University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ENVIRONMENTAL EFFECTS FROM A LARGE-SCALE ADOPTION OF ELECTRIC VEHICLE TECHNOLOGY IN THE CITY OF LOS ANGELES by Jae Duk Kim 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 (INDUSTRIAL AND SYSTEMS ENGINEERING) June 2014 Copyright 2014 Jae Duk Kim 1 Table of Contents 1 INTRODUCTION ............................................................................................................... 10 1.1 Transportation Sector Emissions ................................................................................ 10 1.2 Electric Vehicles to the Rescue? .................................................................................. 11 1.3 Problem Statement and Dissertation Objective ........................................................ 12 1.4 Research Contribution ................................................................................................. 15 2 LITERATURE REVIEW ................................................................................................... 16 2.1 Types of Emissions ....................................................................................................... 16 2.2 Types of Electric Vehicles ............................................................................................ 17 2.3 Vehicle Emissions Models ............................................................................................ 18 2.4 Life-Cycle Analysis of Grid Energy ............................................................................ 20 2.5 Technology Diffusion Models ...................................................................................... 25 2.5.1 Bass innovation diffusion model ............................................................................ 26 2.5.2 Adopter categories in innovation diffusion models ................................................ 28 2.5.3 Bass parameter estimation methods ........................................................................ 30 2.5.4 Extensions to Bass .................................................................................................. 33 2.6 Electric Vehicle Charging Profiles.............................................................................. 35 2.7 Models Projecting Future Transportation Emissions............................................... 38 3 SYSTEM MODELING ....................................................................................................... 43 3.1 Modeling the Supply Side: LADWP’s Energy Grid ................................................. 44 3.1.1 LADWP’s generation sources................................................................................. 44 3.1.2 LADWP’s system load profile ................................................................................ 45 3.1.3 Renewable generation sources ................................................................................ 46 3.1.4 Non-renewable generation sources ......................................................................... 47 3.1.5 Modeling LADWP’s dispatching of generation sources ........................................ 47 3.1.6 Marginal carbon intensity of grid energy................................................................ 49 3.2 Modeling the Demand Side: EV Charging Scenarios ............................................... 51 3.2.1 Los Angeles travel pattern and charging scenarios................................................. 52 3.2.2 EV electric consumption rate .................................................................................. 53 3.2.3 Charge power .......................................................................................................... 54 3.3 Modeling the EV Fleet Size ......................................................................................... 55 2 3.3.1 Vehicle sales projection .......................................................................................... 55 3.3.2 EV sales forecast based on Bass diffusion model ................................................... 59 3.3.3 Vehicle survival rate ............................................................................................... 60 3.3.4 Los Angeles vehicle fleet projection....................................................................... 62 3.4 GHG Emissions of Gasoline Vehicles ......................................................................... 63 3.4.1 Fuel efficiency ........................................................................................................ 64 3.4.2 Fuel carbon intensity ............................................................................................... 64 4 RESULTS AND DISCUSSION .......................................................................................... 65 4.1 EV Energy Loads in Year 2020 ................................................................................... 65 4.1.1 Daily EV energy load.............................................................................................. 65 4.1.2 System energy load profile ..................................................................................... 66 4.1.3 EV-related GHG emissions..................................................................................... 68 4.2 EV energy load requirement scenarios in year 2030 ................................................ 70 4.2.1 Daily EV energy load.............................................................................................. 70 4.2.2 System energy load profile ..................................................................................... 71 4.2.3 EV-related GHG emissions..................................................................................... 74 4.3 Discussion on EV Energy Loads and GHG Emissions ............................................. 76 4.4 Emissions Mitigation Potential of EV Policy Levers................................................. 80 4.4.1 Baseline case ........................................................................................................... 81 4.4.2 EV technology adoption rate .................................................................................. 83 4.4.3 Renewable energy adoption .................................................................................... 85 4.4.4 EV charging behavior ............................................................................................. 86 4.4.5 Policy Implications ................................................................................................. 88 5 SUMMARY AND CONCLUSION .................................................................................... 89 6 DIRECTIONS FOR FUTURE RESEARCH .................................................................... 92 7 REFERENCES .................................................................................................................... 94 8 APPENDIX......................................................................................................................... 105 8.1 MATLAB Scripts ....................................................................................................... 105 8.2 MATLAB Functions .................................................................................................. 125 3 List of Figures and Tables Figure 1. Comparison of GHG emissions between U.S. and California (Source: EPA, 2010a; CARB, 2010) ................................................................................................................................ 11 Figure 2. U.S. electricity generation by fuel type (Source: EPA, 2011) ....................................... 22 Figure 3. Adoptions due to external and internal influences in the Bass model (Mahajan et al. 1990b) ........................................................................................................................................... 27 Figure 4. Rogers's technology diffusion curve (adapted from Rogers 1995) ............................... 28 Figure 5. Adopter categories based on Bass model (revised from Mahajan et al. 1990a)............ 29 Figure 6. Typical daily California system load profile ................................................................. 36 Figure 7. Model overview of estimating hourly GHG emissions from EV charging in LA ........ 43 Figure 8. LADWP's seasonal hourly energy load profile for the year 2011 (FERC, 2013) ......... 46 Figure 9. Overview of resource dispatching model ...................................................................... 49 Figure 10. Trip distribution by time of the day in Los Angeles (Fehr, 2009) .............................. 