_COM = [rv_D_COM_Comp; rv_D_COM_SUV]; rv_SSOC_COM = [rv_SSOC_COM_Comp; rv_SSOC_COM_SUV]; D_actual_COM = [D_actual_COM_Comp; D_actual_COM_SUV]; rv_D_COM_freq = zeros(l2,n2); rv _D _ COM_Hr = zeros(l2,l); for i=0:0.5:5.5 rv_D_COM_Hr(i*2+ l )=i; end for j=l:n2 for i=l:length(rv_D_COM) for k=1:12 if rv _D _ COM(i,j )>rv _D _ COM_Hr(k) && rv _D _ COM(i,j)<=rv _D _ COM_Hr(k)+0.5 rv _ D _COM_ freq(k,j) = rv _ D _COM_ freq(k,j )+ 1; end end end end D_actual_COM_freq = zeros(l2,n2); for j=l:n2 for i=l:length(D_actual_COM) for k=1:12 ifD_actual_COM(i,j)>rv_D_COM_Hr(k) && D _actual_ COM(i,j)<=rv _ D _COM_ Hr(k)+0.5 D_actual_COM_freq(k,j) = D_actual_COM_freq(k,j)+ l ; end end end end 310 USC Viterbi figure subplot(2,2, 1) hist(rv _CST_ COM,48); set(gca,'xtick', 0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); School of Engineering title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(2,2,2) % hist(rv_D_COM); % set(gca,'xtick',0:7); % xlim([O 7]); bar(rv _ D _COM_ Hr,rv _ D _COM_ freq); xlabel('Plug-in Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) % hist(D_actual_COM); % set(gca,'xtick',0:6); % xlim([O.l 6]); bar(rv_D_COM_Hr,D_actual_COM_freq); xlabel('Charging Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv _ SSOC _COM, 100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') figure subplot(2,2, 1) hist(rv _CST_ COM_ Comp,48); set(gca,'xtick',0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(2,2,2) hist(rv _D_COM_Comp); set(gca,'xtick',0:7); xlim([O 7]); xlabel('Plug-in Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) hist(D _actual_ COM_ Comp); set(gca,'xtick',0:6); xlim([O.l 6]); xlabel('Charging Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv _ SSOC _COM_ Comp, 100); 311 PhD Thesis xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); USC Viterbi School of Engineering title('Histogram of Monte Carlo Generated Charging Start SOC Variables') figure subplot(2,2, 1) hist(rv _CST_ COM_SUV,48); set(gca,'xtick', 0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(2,2,2) hist(rv _ D _COM_ SUV); set(gca,'xtick',0:7); xlim([O 7]); xlabel('Plug-in Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Plug-in Duration Variables') subplot(2,2,3) hist(D _actual_ COM_ SUV); set(gca,'xtick',0:6); xlim([O. l 6]); xlabel('Charging Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(2,2,4) hist(rv_SSOC_COM_SUV,100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') PhD Thesis % Convert EV Demand Forecast into 24-hr Format and Plot, then compare with Actual Weekday EV Demand (05/20115-02/04/17) EV_ Demand_ 24hr _COM_ Comp= zeros(n2,24 ); EV1_Demand_72hr _COM_ Comp= zeros(n2, 72); EV2_Demand_72hr_COM_Comp = zeros(n2,72); EV3_Demand_72hr_COM_Comp = zeros(n2,72); EV4_Demand_72hr_COM_Comp = zeros(n2,72); EV5 _Demand_ 72hr _COM_ Comp= zeros(n2, 72); EV _Demand_24hr_avg_COM_Comp = zeros(l ,24); EV_ Demand_ 24hr _COM= zeros(n2,24 ); EV_ Demand_ 24hr _avg_ COM = zeros(l,24 ); for k=l :n2 for j=l:nl *p2 for i=l :72 % Charging Event = 1 iffloor(rvl_CST_COM_Comp(j,k)) == floor(rvl_CST_COM_Comp(j,k) + Dl_actual_COM_Comp(j,k)) && floor(rvl_CST_COM_Comp(j ,k)) == i-1 % if Charging Start Hr= Charging End Hr && Charging Start Hr= i-1 312 USC Viterbi School of Engineering PhD Thesis EVl_Demand_72hr_COM_Comp(k,i) = EVl_Demand_72hr_COM_Comp(k,i) + EVl_demand_COM_Comp(j,k); % assign all EV demand to this hour slot else if floor(rvl_ CST_ COM_ Comp(j,k)+Dl_ actual_ COM_ Comp(j,k))>floor(rvl_ CST_ COM_ Comp(j,k)) % else if CEhr (charging ending hr) > CS hr (charging start hr) if floor(rvl_CST_COM_Comp(j,k)) == i-1 % ifCShr = (i-l)th hour EVl_Demand_72hr_COM_Comp(k,i) = EVl_Demand_72hr_ COM_ Comp(k,i) + (ceil(rvl_ CST_ COM_ Comp(j,k))- rvl_ CST_ COM_ Comp(j,k))*EVl_demand_ COM_ Comp(j,k)/Dl_actual_ COM_ Comp(j,k); % EV demand at this hour = ( ceil(CST) - CST)*Demand/Duration end if floor(rvl_ CST_ COM_ Comp(j,k)+Dl_actual_ COM_ Comp(j,k)) == i-1 % ifCEhr = (i-l)th hour