53 Figure 11. Overview of LA's EV fleet projection model .............................................................. 55 Figure 12. LA's projected population using linear regression ...................................................... 56 Figure 13. Projected working age population in LA and Orange County .................................... 57 Figure 14. Estimated vehicle sales in LA based on County sales data ......................................... 58 Figure 15. EV sales scenarios in Los Angeles based on Bass model ........................................... 60 Figure 16. Vehicle survival rate of two different models (NHTSA, 2006; Davis et al., 2012) .... 62 Figure 17. EV fleet size with respect to total vehicle fleet in Los Angeles .................................. 63 Figure 18. Overview of the calculation for ICEV GHG emissions per day ................................. 64 Figure 19. Hourly energy load requirement from EV charging in LA in 2020 ............................ 66 Figure 20. LADWP’s hourly energy load profile during summer months for Case 1 (off-peak) 67 Figure 21. LADWP's hourly energy load profile during summer months for Case 2 (controlled peak) .............................................................................................................................................. 67 Figure 22. LADWP's hourly energy load profile during summer months for Case 3 (uncontrolled) ................................................................................................................................ 68 Figure 23. LADWP's marginal carbon intensity for Case 1 charging (off-peak) ......................... 69 Figure 24. LADWP's marginal carbon intensity for Case 2 charging (controlled peak) .............. 69 Figure 25. LADWP’s marginal carbon intensity for Case 3 and baseline EV adoption (uncontrolled) ................................................................................................................................ 70 Figure 26. Hourly energy load requirement from EV charging in LA in 2030 ............................ 71 Figure 27. LADWP's hourly energy load profile during summer months for Case 1 (off-peak) . 73 Figure 28. LADWP's hourly energy load profile during summer months for Case 2 (controlled peak) .............................................................................................................................................. 73 Figure 29. LADWP's hourly energy load profile during summer months for Case 3 (uncontrolled) ................................................................................................................................ 74 Figure 30. LADWP's marginal carbon intensity for Case 1 charging (off-peak) ......................... 75 Figure 31. LADWP's marginal carbon intensity for Case 2 charging (controlled peak) .............. 75 4 Figure 32. LADWP’s marginal carbon intensity for Case 3 and baseline EV adoption (uncontrolled) ................................................................................................................................ 76 Figure 33. System load scenarios in 2020 with respect to generation sources in general dispatching order ........................................................................................................................... 79 Figure 34. System load scenarios in 2030 with respect to generation sources in general dispatching order ........................................................................................................................... 80 Figure 35. Average seasonal marginal carbon intensity in year 2020 .......................................... 82 Figure 36. Average seasonal marginal carbon intensity in year 2030 .......................................... 83 Figure 37. LA's EV fleet GHG emissions per day and potential mitigation of ICEV GHG emissions per day in 2020 ............................................................................................................. 84 Figure 38. LA's EV fleet GHG emissions per day and potential mitigation of ICEV GHG emissions per day in 2030 ............................................................................................................. 85 Figure 39. Change in GHG emissions mitigation potential from 2020 to 2030 ........................... 86 Figure 40. GHG emissions per EV based on start time of charging event in 2020 ...................... 87 Figure 41. GHG emissions per EV based on start time of charging event in 2030 ...................... 88 Figure 42. An overview of the new energy system ...................................................................... 92 Table 1. Energy Portfolio of LAWDP in 2008 (Source: LADWP, 2010) .................................... 13 Table 2. Global warming potentials (GWP) over 100 year time horizon (Source: IPCC, 2003) . 16 Table 3. Types of electric vehicles ............................................................................................... 17 Table 4. Overview of EMFAC and MOVES (Source: CARB, 2011; EPA, 2012b) .................... 19 Table 5. Electricity emission rates (lbs/MWh) (Source: EIA, 2007; EPA, 2011) ........................ 21 Table 6. Lifecycle GHG emissions of different energy sources (kg/MWh) ................................. 24 Table 7. Adopter category percentages in Rogers's diffusion of innovation (Rogers, 1995) ....... 28 Table 8. Analytical expression for the adopter categories (revised from Mahajan et al., 1990a) 30 Table 9. Major published works related to emission mitigation potential of transportation technologies .................................................................................................................................. 39 Table 10. Overview of past works on future transportation emissions from electric-propulsion vehicle adoption ............................................................................................................................ 41 Table 11. List of LADWP's generation sources and net dependable capacity in 2012 (LADWP, 2012) ............................................................................................................................................. 45 Table 12. Lifecycle emission rates of power generation plants used in the study ........................ 51 Table 13. Summary of charging scenarios .................................................................................... 53 Table 14. Specifications of electric vehicle chargers (SAE, 2011) .............................................. 54 Table 15. Projected population changes and associated new car sales for Los Angeles .............. 58 Table 16. Model parameters for three adoption cases .................................................................. 59 Table 17. Daily GHG emissions from ICEVs in LA in year 2020 and 2030 ............................... 81 5 List of Acronyms and Abbreviations BEV – Battery electric vehicle CAFE – Corporate average fleet economy CARB – California Air Resources Board CCS – Carbon capture and storage CD – Compact disk CEC – California Energy Commission CH 4 – Methane CNG – Compressed natural gas CO 2 – Carbon dioxide CO 2eq – Carbon dioxide equivalent DOE – Department of Energy EFD – Emissions factor database EIA – Energy Information Administration EMFAC – EMission FACtors EPA – Environmental Protection Agency ERCOT – Electric Reliability Council of Texas EV – Electric vehicle FCHV – Fuel-cell hydrogen vehicle GHG – Greenhouse gases GREET - Greenhouse gases, Regulated Emissions, and Energy use in Transportation GWP – Global warming potential HV – Hybrid vehicle IC – Internal combustion ICEV – Internal combustion engine vehicle 6 IPCC – Intergovernmental Panel on Climate Change ISO – Independent System Operators kWh – Kilo-Watt-hour LA – City of Los Angeles LADWP – Los Angeles Department of Water and Power LCA – Life-cycle Assessment LCI – Life-cycle Inventory LDV – Light-duty vehicle LEVERS – Long-term Evaluation of Vehicle Emissions Reduction Strategies LPG – Liquid propane gas MARKAL – MARKet ALocation MLE – Maximum likelihood estimation MOVES – Motor Vehicle Emissions Simulator MWh – Mega-Watt-hour N 2O – Nitrous oxide NGCT – Natural gas combustion-turbine NGST – Natural gas steam turbine NLLS – Nonlinear least squares NOx – Oxides of nitrogen NREL – National Renewable Energy Laboratory OLS – Ordinary least squares Pb – Lead PHEV – Plug-in electric vehicle PM – Particulate matter ROG – Reactive organic gas 7 RTO – Regional Transmission Organization SHO – Source hours operating SOx – Oxides of sulfur TOG – Total organic gas THC – Total hydrocarbons UNFCCC – United Nations Framework Convention on Climate Change VMT – vehicle miles traveled 8 ABSTRACT Greenhouse gas (GHG) emissions reduction has become an important component of policy decisions on transportation systems design, research and development, and implementation. Particularly in major urban centers, increasing use of electric vehicles (EVs) is being encouraged to support the overall objective of reduction in transportation emissions. This encouragement ranges from consumer tax credits to research and development grants for advanced EV technologies. Assessing the net effect EVs on actual emission mitigation potential, however, depends on three main factors: 1) energy portfolio of power providers, 2) consumer adoption rate, and 3) battery charging patterns. Unfortunately, the current U.S. energy grid is predominantly composed of coal-fired plants that emit high concentrations of GHGs. Therefore, EVs essentially push emissions upstream to the electricity generation sources. EVs represent a dramatic paradigm shift in transportation such that forecasting their adoption requires adaptations to the innovation diffusion models. The charging patterns also affect the emission mitigation potential because the use of “peak” versus “off-peak” power changes the grid energy emissions significantly. This study seeks to quantify the emissions mitigation potential of these three main influencing factors. In order to answer the main research questions, an integrated emissions model is developed for the City of Los Angeles. The model incorporates modules such as changes in population and mobility patterns, consumer technology adoption, vehicle charging patterns, and lifecycle emissions of GHGs from electricity generation. Some of the model’s main outputs are the daily EV energy loads, daily system load profile, hourly average marginal grid energy carbon intensity, and the types of energy generation dispatched at every hour. For 2020, model results show that the EV charging loads will be modest with negligible effects on the overall system load profile. Results indicate that high EV adoption results in greater emissions 9 mitigation potential. However, the type of charging has a significant impact on the scale of mitigation at all levels of adoption. Contrary to previous study results, the average marginal carbon intensity is higher if EV charging occurs during off-peak hours. These results demonstrate that the charging decision in terms of the time of day matters in GHG emissions mitigation efforts. The short-term incentives for off-peak charging may not only result in greater emissions but also deter EV technology adoption which would lower the overall emissions mitigation potential. Encouraging restrictive charging behavior in the short-run may be counterproductive to GHG emissions reduction policies. Model results for 2030 show that EV charging loads increase significantly resulting in potential generation shortages. There are also significant grid operation challenges as the region’s energy grid is required to ramp up and down rapidly to meet EV loads. For 2030, the average marginal carbon intensity for off-peak charging becomes lower than peak charging. Increasing use of renewable generation sources leads to greater GHG emissions mitigation but the greatest effect arises from the removal of coal generation sources. The study concludes with remarks on further research into the optimal distribution of renewable energy and EV-grid interactions as major research areas to enhance understanding of the EV’s effectiveness as GHG emissions mitigating technology. 10 1 INTRODUCTION 1.1 Transportation Sector Emissions In 2009, the U.S. transportation sector accounted for more than 1,700 million metric tons of greenhouse gas (GHG) emissions that contribute to global warming (EPA, 2012a). This constitutes 27% of annual U.S. GHG emissions and ranks second only to electricity generation (EPA, 2012a). The sector’s high emissions are directly related to its heavy dependency on fossil fuels. It constitutes about 60% of all U.S. fossil fuel consumption or 4.9 billion barrels in 2009 (EIA, 2012a). Within the U.S. transportation sector, light-duty vehicles (e.g., passenger cars and light-duty trucks) are the dominant source of GHG emissions accounting for about 65% of the sector’s total (EPA, 2010a). Despite improvements in fuel efficiency and vehicle technologies, emissions from light-duty vehicles increased from 994 to 1185 million metric tons from 1990 to 2008 (EPA, 2010a). Transportation’s emission footprint in California is even higher. In 2008, light-duty vehicles (LDV) accounted for nearly 130 million metric tons of CO 2-equivalent (CO 2e) emissions or 73% of all transportation emissions (CARB, 2010). Since transportation is the largest source of emissions in California (37%), the contribution from light-duty vehicles to the region’s overall emissions is disproportionally higher than the national average as shown in Figure 1 (CARB, 2010; EPA, 2010a). 11 Figure 1. Comparison of GHG emissions between U.S. and California (Source: EPA, 2010a; CARB, 2010) 1.2 Electric Vehicles to the Rescue? Due to the transportation sector’s significant contribution to overall emissions, there has been strong interest in emissions mitigation technologies such as the electric vehicle (EV). With greater economic incentives and technology maturity, a strong market for lower emission vehicles especially the EV is predicted to grow. Adoption of EVs has been frequently identified as an important strategy for reducing both petroleum use and greenhouse gas (GHG) emissions (DOE, 2010; MIT, 2010; Slezak, 2013). In recent years, improved technologies and favorable government policies have helped increase EV adoption. For example, new EV car buyer incentives such as tax credits of up to $7,500 have lowered the upfront costs for adoption (DOE, 2013). This market is being encouraged by a number of regulatory measures. For example, California’s Air Resources Board (CARB) has set a mandate requiring that one out of seven new cars sold (or 1.4 million) must be emission-free by 2025 (CARB, 2012a). Federal regulatory measures such as increasing Corporate Average Fleet Economy (CAFE) standards are changing the policy landscape in support of new EV technology investments among top auto manufacturers. The government’s ambitious 9% 8% 6% 5% 3% 6% 21% 20% 10% 9% 27% 17% 24% 34% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% CALIFORNIA U.S. Electricity Generation Light-Duty Vehicles Other Transportation Industrial Commercial Residential Other 12 goal of putting significant number of EVs on the road quickly (e.g., President Obama’s call for 1 million PEVs by 2015) signals a continuation of these favorable policies. 1.3 Problem Statement and Dissertation Objective Despite the elimination of tailpipe emissions, the net effect of a large-scale adoption of EV technology on GHG emissions is unclear and assessing the full impacts is complex. Emissions are not completely eliminated but rather, the emissions mitigation potential depends on multiple factors. Previous studies have explored a large number of potential factors from battery attributes to consumer behaviors (Axsen et al., 2011; Weiller, 2011; Kelly et al., 2012). Based on literature review, the following are some of the major factors explored in this study. With greater adoption of EVs, the energy demand for battery charging inevitably increases resulting in greater electricity generation and GHG emissions. For this reason, some critics have proclaimed that an EV is actually an EEV or “emissions-elsewhere” vehicle (O’Keefe, 2010). The current U.S. electricity generation process generates 41% of fossil fuel GHG emissions in the U.S. (EPA, 2012a). This “upstream” GHG emissions will grow with increasing charging demands from EVs. Therefore, the net effect of EVs on GHG emissions is directly affected by the type and scale of energy generation resources consumed in the generation process. For Los Angeles (LA), the sole power utility is the Los Angeles Department of Water and Power (LADWP). LADWP’s annual energy generation is composed predominantly of coal and natural gas (see Table 1). Despite aggressive plans for renewable energy adoption by LADWP, coal and natural gas are expected to remain significant in its energy portfolio. Therefore, with greater EV adoption in LA, the increased energy consumption would create greater demand for electricity generation from both nonrenewable and renewable energy sources. Furthermore, the specific emissions intensity for LADWP’s energy portfolio must be calculated. Numerous studies have shown that EVs result in 13 significantly lower GHG emissions based on the assumption that the average electric grid carbon intensity is low and constant (Campbell et al., 