EVl_Demand_72hr_COM_Comp(k,i) = EVl_Demand_72hr_COM_Comp(k,i) + (rvl_CST_COM_Comp(j,k)+Dl_actual_COM_Comp(j,k) floor(rvl_ CST_ COM_ Comp(j,k)+Dl _actual_ COM_ Comp(j,k)))*EVl _demand_ COM_ Comp(j,k)/Dl _ actu al_ COM_ Comp(j,k); % EV demand at this hour= (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rvl_CST_COM_Comp(j,k)) && i-1 < floor(rvl _CST_COM_Comp(j,k)+Dl_actual_COM_Comp(j,k)) % ifCShr < (i-l)th hour < CEhr EVl_Demand_72hr_COM_Comp(k,i) = EVl_Demand_72hr_ COM_ Comp(k,i) + EVl_demand_ COM_ Comp(j,k)/Dl_actual_ COM_ Comp(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 2 iffloor(rv2_CST_COM_Comp(j,k)) == floor(rv2_CST_COM_Comp(j,k) + D2_actual_COM_Comp(j,k)) && floor(rv2_CST_COM_Comp(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV2_Demand_72hr_COM_Comp(k,i) = EV2_Demand_72hr_COM_Comp(k,i) + EV2 _demand_ COM_ Comp(j,k); % assign all EV demand to this hour slot else if floor(rv2 _CST_ COM_ Comp(j,k)+D2 _actual_ COM_ Comp(j,k))>floor(rv2_ CST_ COM_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv2_CST_COM_Comp(j,k)) == i-1 % ifCShr = (i-l)th hour EV2 _Demand_ 72hr _COM_ Comp(k,i) = EV2_Demand_72hr_ COM_ Comp(k,i) + (ceil(rv2_ CST_ COM_ Comp(j,k))- rv2_ CST_ COM_ Comp(j,k))*EV2_demand_ COM_ Comp(j,k)/D2_actual_ COM_ Comp(j,k); % EV demand at this hour = ( ceil( CST) - CS T)*Demand/Duration end if floor(rv2_ CST_ COM_ Comp(j,k)+D2_actual_ COM_ Comp(j,k)) == i-1 % ifCEhr = (i-l)th hour EV2 _Demand_ 72hr _COM_ Comp(k,i) = EV2_Demand_72hr_COM_Comp(k,i) + (rv2_CST_COM_Comp(j,k)+D2_actual_COM_Comp(j,k) floor(rv2_ CST_ COM_ Comp(j,k)+D2 _actual_ COM_ Comp(j,k)))*EV2 _demand_ COM_ Comp(j,k)/D2 _ actu al_COM_Comp(j,k); % EV demand at this hour= (CST+D-CEhr)*Demand/Duration end if i-1 > floor(rv2_CST_COM_Comp(j,k)) && i-1 < floor(rv2 _CST_ COM_ Comp(j,k)+D2 _actual_ COM_ Comp(j,k)) % if CS hr < (i-1 )th hour < CEhr EV2_Demand_72hr_COM_Comp(k,i) = EV2_Demand_72hr_COM_Comp(k,i) + EV2_demand_COM_Comp(j,k)/D2_actual_COM_Comp(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied 313 USC Viterbi School of Engineering PhD Thesis end end end % Charging Event = 3 iffloor(rv3_CST_COM_Comp(j,k)) == floor(rv3_CST_COM_Comp(j,k) + D3_actual_COM_Comp(j,k)) && floor(rv3_CST_COM_Comp(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV3_Demand_72hr_COM_Comp(k,i) = EV3_Demand_72hr_COM_Comp(k,i) + EV3_demand_COM_Comp(j,k); % assign all EV demand to this hour slot else if floor(rv3 _CST_ COM_ Comp(j,k)+D3 _actual_ COM_ Comp(j,k))>floor(rv3 _CST_ COM_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv3_CST_COM_Comp(j ,k)) == i-1 % if CShr = (i-1 )th hour EV3 _Demand_ 72hr _COM_ Comp(k,i) = EV3_Demand_72hr_COM_Comp(k,i) + (ceil(rv3_CST_COM_Comp(j,k))- rv3_ CST_ COM_ Comp(j,k))*EV3 _demand_ COM_ Comp(j,k)/D3 _actual_ COM_ Comp(j,k); % EV demand at this hour= ( ceil(CST) - CST)*Demand/Duration end if floor(rv3 _CST_ COM_ Comp(j,k)+D3_actual_ COM_ Comp(j,k)) == i-1 % ifCEhr = (i-l)th hour EV3 _Demand_ 72hr _COM_ Comp(k,i) = EV3_Demand_72hr_COM_Comp(k,i) + (rv3_CST_COM_Comp(j,k)+D3_actual_COM_Comp(j,k) floor(rv3_ CST_ COM_ Comp(j,k)+D3 _actual_ COM_ Comp(j,k)))*EV3 _demand_ COM_ Comp(j,k)/D3 _actu al_ COM_ Comp(j,k); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv3_CST_COM_Comp(j,k)) && i-1 < floor(rv3 _CST_ COM_ Comp(j,k)+D3 _actual_ COM_ Comp(j,k)) % if CS hr < (i-1 )th hour < CEhr EV3_Demand_72hr_COM_Comp(k,i) = EV3_Demand_72hr_COM_Comp(k,i) + EV3_demand_COM_Comp(j,k)/D3_actual_COM_Comp(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event = 4 iffloor(rv4_CST_COM_Comp(j,k)) == floor(rv4_CST_COM_Comp(j,k) + D4_actual_COM_Comp(j,k)) && floor(rv4_CST_COM_Comp(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr= i-1 EV4_Demand_72hr_COM_Comp(k,i) = EV4_Demand_72hr_COM_Comp(k,i) + EV4_demand_COM_Comp(j,k); % assign all EV demand to this hour slot else if floor(rv4 _CST_ COM_ Comp(j,k)+D4 _actual_ COM_ Comp(j,k))>floor(rv4 _CST_ COM_ Comp(j,k)) % else if CE hr (charging ending hr) > CS hr (charging start hr) if floor(rv4_CST_COM_Comp(j ,k)) == i-1 % if CShr = (i-1 )th hour