2009; Yang et al., 2009; Kim, Rahimi and Newell, 2012). However, this assumption can be problematic because the electricity grid is not homogeneous; each geographic region has varying energy mix comprising of coal, natural gas, and other power generating sources. Consequently, the consumption of electricity in different regions may result in significant differences in the type and amount of emissions. Table 1. Energy Portfolio of LAWDP in 2008 (Source: LADWP, 2010) Nonrenewable: Coal 44% Natural Gas 26% Nuclear* 9% Large Hydroelectric 7% Total: 86% Renewable Wind 6% Small Hydroelectric 5% Solar <1% Geothermal 2% Biomass & Waste 1% Total: 14% *Nuclear is not recognized as an eligible renewable by the California Energy Commission Consumer behavior in terms of how EVs are charged can also change the emissions mitigation potential. That is, consumers may charge their EVs during the day using “peak power” rather than night time using “off-peak power.” The consequence will be a greater demand for generation capacity from “peaker plants,” which are typically gas turbine plants that burn natural gas or diesel fuel. There is growing consensus that the true assessment of EV emission impacts requires the consideration of the marginal grid carbon intensity, which has hourly variance depending on the system load (McCarthy and Yang, 2010; Jansen et al., 2010; Elgowainy et al., 2012). As more EVs connect to the grid for charging, the hourly carbon 14 intensity changes depending on the overall system load. Therefore, a deeper understanding of the effect of the time of charging by EVs on GHG emissions is critical. The rate of EV adoption may also play a significant role. High adoption of EVs represents a paradigm shift in transportation infrastructure. Therefore, the adoption may not be analogous to a simple market forecast for a new vehicle taking market share from existing vehicle types (i.e., gasoline vehicles). Technology diffusion models that characterize the diffusion of innovation may be more compatible to forecast EV’s rate of adoption. Due to the complex dynamics of these major factors, the net effects on GHG emissions resulting from a large-scale adoption of the EV technology are unclear. Therefore, it is critical to quantity the potential effects on emissions in a comprehensive framework. There is a need for a deeper understanding of the complex interlinks between EV adoption, electricity grid mix, and charging behavior, on a localized basis. The overall research objective of this study is to test the following hypothesis: In an electrified transportation sector, consumer behavior in terms of the time-of-day charging is a greater factor to GHG emissions reduction than the rate of technology adoption and renewable energy sources in electricity generation. The approach for this research is to develop an integrated transportation emissions model for GHG for the City of Los Angeles that includes 1) the renewable portfolio of the power supplier, 2) technology adoption rate, and 3) consumer charging patterns. In order to test the above assertion, this study addresses the following key research questions: 1) What are the net effects on GHG emissions from a high adoption of EVs in LA? What are the direct and indirect effects? 2) What is the impact of peak and off-peak charging on emissions? How much does the baseline load need to increase to mitigate potential effects from EV adoption? 15 3) What is the nature and extent of the effects on daily load on the energy grid from the required large-scale increases in power generation capacities? 4) What are some ways to maximize or minimize the effect of EVs on the energy grid? What are the sensitivities of the net GHG emissions mitigation potential? 1.4 Research Contribution This study is designed to assess the potential impacts of a large-scale adoption of EVs in the City of Los Angeles (LA) on grid energy demand and the resultant GHG emissions. Unlike much of California, LA’s electric grid is owned and operated by a single vertically integrated public utility that controls the type and scale of renewable and nonrenewable energy sources and dispatching. LA’s unique car culture also makes early EV adoption disproportionately higher than the rest of the country suggesting that previous results may be significantly underestimating the effects from EV charging. First, a model is created for the estimation of GHG emissions from power generation. This is accomplished by projecting the changes to the region’s hourly GHG emission rate or “grid carbon intensity” based on planned energy generation projects and resource dispatching. A second model is created to forecast the demand side by analyzing the travel patterns, technology attributes, and adoption scenarios. Charging scenarios are created based on local trip generation data. Technological attributes such as the EV battery discharge rate and the charge power are also included in the model to see potential effects on emissions. Another model takes into account the changes in the EV fleet size based on working age population projections, vehicle scrap rates, and the region’s historical vehicle sales data. Here, the assumption is that EV adoption is primarily composed of passenger vehicles (i.e., excludes large trucks, buses, etc.) using recent sales trends indicating changing consumer preferences for smaller vehicles. One important and unique feature of this study is the inclusion of the changes in the renewable portfolio to specific 16 types of power sources, hourly emissions, and vehicle fleet composition. By doing so, the model is able to directly attribute GHG emissions from increasing EV adoption to a specific type and scale of power generation source on an hourly basis. The inclusion of these key factors in the methodology leads to significantly improved results about the net effects of a large scale EV adoption in LA. 2 LITERATURE REVIEW Because of the interdisciplinary nature of this dissertation, this literature review explores multiple research areas related to the current transportation system. The following sections contain reviews of relevant literature on emissions modeling, grid energy, technology diffusion, and electric vehicle charging, which are the main variables included in the systems analysis. 2.1 Types of Emissions Transportation-related GHG emissions includes carbon dioxide (CO 2), methane (CH 4), and nitrous oxide (N 2O) from the combustion of fuel. There is also fluorinated gases (F-gases) from the air conditioning system (Kahn Ribeiro et al., 2007). Each type of GHG has different global warming potential (GWP) with CO 2 having a GWP of 1 (see Table 2 ). Therefore, these emissions are often weighted with GWP factors and cumulated for a single global warming potential number commonly referred to as “CO 2eq.” In this dissertation, the variable “CO 2eq” is used interchangeably with “GHG emissions” as equivalent measures. Table 2. Global warming potentials (GWP) over 100 year time horizon (Source: IPCC, 2003) Greenhouse Gas GWP CO 2 1 N 2O 296 CH 4 23 HFC-125 3,400 HFC-134a 1,300 HFC-143a 4,300 HFC-152a 120 17 2.2 Types of Electric Vehicles There are different types of EVs with varying levels of electrification (see Table 3). A battery-electric vehicle (BEV) has electric motors and battery packs as its only means of propulsion. Typically, their range is less than 100 miles (e.g., Nissan Leaf) but there are ones with ranges exceeding 200 miles (e.g., Tesla Roadster S). A plug-in hybrid vehicle (PHEV) acts like a BEV for a limited range before switching to an internal combustion (IC) engine that either directly drives the wheels or acts like a generator to supply electricity to keep the electric motors running. A typical PHEV can have an electric-only range for the first 40 to 50 miles. With the backup IC engine, a PHEV can have an extended range of 350 to 500 miles. A hybrid vehicle (HV) uses both gasoline and electric drivetrains. HV’s electric motors either assist the IC engine or directly drive the vehicle for short bursts. A fuel-cell hydrogen vehicle (FCHV) is an EV but with a fuel-cell instead of a battery pack and hydrogen as fuel instead of gasoline. In this study, the focus is on BEV and PHEV technologies because future prospects of HV and FCHV are bleak. Therefore, for brevity, the term “electric vehicle” (EV) is used to represent BEVs and PHEVs in electric-mode only. Table 3. Types of electric vehicles Type Description Notable cars (battery pack) Battery- Electric Vehicle (BEV) Pure electric drive powered by battery pack Nissan Leaf (24 kWh) Tesla Model S (90 kWh) Ford Focus Electric (23 kWh) Plug-In Electric Vehicle (PHEV) Partial electric drive for limited range; IC engine as backup generator to drive electric motors or wheels for extended range Chevy Volt (16 kWh) Toyota Prius Plug-In (4.4 kWh) Fisker Karma (20 kWh) Hybrid Vehicle (HV) IC engine is the primary with electric motors assisting or providing short bursts of electric-only drive modes Toyota Prius (1.3 kWh) Ford C-Max (1.4 kWh) Fuel-Cell Hydrogen Vehicle (FCHV) Hydrogen as fuel to drive electric motors Chevy Equinox Honda FCX Clarity Hyundai Tucson FCEV 18 2.3 Vehicle Emissions Models The detrimental impacts from transportation-related emissions have prompted the development of emissions modeling tools. Different models have been built upon decades of work by researchers to estimate the emissions of pollutants from transportation activities. From those efforts, different vehicle emissions models have emerged that have been further modified and strengthened. Although no model is perfect, each approach has its advantages and disadvantages as well as gaps and missing data sets. Currently, there are three major modeling tools used by researchers and policymakers to assess vehicle emissions. Argonne National Laboratory created the Greenhouse gases, Regulated Emissions, and Energy use in Transportation (GREET) model that estimates the life-cycle energy use and emissions associated with various transportation fuels and advanced vehicle technologies. Many studies have used GREET to estimate the “well-to-wheels” emissions and energy use for different types of fuels (Wang, 2002; Wu et al., 2008; Plevin, 2009). The U.S. Environmental Protection Agency (EPA) has created the Motor Vehicle Emissions Simulator (MOVES) to estimate emissions at the national and sub-regional levels. The California Air Resources Board (CARB) has created the EMFAC tool that estimates emissions at the state, county, and sub-county levels. MOVES and EMFAC are today’s de facto vehicle emissions modeling tools for emissions estimations for regulation compliance and other agency requirements. These two models, however, both have significant deficiencies and gaps that limit their effectiveness in answering important research questions. MOVES can be used to estimate air pollution emissions from cars, trucks, motorcycles, and buses including GHGs as well as associated energy consumption. MOVES also incorporates the “upstream” emissions and energy consumption required for the production of each fuel type. 19 EMFAC is a regional model (i.e., California) that provides the most of the outputs as those from MOVES. Both models estimate outputs based on estimates of vehicle activities, emission rates, and various adjustment factors depending on the type of vehicle, fuel type, and other parameters. Table 4 highlights some of the similarities and differences between MOVES and EMFAC. Although both MOVES and EMFAC does account for PHEVs in the modeling, these tools do not integrate the geospatial electricity generation information. These models also ignore or make simplistic assumptions about the potential changes in electric battery technology in future projections. Furthermore, these models do not currently have a way to incorporate the EV charging patterns into the emissions calculations. GREET takes a step further by incorporating the average U.S. grid energy mix as the emissions factor for electric vehicles but such broad generalization does not represent the realities especially in LA. Table 4. Overview of EMFAC and MOVES (Source: CARB, 2011; EPA, 2012b) EMFAC 2011 MOVES 2010a Geographical area  State (California)  Air Basin  District  County  Nationwide  State  County  Link (road type) Road  N/A  Rural roadways with options  Urban roadways with options  Off-network Temporal scale  Analysis year: 1990-2035  Season: summer/winter/annual  Month: each month  Analysis year: 1990, 1999- 2050  Month: each month  Day: weekdays or weekends  Hour: each hour of the day Vehicle model  1965-2040  1960-2050 Vehicle class  Passenger cars  Light-duty trucks  Medium-duty trucks  Light-heavy-duty trucks  Medium-heavy-duty trucks  Passenger cars  Passenger trucks  Light commercial trucks  Single unit short-haul trucks  Single unit long-haul trucks  Combination short-haul trucks 20  Heavy-heavy-duty trucks  Urban buses  School buses  Other buses  Motorcycles  Motor homes  Combination long-haul trucks  Intercity buses  Transit buses  School buses  Motorcycles  Motor homes Fuel type  Gasoline  Diesel  Electricity  Gasoline  Diesel  Ethanol  Methanol  Compressed natural gas (CNG)  Liquid propane gas (LPG)  Gaseous hydrogen  Liquid hydrogen  Electricity Pollutant  Hydrocarbons (TOG, ROG, THC and CH 4)  Carbon monoxide (CO)  Nitrogen oxides (NOx)  Sulfur oxides (SOx)  Lead (Pb)  Particulate matter (PM30, PM10, PM2.5)  Methane (CH 4)  Nitrous oxide (N 2O)  Atmospheric carbon dioxide (CO 2) Primary activity measurement  Vehicle miles traveled (VMT)  Vehicle operating time (SHO – Source Hours Operating) Grid energy specification  N/A  N/A Electric battery projections  N/A  N/A EV charging profiles  N/A  N/A 2.4 Life-Cycle Analysis of Grid Energy The emissions rate of grid electricity is dependent on the source and type of the electricity generation process. For this reason, it is common in emissions inventory reports to utilize average emissions factors (or grid emission intensity) on a “per-kilowatt-hour” (kWh) basis. These factors represent an aggregate estimate of emissions from a broad set of electricity generation processes. Unfortunately, there is no standard protocol for accounting and calculating these emissions factors, 21 resulting in an assortment of methodologies and ranges of values. For example, the United Nations (UNFCCC) has established its own framework to calculate emission factors from grid electricity and has created the Emissions Factor Database (EFD) as a library of emission factors (UNFCCC, 2009). Several other studies have attempted to offer different protocols (see Marnay et al., 2002; BSI, 2008), but high degrees of uncertainty exist on the proper geographical and temporal scales in calculating these emission factors. Governmental agencies such as EPA and EIA also publish emission rates for grid regions and states (see Table 5). There are significant differences in the emission rates because of all of the reasons mentioned above as well as the time lapse between each agency’s publications. Table 5. Electricity emission rates (lbs/MWh) (Source: EIA, 2007; EPA, 2011) California (lbs/MWh) U.S. (lbs/MWh) EIA (2002) 774 1,498 EPA eGrid (2007) 568 1,300 In order to assess EV adoption’s effects on emissions from greater electricity generation, one may argue that a simple calculation using the national U.S. average emission rate (e.g., EPA eGrid and EIA) is sufficient. Many studies make explicit assumptions that the grid emission intensity is homogeneous at the national or state level to assess GHG impacts (Bandivadekar et al., 2008a; Grimes-Casey et al., 2008; Hadley and Tsvetkova, 2009; McCollum and Yang, 2009; Shiau et al., 2009; Sioshansi and Denholm, 2009; Yang et al., 2009). This approach, however, has major drawbacks because of significant regional variances. For example, the average national U.S. grid energy emissions intensity differs significantly from that of any individual state (see Figure 2). State energy portfolios in Indiana, Ohio, and West Virginia are all nearly completely composed of coal. Washington’s energy portfolio, however, is composed mostly of hydro power. The emission intensities across each individual state also diverges greatly because of the diverse energy 22 generation sources for each state, respectively. States are poor boundaries of electricity emissions because grid balancing authorities such as the independent system operators (ISOs) and regional transmission organizations (RTOs) operate across state boundaries. 