EV4_Demand_72hr_COM_Comp(k,i) = EV4_Demand_72hr_COM_Comp(k,i) + (ceil(rv4_CST_COM_Comp(j,k)) rv4_CST_COM_Comp(j,k))*EV4_demand_COM_Comp(j,k)/D4_actual_COM_Comp(j,k); % EV demand at this hour = ( ceil( CST) - CS T)*Demand/Duration end if floor(rv4_ CST_ COM_ Comp(j,k)+D4_actual_ COM_ Comp(j,k)) == i-1 % if CEhr = (i-l)th hour EV4_Demand_72hr_COM_Comp(k,i) = EV4_Demand_72hr_COM_Comp(k,i) + (rv4_CST_COM_Comp(j,k)+D4_actual_COM_Comp(j,k)- 314 USC Viterbi School of Engineering PhD Thesis floor(rv4 _CST_ COM_ Comp(j,k)+D4_ actual_ COM_ Comp(j,k)))*EV4 _demand_ COM_ Comp(j,k)/D4 _ actu al_ COM_ Comp(j,k); % EV demand at this hour= (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv4_CST_COM_Comp(j,k)) && i-1 < floor(rv4_CST_COM_Comp(j,k)+D4_actual_COM_Comp(j,k)) % ifCShr < (i-l)th hour < CEhr EV4_Demand_72hr_COM_Comp(k,i) = EV4_Demand_72hr_COM_Comp(k,i) + EV4_demand_COM_Comp(j,k)/D4_actual_COM_Comp(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 5 iffloor(rv5_CST_COM_Comp(j,k)) == floor(rv5_CST_COM_Comp(j,k) + D5_actual_COM_Comp(j,k)) && floor(rv5_CST_COM_Comp(j,k)) == i-1 % if Charging Start Hr= Charging End Hr && Charging Start Hr = i-1 EV5_Demand_72hr_COM_Comp(k,i) = EV5_Demand_72hr_COM_Comp(k,i) + EV5_demand_COM_Comp(j,k); % assign all EV demand to this hour slot else if floor(rv5 _CST_ COM_ Comp(j,k)+D5 _actual_ COM_ Comp(j,k))>floor(rv5 _CST_ COM_ Comp(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv5_CST_COM_Comp(j,k)) == i-1 % ifCShr = (i-l)th hour EV5_Demand_72hr_COM_Comp(k,i) = EV5 _Demand_72hr_ COM_ Comp(k,i) + (ceil(rv5 _CST_ COM_ Comp(j,k))- rv5 _CST_ COM_ Comp(j,k))*EV5 _demand_ COM_ Comp(j,k)/D5 _actual_ COM_ Comp(j,k); % EV demand at this hour = ( ceil( CST) - CS T)*Demand/Duration end if floor(rv5 _CST_ COM_ Comp(j,k)+D5 _actual_ COM_ Comp(j,k)) == i-1 % ifCEhr = (i-l)th hour EV5_Demand_72hr_COM_Comp(k,i) = EV5_Demand_72hr_COM_Comp(k,i) + (rv5_CST_COM_Comp(j,k)+D5_actual_COM_Comp(j,k) floor(rv5_ CST_ COM_ Comp(j,k)+D5 _actual_ COM_ Comp(j,k)))*EV5 _demand_ COM_ Comp(j,k)/D5 _ actu al_COM_Comp(j,k); % EV demand at this hour= (CST+D-CEhr)*Demand/Duration end if i-1 > floor(rv5_CST_COM_Comp(j,k)) && i-1 < floor(rv5_CST_COM_Comp(j,k)+D5_actual_COM_Comp(j,k)) % ifCShr < (i-l)th hour < CEhr EV5_Demand_72hr_COM_Comp(k,i) = EV5 _Demand_ 72hr _COM_ Comp(k,i) + EV5 _demand_ COM_ Comp(j,k)/D5 _actual_ COM_ Comp(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end end end end for k= l :n2 for i=1:24 EV _Demand_24hr_COM_Comp(k,i) = EVl_Demand_72hr_COM_Comp(k,i) + EVl_Demand_72hr_COM_Comp(k,i+24) + EVl_Demand_72hr_COM_Comp(k,i+48) + EV2_Demand_72hr_COM_Comp(k,i) + EV2_Demand_72hr_COM_Comp(k,i+24) + EV2_Demand_72hr_COM_Comp(k,i+48) + EV3_Demand_72hr_COM_Comp(k,i) + EV3 _Demand_72hr_ COM_ Comp(k,i+24) + EV3_Demand_72hr_ COM_ Comp(k,i+48) + EV4_Demand_72hr_COM_Comp(k,i) + EV4_Demand_72hr_COM_Comp(k,i+24) + EV4_Demand_72hr_COM_Comp(k,i+48) + EV5_Demand_72hr_COM_Comp(k,i) + 315 USC Viterbi School of Engineering PhD Thesis EV5_Demand_72hr_COM_Comp(k,i+24) + EV5_Demand_72hr_COM_Comp(k,i+48); % convert charging start at late night and ends tomorrow into 24 hour format end end for i=1:24 EV _Demand_24hr_avg_ COM_ Comp(l,i) = mean(EV _Demand_24hr_ COM_ Comp(:,i)); end EV_ Demand_ 24hr _COM_ SUV= zeros(n2,24 ); EV1_Demand_72hr _COM_ SUV = zeros(n2, 72); EV2 _Demand_ 72hr _COM_ SUV= zeros(n2, 72); EV3_Demand_72hr_COM_SUV = zeros(n2,72); EV _Demand_24hr_avg_COM_SUV = zeros(l,24); for k= l :n2 for j=l:nl *p3 for i=l :72 % Charging Event = 1 iffloor(rvl_CST_COM_SUV(j,k)) == floor(rvl_CST_COM_SUV(j,k) + Dl_actual_COM_SUV(j,k)) && floor(rvl_CST_COM_SUV(j ,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr= i-1 EV1_Demand_72hr_COM_SUV(k,i) = EV1_Demand_72hr_COM_SUV(k,i) + EVl_demand_COM_SUV(j,k); % assign all EV demand to this hour slot else if floor(rvl_ CST_ COM_SUV(j,k)+Dl_actual_ COM_ SUV(j,k))>floor(rvl_ CST_ COM_SUV(j,k)) % else if CEhr (charging ending hr) > CS hr (charging start hr) if floor(rvl _CST_COM_SUV(j,k)) == i-1 % ifCShr = (i-l)th hour EVl_Demand_72hr_COM_SUV(k,i) = EVl_Demand_72hr_COM_SUV(k,i) + (ceil(rvl_CST_COM_SUV(j,k)) rvl_CST_COM_SUV(j,k))*EVl_demand_COM_SUV(j,k)/Dl_actual_COM_SUV(j,k); % EV demand at this hour= (ceil(CST) - CST)*Demand/Duration end if floor(rvl_CST_COM_SUV(j,k)+Dl_actual_COM_SUV(j,k)) == i-1 % ifCEhr = (i-l)th hour EVl_Demand_72hr_COM_SUV(k,i) = EVl_Demand_72hr_COM_SUV(k,i) + (rvl_CST_COM_SUV(j,k)+Dl_actual_COM_SUV(j,k) floor(rvl _ CST_ COM _SUV(j,k)+Dl _actual_ COM _SUV(j,k)))*EVl _demand_ COM_ SUV(j,k)/Dl _actual_ COM_ SUV(j,k); % EV demand at this hour= (CST +D-CEhr )*Demand/Duration end if i-1 > floor(rvl_CST_COM_SUV(j,k)) && i-1 < floor(rvl_CST_COM_SUV(j ,k)+Dl_actual_COM_SUV(j,k)) % ifCShr < (i-l)th hour < CEhr EV1_Demand_72hr_COM_SUV(k,i) = EVl_Demand_72hr_COM_SUV(k,i) + EVl_demand_COM_SUV(j,k)/Dl_actual_COM_SUV(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event= 2 iffloor(rv2_CST_COM_SUV(j,k)) == floor(rv2_CST_COM_SUV(j,k) + D2_actual_COM_SUV(j,k)) && floor(rv2_CST_COM_SUV(j,k)) == i-1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV2_Demand_72hr_COM_SUV(k,i) = EV2_Demand_72hr_COM_SUV(k,i) + EV2_demand_COM_SUV(j,k); % assign all EV demand to this hour slot 316 USC Viterbi School of Engineering PhD Thesis else if floor(rv2_ CST_ COM_SUV(j,k)+D2_actual_ COM_SUV(j,k))>floor(rv2_ CST_ COM_SUV(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv2_CST_COM_SUV(j,k)) == i-1 % ifCShr = (i-l)th hour EV2_Demand_72hr_COM_SUV(k,i) = EV2_Demand_72hr_ COM_SUV(k,i) + (ceil(rv2_ CST_ COM_SUV(j,k))- rv2_ CST_ COM_SUV(j,k))*EV2_demand_ COM_SUV(j,k)/D2_actual_ COM_SUV(j,k); % EV demand at this hour = (ceil(CST) - CST)*Demand/Duration end if floor(rv2_CST_COM_SUV(j,k)+D2_actual_COM_SUV(j,k)) == i-1 % ifCEhr = (i-l)th hour EV2_Demand_72hr_COM_SUV(k,i) = EV2_Demand_72hr_COM_SUV(k,i) + (rv2_CST_COM_SUV(j,k)+D2_actual_COM_SUV(j,k) floor(rv2_ CST_ COM_ SUV(j,k)+D2 _actual_ COM _SUV(j,k)))*EV2 _demand_ COM_ SUV(j,k)/D2 _actual_ COM_ SUV(j,k); % EV demand at this hour= (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv2_CST_COM_SUV(j,k)) && i-1 < floor(rv2_CST_COM_SUV(j,k)+D2_actual_COM_SUV(j,k)) % ifCShr < (i-l)th hour < CEhr EV2 _Demand_ 72hr _COM_ SUV(k,i) = EV2_Demand_72hr_COM_SUV(k,i) + EV2_demand_COM_SUV(j,k)/D2_actual_COM_SUV(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end end % Charging Event = 3 iffloor(rv3_CST_COM_SUV(j,k)) == floor(rv3_CST_COM_SUV(j,k) + D3_actual_COM_SUV(j,k)) && floor(rv3_CST_COM_SUV(j ,k)) == i-1 % if Charging Start Hr= Charging End Hr && Charging Start Hr = i-1 EV3_Demand_72hr_COM_SUV(k,i) = EV3_Demand_72hr_COM_SUV(k,i) + EV3_demand_COM_SUV(j,k); % assign all EV demand to this hour slot else if floor(rv3 _CST_ COM _SUV(j,k)+D3 _actual_ COM _SUV(j,k))>floor(rv3 _CST_ COM_ SUV(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rv3_CST_COM_SUV(j,k)) == i-1 % if CShr = (i-1 )th hour EV3_Demand_72hr_COM_SUV(k,i) = EV3 _Demand_72hr_ COM_SUV(k,i) + ( ceil(rv3 _CST_ COM_SUV(j,k))- rv3_ CST_ COM_SUV(j,k))*EV3_demand_ COM_SUV(j,k)/D3 _actual_ COM_SUV(j,k); % EV demand at this hour= (ceil(CST) - CST)*Demand/Duration end if floor(rv3 _CST_ COM_SUV(j,k)+D3 _actual_ COM_SUV(j,k)) == i-1 % ifCEhr = (i-l)th hour EV3 _Demand_ 72hr _COM_ SUV(k,i) = EV3_Demand_72hr_COM_SUV(k,i) + (rv3_CST_COM_SUV(j,k)+D3_actual_COM_SUV(j,k) floor(rv3_ CST_ COM _SUV(j,k)+D3 _actual_ COM _SUV(j,k)))*EV3 _demand_ COM _SUV(j,k)/D3 _actual_ COM_SUV(j,k); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rv3_CST_COM_SUV(j,k)) && i-1 < floor(rv3_CST_COM_SUV(j ,k)+D3_actual_COM_SUV(j,k)) % ifCShr < (i-l)th hour < CEhr EV3_Demand_72hr_COM_SUV(k,i) = EV3_Demand_72hr_COM_SUV(k,i) + EV3_demand_COM_SUV(j,k)/D3_actual_COM_SUV(j,k); % EV demand at this hour= Demand/Duration , it means this whole hour will be fully occupied end end 317 USC Viterbi School of Engineering PhD Thesis end end end end for k= l :n2 for i=1:24 EV _Demand_24hr_COM_SUV(k,i) = EVl_Demand_72hr_COM_SUV(k,i) + EVl_Demand_72hr_COM_SUV(k,i+24) + EVl_Demand_72hr_COM_SUV(k,i+48) + EV2_Demand_72hr_COM_SUV(k,i) + EV2_Demand_72hr_COM_SUV(k,i+24) + EV2_Demand_72hr_COM_SUV(k,i+48) + EV3_Demand_72hr_COM_SUV(k,i) + EV3_Demand_72hr_COM_SUV(k,i+24) + EV3_Demand_72hr_COM_SUV(k,i+48); % convert charging start at late night and ends tomorrow into 24 hour format end end for i=1:24 EV _Demand_24hr_avg_ COM_SUV(l,i) = mean(EV _Demand_24hr_ COM_SUV(:,i)); end for k=l:n2 for i=1:24 EV _Demand_24hr_COM(k,i) =EV _Demand_24hr_COM_Comp(k,i) + EV _Demand_24hr_COM_SUV(k,i); end end for i=1:24 EV _Demand_24hr_avg_ COM(l,i) = mean(EV _Demand_24hr_ COM(:,i)); end figure plot(O:l:23, EV _Demand_24hr_COM); xlabel('Time of Day'); ylabel('EV Charging Demand in kW' ); strl = sprintf('Commercial EV Charging (Level 2) Demand Forecast Based on Monte Carlo Simulation (o/od EVs: o/od %% Compact + o/od %% SUV, o/od times)',nl,p2*100,p3*100,n2); strl 1 =textwrap( {strl },85); title( str 11 ); set(gca,'xtick',0:23 ); % figure % plot(O: 1 :23,weekday_1516*nl *pl/26,0: 1:23,EV _Demand_24hr,'ro-'); % legend('Scaled Historical Weekday','Forecast'); % xlabel('Time of Day'); % ylabel('EV Charging Demand in kW'); % str2 = sprintf('Commercial Compact EV Scaled Historical Weekday Demand (05/20/15 - 02/04/17) vs EV Charging Demand Forecast (o/od EVs, o/od times)',nl *pl, n2); % title(str2); % set(gca,'xtick',0:23); figure % plot surf of EV_ Demand_ 24hr matrix hr = [O: 1:23]'; MC_num = [l:l:n2]'; surf(hr, MC_num, EV _Demand_24hr_COM); color bar 318 USC Viterbi School of Engineering PhD Thesis str3 = sprintf('Surf Plot of Commercial EV Charging Demand Forecast Based on Monte Carlo Simulation (%d EVs: o/od %% Compact + %d %% SUV, %d times)',nl,p2*100,p3*100,n2); str33=textwrap( { str3} ,85); title( str33 ); xlabel('Time of Day'); ylabel('Monte Carlo Simulation Time'); zlabel('EV Charging Demand in kW' ); hrq = 0:0.1 :23; MC_numq = l:l:n2; [hrq, MC_ numq] = meshgrid(hrq, MC_ numq); EV_ Demand_ 24hrq = interp2(hr, MC_ num, EV_ Demand_ 24hr _COM, hrq, MC_ numq, 'spline'); figure surf(hrq, MC_numq, EV _Demand_24hrq); color bar str4 = sprintf('Surf Plot with Spline of Commercial Compact EV Charging Demand Forecast Based on Monte Carlo Simulation (o/od EVs: %d %% Compact+ %d %% SUV, o/od times)',nl,p2*100,p3*100,n2); str44=textwrap( { str4}, 90); title( str44 ); xlabel('Time of Day'); ylabel('Monte Carlo Simulation Time'); zlabel('EV Charging Demand in kW'); 319 USC Viterbi School of Engineering E.3 Model for Residential Area %% Generate Random Number U, Ul, U2, U3 follows three pdf prompt= 'Please enter the number ofEVs'; nl = input(prompt); % nl =# of Compact EVs to run Monte Carlo prompt2 = 'Please enter Monte Carlo Simulation times>>'; n2 = input(prompt2); L2_P = 3.3; % Level 2 charging power= 3.3 kW bat= 24; % battery size of Compact EV= 24 kWh mg(' shuffle'); rvl_CST = ones(nl,n2)*-l; rvl_SSOC = ones(nl,n2)*-l ; wl CE =0.6; w2 CE= 0.4; mul_CST = 21; % meanl_CST =bl in Gaussian pdf stdl_CST = 2; % standard deviation 1 CST= cl/sqrt(2) in Gaussian pdf wl_CST = l; % weightl of CST= coefl/sum(coefl +coef2+coef3); SSOC_al = 0.2244;% (-0.4808, 0.9296) SSOC_bl = 219;% (52.12, 385.8) SSOC cl = 87.01;% (36.52, 137.5) mul_SSOC = 50; % meanl_SSOC =bl in Gaussian pdf stdl_SSOC = 30; % standard deviation 1 SSOC = cl/sqrt(2) in Gaussian pdf wl_SSOC = l ; % weight! ofSSOC = coefl/sum(coefl+coef2+coef3); for j=l:n2 for i=l:nl PhD Thesis ul = rand; % ul =random variable to check how many times the EV is charging today u2 = rand; % u2=random variable follows CST pdf u4 = rand; % u4=random variable follows SSOC pdf iful <= wl_ CE % if ul > probability of 0 charging eventtoday, (or CE >= 1 ), generate CSTl, Dl, SSOCl end J end end while rvl_CST(i,j) <= 0 II rvl_CST(i,j) > 27 rvl_CST(i,j) = stdl_CST*randn + mul_CST; if rvl_ CST(i,j ) >= 23.5 rvl_CST(i,j) = rvl_CST(i,j)- 23.5; end end while rvl_SSOC(i,j) <= 0 II rvl_SSOC(i,j) > 100 rvl_SSOC(i,j) = stdl_SSOC*randn + mul_SSOC; end Dl_actual = ones(nl,n2)*-l; EVl_demand = zeros(nl,n2); for j=l:n2 for i=l:nl 320 USC