1 Therefore, the use of average emissions intensity of a specific geographical region for a different region can significantly misrepresent the actual emissions. Weber et al. demonstrated the inaccuracy in the grid energy emission intensity as the region of interest is generalized to a larger region (2010). For the LA region, the use of the U.S. or California average emission intensity would yield an invalid result since the power supplier for LA (i.e., LADWP) has a significantly different energy portfolio (see Table 1). Therefore, the grid energy emission intensity must be calculated with respect to the power provider’s specific energy portfolio. Figure 2. U.S. electricity generation by fuel type (Source: EPA, 2011) 1 Texas is an exception because the Electric Reliability Council of Texas (ERCOT) is located solely within its borders, covers nearly the entire state, and is not synchronously interconnected to the rest of the U.S. 23 One critical factor in grid electricity emissions is the inclusion of the full lifecycle emissions particularly for renewable energy. With greater adoption of renewable energy sources like solar and wind, the calculation of emission rates is becoming even more complex. Many agencies often associate zero emissions for renewable energy such as wind when computing emission rates (e.g., see Kim J.D. et al., 2012). The result is an artificially low estimate of the actual emission rate, which underestimates the full lifecycle emissions from electricity generation. To avoid this inconsistency, a lifecycle perspective must be adopted to account for the full emission impacts from electricity generation because each energy source has varying levels of emissions at different lifecycle stages. For example, coal plants have high emissions during their usage phase while wind energy plants have high emissions during the production phase. Therefore, the lifecycle assessment (LCA) methodology is adopted because it seeks to track environmental impacts throughout the life-cycle, including raw material extraction, production, processing or manufacturing, transportation, distribution, storage, use, and disposal (i.e., lifecycle phases). The standard LCA method consists of sequential steps: definition of goal and functional unit, delimitation of scope or system boundary, lifecycle inventory (LCI), and lifecycle impact assessment (see Keoleian and Spitzley, 2006; Jimenez-Gonzalez, 2000; Curran, 1996). LCI refers to the accounting of pollution and resource extraction in each life-cycle phase. From a LCA perspective, studies differ in emission factors from power generation processes (e.g., see Hondo, 2005; Pacca and Horvath, 2002). Emission factors for power generation from renewable sources also differ significantly based on the type and power capacity of the plant. An extensive literature review of life-cycle studies on electric generation process reveals a wide range of emission rates for each energy source. Table 6 lists some of the results of different studies. The wide range of emission rates can be attributed to uncertainties. More 24 importantly, however, the range of values is a direct result of the differences in the attributes such as power generation capacity and location. Table 6. Lifecycle GHG emissions of different energy sources (kg/MWh) Source (year) Coal Natural Gas Solar PV Nuclear Wind Hydro Fthenakis et al. (2008) - - 20-55 - - - Spath et al. (1999) 1022 - - - - - Spitzley and Keoleian (2005) 968 505 19 - 3 26 Storm van Leeuwen and Smith (2002) - - - 64 - - Bergerson and Lave (2002) 998 580 - - - 8 NAP (2010) 681-1050 245-608 - 15-25 - - IER (1997) 815 362 53 20 7 - UK SDC (2006) 891 356 - 16 - - Meier (2002) 974 469 39 15 14 - Dones et al. (2007) 949 485 79 8 14 3 UK DTI(2006) 755 385 - 11-22 11-37 - Vattenfall (1999) 980 450 50 6 6 3 British Energy (2005) - - - 5 - - Hondo (2005) 975 608 53 24 29 44 Pacca (2007) - - - - - 128-380 The high variability in the emission factors of some of the energy generation sources can be problematic in estimating the GHG emission factor for a specific power plant. The choice of a particular emission factor from a specific LCA study may not correspond to the actual type and capacity of the generation plant. As an attempt to narrow some of variability in emission factors, the National Renewable Energy Laboratory (NREL) has conducted an extensive review of published LCA studies on grid energy then “harmonized” the results (Burkhardt et al., 2012; Dolan and Heath, 2012; Hsu et al., 2012; Kim et al., 2012; Warner and Heath, 2012; Whitaker et al., 2012; O’Donoughue et al. 2014). These studies were able to identify some of the causes of high variability in the estimates of lifecycle GHG emissions of electricity generation. In this dissertation, some of emission factors from the LCA harmonization results are utilized for specific energy generation sources. 25 2.5 Technology Diffusion Models Mass adoption of the EV has not occurred at an accelerated pace since its introduction. Many have proclaimed its premature death repeatedly (see Who Killed the Electric Car?, 2006). EVs, however, have made a comeback and are expected to gain significant market share aided by favorable government policies (see DOE, 2013; CARB, 2014). Many emissions studies have been conducted on an explicit assumption that EV adoption will become increasingly significant (Bandivadekar et al., 2008a; Samaras and Meisterling, 2008). In these studies, the EV market projections, however, are fairly simplistic often with an arbitrary market share percentage or a fixed number of vehicles (e.g., 1 million EVs) without much justification. Therefore, the resulting emission forecasts may be greatly overestimated or underestimated for any given time forecast. Technology diffusion theory has been previously applied to numerous technology innovations. 2 It has been successful in predicting the adoption pattern of paradigm changing technologies, thus may serve as an effective methodology for forecasting EV adoption. The most appealing features of diffusion models are their insights into the market penetration and saturation points based on the speed and shape of the adoption curve (Mahajan and Muller, 1979). Multiple diffusion models have been suggested in the strategic decisions for product launches in various industries for different technology innovations (Bass, 1969; Mansfield, 1961; Fourt and Woodlock, 1960; Mahajan and Peterson, 1979). For example, firms such as Eastman Kodak, IBM, Sears, and AT&T have all used various diffusion models for forecasting (Mahajan et al., 1990b). Diffusion models typically have a sigmoid shape (“S-curve”) but parameters such as cost, changing market size, and substitutions can significantly change the spread and time scales (e.g., Kalish, 1985; Kamakura and Balasubramanian, 1988). 2 Although technology “adoption” and “diffusion” may have different technical definitions, the terms are used interchangeably in this study to represent the rate of market penetration and saturation in time and space. 26 The most famous diffusion models are those of Fourt and Woodlock (1960), Mansfield (1961), and Bass (1969). Fourt and Woodlock used consumer panel statistics to create a simple model for predicting the success of new grocery products (1960). Mansfield looked at the spread of twelve different innovations in four different industrial sectors (coal, iron and steel, brewing, and railroad) to create a deterministic model to predict the “rate of imitation” or adoption across the industry of a particular technology innovation (1961). Bass used sales data of eleven different technology consumer durable goods to create a mathematical model to predict the rate of adoption at any given point in time (1969). Since its inception, Bass model has been the de facto framework that subsequent works on technology diffusion have extended upon. This dissertation also extends the fundamental Bass model to predict EV adoption in the emissions model. 2.5.1 Bass innovation diffusion model In his seminal work, Bass formulated a mathematical model capturing the diffusion of technology innovations (1969). Building upon an early work by Rogers (1995), Bass’s model defines the diffusion of innovation as a process in which an innovation is communicated through certain channels in a social system. Bass identified mass media and the “word of mouth” as the two basic communication channels that influence the adoption of an innovation (1969). The first group of adopters is “innovators” who are influenced by the external factors (i.e., mass media) and the second group of adopter is “imitators” who are influenced by the internal factors (i.e., word- of-mouth) (Bass, 1969). The following is the first-purchase Bass model of technology diffusion: 𝑛 (𝑡 ) ∶= 𝑑𝑁 (𝑡 ) 𝑑𝑡 = (𝑝 + 𝑞 𝑚 𝑁 (𝑡 )) [𝑚 − 𝑁 (𝑡 )] (1) 𝑓 (𝑡 ) ∶= 𝑑𝐹 (𝑡 ) 𝑑𝑡 = (𝑝 + 𝑞𝐹 (𝑡 ))(1 − 𝐹 (𝑡 )) (2) 𝐹 (𝑡 ) = 1−𝑒 −( 𝑝 +𝑞 )𝑡 1+ 𝑝 𝑞 𝑒 −( 𝑝 +𝑞 )𝑡 (3) where 27 𝑁 (𝑡 ) ∶= 𝑐𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑎𝑑𝑜𝑝𝑡𝑒𝑟𝑠 𝑎 𝑡 𝑡𝑖𝑚𝑒 𝑡 𝐹 (𝑡 ) ∶= 𝑁 (𝑡 ) 𝑚 , 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑎𝑑𝑜𝑝𝑡𝑒𝑟𝑠 𝑤 ℎ𝑜 𝑎𝑑𝑜𝑝𝑡 𝑡 ℎ𝑒 𝑝𝑟𝑜𝑑𝑢𝑐𝑡 𝑏𝑦 𝑡𝑖𝑚𝑒 𝑡 𝑚 ∶= 𝑚𝑎𝑟𝑘𝑒𝑡 𝑝𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 (𝑐𝑒𝑖𝑙𝑖𝑛𝑔 ) 𝑝 ∶= 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑜𝑓 𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛 𝑞 ∶= 𝑐𝑜𝑒 𝑓 𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑜𝑓 𝑖𝑚𝑖𝑡𝑎𝑡𝑖𝑜𝑛 Equation 1 can be rewritten to reflect the number of adoptions by two groups: 𝑛 (𝑡 ) = 𝑑𝑁 (𝑡 ) 𝑑𝑡 = (𝑝 − 𝑚𝑁 (𝑡 )) + 𝑞 𝑚 𝑁 (𝑡 )[𝑚 − 𝑁 (𝑡 )] (4) The first term in Equation 4, (𝑝 − 𝑚𝑁 (𝑡 )) , represents the adoption from buyers who are not influenced in their timing of purchase by the previous buyers, thus primarily dependent on external influences and proportional to the remaining non-adopters. Bass described the parameter 𝑝 as the “coefficient of innovation.” The second term, ( 𝑞 𝑚 𝑁 (𝑡 )[𝑚 − 𝑁 (𝑡 )]) , represents the adoption from buyers who are influenced by the number of previous buyers, thus primarily dependent on internal influences and proportional to both the number of non-adopters and previous adopters. Bass described the parameter 𝑞 as the “coefficient of imitation.” Figure 3 conceptually shows the Bass model. Figure 3. Adoptions due to external and internal influences in the Bass model (Mahajan et al. 1990b) 28 2.5.2 Adopter categories in innovation diffusion models Diffusion models can be segmented to categorize adopters into specific segments. Rogers suggested a normal distribution curve and classified adopters into five categories – innovators, early adopters, early majority, late majority, and laggards (1995). Figure 4 shows Rogers’s diffusion curve with the adopter categories. Figure 4. Rogers's technology diffusion curve (adapted from Rogers 1995) The two main parameters of the diffusion curve and the adopter categories are the mean time of adoption (t) and standard deviation (σ). These two parameters are used to attribute percentages to each adopter category (see Table 7). This characterization’s main advantage is its simplicity and mutual exclusivity of the adopter groups. Table 7. Adopter category percentages in Rogers's diffusion of innovation (Rogers, 1995) Adopter category % adopters Area covered under normal curve Innovators 2.5 Beyond t - 2σ Early Adopters 13.5 Between t - σ and t - 2σ Early Majority 34.0 Between t and t – σ Late Majority 34.0 Between t and t + σ Laggards 16.0 Beyond t + σ 29 Although Rogers’s normal distribution categorization has advantages, the main drawbacks are that many technology diffusions may not follow a normal distribution and the adopter categories do not necessarily possess fixed percentages (Mahajan and Peterson, 1985). To overcome the shortcomings in Rogers (1995), Mahajan et al. suggested an alternate approach based on the Bass model (1990a). The approach does not assume a normal distribution, rather, the diffusion curve is data-specific (Mahajan et al., 1990a). Therefore, the adopter categories are also based on the data rather than a predefined percentages. The categories are created based on the inflection points of 𝑓 (𝑡 ), ( 𝑑𝑓 (𝑡 ) 𝑑𝑡 ), and ( 𝑑 2 𝑓 𝑑𝑡 2 ) of the original Bass model. The result is a categorization into four groups – early adopters, early majority, late majority, and laggards – defined analytically. Figure 5 shows the distribution and range of percentages for each of the four categories. Figure 5. Adopter categories based on Bass model (revised from Mahajan et al. 1990a) The important result is the analytical categorization of the adopter groups with ranges of percentages for each group. Table 8 shows the analytical expressions for the time intervals and size for each adopter category. 30 Table 8. Analytical expression for the adopter categories (revised from Mahajan et al., 1990a) Adopter Category Time Intervals Adopter Category Size Early Adopters 1 (𝑝 + 𝑞 ) ln [(2 + √3) 𝑝 𝑞 ] 1 2 (1 − 𝑝 𝑞 ) − 1 √12 (1 + 𝑝 𝑞 ) Early Majority 1 (𝑝 + 𝑞 ) ln(2 + √3) 1 √12 (1 + 𝑝 𝑞 ) Late Majority 1 (𝑝 + 𝑞 ) ln(2 + √3) 1 √12 (1 + 𝑝 𝑞 ) Laggards 1 2 (1 + 𝑝 𝑞 ) − 1 √12 (1 + 𝑝 𝑞 ) Unlike Rogers, this result has inconsistent category sizes depending on the Bass parameters (Mahajan et al., 1990a). In general, this categorization also results in smaller sizes for early majority and late majority categories while a larger size for laggards (Mahajan et al., 1990a). 2.5.3 Bass parameter estimation methods The use of the Bass model requires estimating three parameters (𝑚 , 𝑝 , 𝑞 ) so at least three data points are required. There are three time-invariant estimation approaches. Bass originally suggested the ordinary least squares (OLS) approach (1969). Schmittlein and Mahajan suggested the maximum likelihood estimation (MLE) approach (1982). Srinivasan and Mason suggested the nonlinear least squares (NLLS) estimation approach (1986). Bass’s OLS approach is a linear regression using discrete time-series data. Equation 1 is rewritten as the following: 𝑁 (𝑡 + 1) − 𝑁 (𝑡 ) = 𝑝𝑚 + (𝑞 − 𝑝 )𝑁 (𝑡 ) − 𝑞 𝑚 𝑁 2 (𝑡 ) + 𝜀 𝑡 (5a) = 𝛼 1 + 𝛼 2 𝑁 (𝑡 ) + 𝛼 3 𝑁 2 (𝑡 ) + 𝜀 𝑡 (5b) 31 where 𝛼 1 = 𝑝𝑚 , 𝛼 2 = 𝑞 − 𝑝 , 𝛼 3 = − 𝑞 𝑚 ⁄ . Equation 5b can be used to perform regression analysis to find 𝛼 ′𝑠 , which can then be used to find the original Bass parameters. There are, however, three major drawbacks to this approach (Schmittlein and Mahajan, 1982). First, the correlation between the variables in Equation 5b may yield parameter estimates that are unstable or wrong in sign. Second, the method does not provide any standard errors for the parameters so their statistical significance cannot be assessed. Third, because the method relies on discrete time-series data, the resulting continuous model may have time-interval bias, which tends to overestimate (or underestimate) adoption if the adoption occurs sharply (or slowly). To overcome the drawbacks in OLS, Schmittlein and Mahajan suggested the MLE procedure (1982). MLE method uses the solution of the differential equations in the Bass model (Equations 1 to 4) to estimate the parameters. For a sample size 𝑀 , the expected number of adopters can be written as the following (using Equation 3): 𝐸 [𝑁 (𝑡 )] = 𝑐𝑀𝐹 (𝑡 ) = 𝑐𝑀 1−𝑒 −𝑏𝑡 1+𝑎 𝑒 −𝑡 (6) where 𝑎 = 𝑞 𝑝 ⁄ , 𝑏 = 𝑝 + 𝑞 , and 𝑐 is the probability of eventual adoption. The maximum likelihood estimates are generated first for 𝑎 , 𝑏 , and 𝑐 . Then estimates for 𝑞 , 𝑝 , and 𝑚 are generated using the following: 𝑝 ̂= 𝑏 ̂ (𝑎 ̂+1) , 𝑞 ̂ = 𝑎 ̂𝑏 ̂ (𝑎 ̂+1) , 𝑚 ̂ = 𝑐 ̂ 𝑀 (7) The main drawbacks to MLE, however, are the dependency on good initial parameter estimates and its underestimation of the standard errors of the estimated parameters that may lead to wrong conclusions about the parameters’ statistical significance (Srinivasan and Mason, 1986). This disadvantage arises from the MLE method’s dependence on only the sampling errors and its inability to account for other external factors that may significantly influence adoption. 32 Srinivasan and Mason suggested the NLLS method to overcome the drawbacks in MLE (1986). The NLLS method adds an additive error term to the expression that represents the number of adopters in a given interval: 𝑁 (𝑡 ) = 𝑚 [𝐹 (𝑡 ) − 𝐹 (𝑡 − 1)] + 𝑢 𝑡 (8) This additional error term may represent the sampling error, misspecification of the density function (see Easingwood et al., 1983), and excluded variables such as advertising, pricing, and other marketing and competitive effects (Srinivasan and Mason, 1986). Like MLE, however, NLLS also requires initial values. Although studies have identified a good range of parameters for different applications (see Sultan et al., 1990), the sensitivity to initial values can be significant in the two estimation methods. Although these methods have been useful, however, results have shown that the model’s performance is highly sensitive to the number of data points (see Hyman, 1988; Tigert and Farviar, 1981). Studies also show that to obtain robust parameters for Bass models, more data points including the peak at the noncumulative adoption curve are required (Srinivasan and Mason, 1986). Waiting for more data points (e.g., the peak in the noncumulative adoption curve) to obtain robust parameters, however, is counterproductive and unrealistic since the diffusion model attempts to predict the penetration rate and saturation point before any significant market adoption. Therefore, predictions by diffusion models are less useful once significant adoption has already occurred. Such paradox is the nature of diffusion models. To overcome some of the dependency on a large set of data, there are also time-varying procedures for estimating Bass parameters. The original Bass model made an implicit assumption that the parameters remain constant. This assumption, however, may be too simplistic considering the technological and market changes over a product’s lifetime. Incremental improvements in 33 technology, decreasing production costs and prices, improving quality, and changing customer preferences and expectations are just some factors that will likely impact the diffusion parameters over time (Horsky, 1990). For EV adoption, the increasing number of charging stations along with increasing battery quality and decreasing prices will have an effect on the diffusion pattern. The two main time-varying approaches are the Bayesian (Sultan et al., 1990) and adaptive feedback approach (Bretschneider and Mahajan, 1980). These methods are designed to update the parameters as more data becomes available. The basic idea of the adaptive feedback approach is to use the forecast error to update the parameter estimates. 