Viterbi School of Engineering if rv l _ S SOC( i,j )>O EVl_demand(i,j) = min(bat, (100-rvl_SSOC(i,j))/lOO*bat); Dl_actual(i,j) = EV1_demand(i,j)/L2_F; end j+n2 end end rv _ D _Hr = zeros(9,1 ); for i=0:0.5:4 rv _ D _ Hr(i *2+ 1 )=i; end D _actual_freq = zeros(9,n2); for j=l:n2 for i=l:length(Dl_actual) for k=1:9 end end j+2*n2 end %% figure subplot(3, 1, 1) ifDl_actual(i,j)>rv _D _Hr(k) && Dl_actual(i,j)<=rv _D _Hr(k)+0.5 D _actual_freq(k,j) = D _actual_freq(k,j)+ 1; end hist(rv l _ CST,48); set(gca,'xtick',0:23 ); xlim([O 23]); xlabel('Time of Day'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start Time Variables') subplot(3,1,2) % hist(D_actual); % set(gca,'xtick',0:5); % xlim([0.1 5]); bar(rv _D _Hr,D _actual_freq); xlabel('Charging Duration'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Duration Variables') subplot(3,1,3) hist(rvl _ SSOC, 100); xlim([O 100]); xlabel('Battery SOC (% )'); ylabel('Frequency'); title('Histogram of Monte Carlo Generated Charging Start SOC Variables') PhD Thesis %% Convert EV Demand Forecast into 24-hr Format and Plot, then compare with Actual Weekday EV Demand (05/20115-02/04/17) EV _Demand_24hr = zeros(n2,24); EV1_Demand_72hr = zeros(n2,72); EV _Demand_24hr_avg = zeros(l,24); for k=l :n2 321 USC Viterbi School of Engineering PhD Thesis for j=l:nl for i=l :72 iffloor(rvl_CST(j,k)) == floor(rvl _CST(j,k) + Dl_actual(j,k)) && floor(rvl_CST(j,k)) == i- 1 % if Charging Start Hr = Charging End Hr && Charging Start Hr = i-1 EV1_Demand_72hr(k,i) = EV1_Demand_72hr(k,i) + EVl_demand(j,k); % assign all EV demand to this hour slot else iffloor(rv l _ CST(j,k)+ Dl _ actual(j,k) )>floor(rv l _ CST(j,k)) % else ifCEhr (charging ending hr) > CShr (charging start hr) if floor(rvl_CST(j ,k)) == i-1 % ifCShr = (i-l)th hour EV1_Demand_72hr(k,i) = EVl_Demand_72hr(k,i) + (ceil(rvl_CST(j,k)) rvl_CST(j,k))*EVl_demand(j,k)/Dl_actual(j,k); % EV demand at this hour= (ceil(CST) CST)*Demand/Duration end if floor(rvl_CST(j ,k)+Dl_actual(j,k)) == i-1 % ifCEhr = (i-l)th hour EV1_Demand_72hr(k,i) = EVl_Demand_72hr(k,i) + (rv 1 _ CST(j,k)+ Dl _ actual(j,k)-floor(rv 1_CST(j,k)+D1 _ actual(j,k)) )*EVl _ demand(j,k)/D 1 _ actual(j,k); % EV demand at this hour = (CST +D-CEhr)*Demand/Duration end if i-1 > floor(rvl_CST(j,k)) && i-1 < floor(rvl _CST(j,k)+Dl_actual(j,k)) % if CShr < (i-1 )th hour < CEhr EV1_Demand_72hr(k,i) = EV1_Demand_72hr(k,i) + EVl_demand(j,k)/Dl_actual(j,k); % EV demand at this hour = Demand/Duration , it means this whole hour will be fully occupied end end k end for k=l:n2 end for i=1:24 end end EV _Demand_24hr(k,i) = EVl_Demand_72hr(k,i) + EVl_Demand_72hr(k,i+24) + EVl_Demand_72hr(k,i+48); % convert charging start at late night and ends tomorrow into 24 hour format end end for i=l:24 EV _Demand_24hr_avg(l ,i) = mean(EV _Demand_24hr(:,i)); end %% figure plot(O: 1:23, EV _Demand_24hr); xlabel('Time of Day'); ylabel('EV Charging Demand in kW' ); strl = sprintf('Residential EV Charging (Level 2) Demand Forecast Based on Monte Carlo Simulation (o/od EVs, %d times)',nl, n2); title( str 1 ); set(gca, 'xtick', 0:23 ); figure % plot surf of EV_ Demand_ 24hr matrix 322 USC Viterbi hr= [O: 1:23)'; MC_num = [l:l:n2]'; surf(hr, MC_num, EV _Demand_24hr); color bar School of Engineering PhD Thesis str3 = sprintf('Surf Plot of Residential EV Charging Demand Forecast Based on Monte Carlo Simulation (o/od EVs, o/od times)', nl, n2); title( str3 ); xlabel('Time of Day'); ylabel('Monte Carlo Simulation Time'); zlabel('EV Charging Demand in kW' ); hrq = 0:0.1 :23; MC_numq = l:l:n2; [hrq, MC_numq] = meshgrid(hrq, MC_numq); EV_ Demand_ 24hrq = interp2(hr, MC_ num, EV_ Demand_ 24hr, hrq, MC_ numq, 'spline'); figure surf(hrq, MC_numq, EV _Demand_24hrq); color bar str4 = sprintf('Surf Plot with Spline of Residential EV Charging Demand Forecast Based on Monte Carlo Simulation (o/od EVs, o/od times)', nl, n2); title( str4 ); xlabel('Time of Day'); ylabel('Monte Carlo Simulation Time'); zlabel('EV Charging Demand in kW' ); 323