3 The equations for the adaptive filter method are the following: 𝑁 (𝑡 ) = 𝛼 1 + 𝛼 2 𝑁 (𝑡 − 1) + 𝛼 3 𝑁 2 (𝑡 − 1) (9a) 𝛼 𝑖 (𝑡 ) = 𝛼 𝑖 (𝑡 − 1) + 𝐴 𝑖 (𝑒 (𝑡 )) (9b) The term 𝐴 𝑖 (𝑒 (𝑡 )) is the “feedback filter” that is a function of the forecast errors (Bretschneider and Mahajan, 1980). Bretschneider and Mahajan applied the method to three different product innovations with encouraging results (1980). These time-varying methods depend on initial estimates of the parameters before data becomes available. The OLE procedure or a set range of parameters (Sultan et al., 1990) may be used for the initial estimate before utilizing the adaptive feedback method to update the parameters. 2.5.4 Extensions to Bass The original Bass model assumes that the diffusion of an innovation is independent of all other innovation. Such assumption, however, is often invalid because innovations do not occur in a vacuum. Often, technology innovations are adopted more rapidly with another innovation or complementary technologies (e.g., adoption of EVs and superfast charging stations). Bayus 3 Adaptive feedback methods originate from control theory in mechanical and electrical engineering disciplines. There are multiple other feedback algorithms (e.g., Kalman filtering) that may be applied to improve performance. 34 characterized the adoption of contingent products by examining the compact disk (CD) industry (1987). Contingent products refer to the situation where the purchase of a product is contingent on the purchase of a primary product. In the CD industry, the hardware (i.e., CD player) is the primary product that must be purchased before CDs, the contingent product, can be purchased. Bucklin and Sengupta modeled the co-diffusion of complementary innovations by examining the adoption of barcode scanner and universal product code (UPC) symbols (1993). Other underlying assumptions in the Bass model have been relaxed and modified. Jain et al. suggested a model to account for effects from supply constraints and applied it to the adoption of new telephones in Israel (1991). Kalish and Lilien suggested a model to account for changing customer perceptions of the product characteristics by defining the coefficient of imitation as a function of perceived product quality (or reputation) and information level (1986). The model was applied to photovoltaics and yielded better results than Bass (Kalish and Lilien, 1986). EVs are currently still in the “innovators” phase since EV’s share of the entire passenger vehicle market share is minimal. With new regulations such as CARB’s zero-emissions vehicle mandate, EV’s market share will increase significantly moving across to the even “early majority” phase in the adoption curve. Previous modeling efforts on vehicle adoption based on Bass have been quite limited. LA is a region where culture, economics, and geography have all factored into the existence of a dominant culture for personal vehicle travel. With an increasing population along with LA’s dominant “car culture” for mobility, personal travel has increased steadily over the last decades. The trend is expected to continue with even greater population growth. 35 2.6 Electric Vehicle Charging Profiles One of the key questions policymakers have debated regarding EV adoption is the availability of electricity in the current grid infrastructure to support increased demand. Numerous studies have assessed the current grid infrastructure’s ability to support the increased demand load from increasing adoption of EVs. Schneider et al. empirically showed that the current U.S. grid infrastructure can feasibly supply the electricity to support 70% of the U.S. light-duty vehicle (LDV) fleet (2008). Many other studies have found that a significant number of EVs in the LDV fleet can be supported without requiring additional power plant or transmission capacity (Kintner- Meyer et al., 2007; Denholm and Short, 2006; Stephan and Sullivan, 2008). Other regional studies have drawn similar conclusions. Hadley and Tsvetkova used the Oakridge Competitive Electricity Dispatch (ORCED) model to examine the impacts on electricity supply from increasing EV market penetration in over 12 regions across the U.S. (2008). In a CEC report, Garland et al. examined the impacts of large scale EVs in terms of marginal electricity generation and CAP emissions in the South Coast Air Basin (1996). The grid effects on EV’s emission mitigation potential have also been studied extensively. Schneider et al. concluded that the current grid has the potential to displace 6.5 million barrels of oil per day or 52% of the nation’s oil imports (2008). Sioshansi and Denholm estimated the emission reductions from different PHEV fleet sizes and flexible generator options in the Texas electric grid along with vehicle-to-grid options (2009). The high emission mitigation potential, however, hinges on a crucial assumption that EV users charge their vehicles during “off-peak” hours (see Figure 6). McCarthy and Yang calculated that California’s marginal electricity supply has varying emission rates depending on when the charging occurs (2010). California’s marginal electricity supply is heavily dependent on natural gas steam turbine (NGST) and natural gas 36 combustion-turbine (NGCT) plants (McCarthy and Yang, 2010). If charging occurs during off- peak hours, then NGST and NGCT plants generate about 21% of the marginal electricity and the emission rate is about 570 gCO 2eq kWh -1 (McCarthy and Yang, 2010). If the charging spreads throughout the day but still remains predominantly at night, then NGST and NGCT plants supply 37% of the marginal electricity and the emission rate is about 625 gCO 2eq kWh -1 (McCarthy and Yang, 2010). Jansen et al. modeled potential changes to the emissions in the California grid GHG intensity with two different EV adoption scenarios (2010). They assumed an “ideal valley filling” scenario where users would charge their vehicles only during the off-peak hours (Jansen et al., 2010). The simulation results showed that the higher adoption of EV actually leads to a higher GHG emission grid intensity rate (Jansen et al., 2010). This is attributed to the fact that marginal generation emissions intensity is about 40% higher than the average intensity for all hours (Jansen et al., 2010). Figure 6. Typical daily California system load profile 4 Prior studies often assumed a consumer who only charges their vehicles at night or off- peak hours. Although many consumers may indeed fit this profile, the reliance on such an overly- 4 Created with data from CAISO OASIS online system load report generator (http://oasishis.caiso.com/) 37 simplified EV consumer charging behavior is overly optimistic and problematic. Axsen and Kurani conducted consumer surveys on drivers across the country to better understand driving behavior and examine how an EV can fit into their lifestyles (2008). The authors conclude that U.S. drivers indeed have the potential to integrate EVs into their lives and charge during the off-peak hours (Axsen and Kurani, 2010). However, the authors point out that the policies aimed at increasing EV adoption such as public charging stations, increases charging during daytime hours (Axsen and Kurani, 2010). Results from other studies further support such notion. In Williams et al., for example, actual driver data from EV owners showed that greater earlier evening and additional daytime charging occurred than previous studies assumed (2011). Furthermore, without such policies that promote daytime charging, significant EV adoption and subsequent increases in electricity demand that displaces gasoline consumption is highly unlikely (Axsen and Kurani, 2010). Therefore, increasing EV adoption as a result of both policies and market forces would most likely entail significant number of public charging stations that would significantly increase daytime electricity demand. The dilemma shows the complexity in the emission mitigation potential and consumer charging behavior. Previous work on vehicle charging effects on emissions show that the consumer’s choice on when to charge has a significant effect on emissions. While studies show the potential impacts for modern EVs in various regions, LA and its power infrastructure has yet to be the subject of a comprehensive impact analysis from modern EV adoption. There is also a lack of understanding of the overall effects on emissions from the region’s power grid mix and consumer purchasing and