Abstract (if available)

Abstract With the growing penetration of the Electric Vehicles (EV) to our daily transportation needs owing to their economic and environmental benefits, there will be both opportunities and challenges to the electric power utilities when adopting plug-in electric vehicles to the distribution network. ❧ The purpose of this thesis is to investigate the impact of electric vehicle battery charging on grid demand and steady state parameters of distribution network in the Los Angeles Department of Water and Power (LADWP) service territory. The research work was conducted under American Recovery and Reinvestment Act (ARRA) Smart Grid Regional Demonstration Project (Federal Grant Number DE-OE0000192) and funded by the US Department of Energy and LADWP. ❧ In order to evaluate the impact of EV infrastructure charging on the distribution grid, three real-world distribution systems have been selected and tested via diverse EV penetration levels and system loading conditions. The designed cases have been tested and verified by two specialized power system analysis software: EDD and OpenDSS. Moreover, various positive and negative effects on the distribution network have been analyzed. ❧ The following part of this thesis examines the possibility of using EVs to provide support to the grid, instead of negative impact discussed in the first part. Specifically, distribution effects of battery aggregation and backfill coupled with renewable energy has been studied. Various EV charging/discharging models, based on the concept of vehicle-to-grid (V2G), are designed and tested on two distribution networks. An integrated algorithm that incorporate EV, PV and storage system is developed and verified to achieve the goal of shaving the peak and filling the valley. ❧ In the next two parts of this thesis, EV charging patterns, power usage of charging stations, and user behaviors have been analyzed based on one year real-world data. The unique data sources and analysis are important for LADWP and other similar utility companies. The results could be used to provide valuable information and practical insights for the utility companies. ❧ In the last part of this thesis, an EV charging demand estimation and forecast model based on Monte Carlo simulation technique has been developed. The model is based on historical charging record and the analysis mentioned above, and can be used to forecast future EV demand for the three customer classes: Industrial, Commercial and Residential.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Utility – customer interface strategy and user application for electric vehicles and distributed generation management

PDF

Distribution system reliability analysis for smart grid applications

PDF

The dynamic interaction of synchronous condensers, SVC, STATCOM and superconducting magnetic energy storage on electric vehicles

PDF

System stability effect of large scale of EV and renewable energy deployment

PDF

An adaptive distance protection scheme for power systems with SFCL

PDF

Reinforcement learning in hybrid electric vehicles (HEVs) / electric vehicles (EVs)

PDF

USC microgrid assessment study

PDF

Curvature analysis of the power grid: congestion, congestion management, and line tripping cascade anticipation

PDF

Environmental effects from a large-scale adoption of electric vehicle technology in the City of Los Angeles

PDF

USC microgrid development conceptual plan

PDF

Developing frameworks to quantify the operational and environmental performance of energy systems within the context of climate change

PDF

A function-based methodology for evaluating resilience in smart grids

PDF

Integration of energy-efficient infrastructures and policies in smart grid

PDF

Predictive real-time assessment for power grid monitoring and situational awareness using synchrophasors with partial PMU observability

PDF

The smart grid network: pricing, markets and incentives

PDF

High-performance distributed computing techniques for wireless IoT and connected vehicle systems

PDF

Electric vehicles & charging stations: Los Angeles County’s roads readiness for California’s transportation electrification

PDF

A joint framework of design, control, and applications of energy generation and energy storage systems

PDF

Discrete optimization for supply demand matching in smart grids

PDF

Architectures and algorithms of charge management and thermal control for energy storage systems and mobile devices

Asset Metadata

Creator Jiang, Zeming (author)

Core Title Electric vehicle integration into the distribution grid: impact, control and forecast

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publication Date 07/24/2017

Defense Date 05/01/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag charging pattern,electric vehicle,grid impact,Monte Carlo simulation,OAI-PMH Harvest,power system,smart grid,user behavior analysis,vehicle to grid

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Beshir, Mohammed (committee chair), Jonckheere, Edmond (committee chair), Maby, Edward (committee member), Meshkati, Najmedin (committee member)

Creator Email kevinzmjiang@gmail.com,zemingji@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-413914

Unique identifier UC11213617

Identifier etd-JiangZemin-5621.pdf (filename),usctheses-c40-413914 (legacy record id)

Legacy Identifier etd-JiangZemin-5621.pdf

Dmrecord 413914

Document Type Dissertation

Format application/pdf (imt)

Rights Jiang, Zeming

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA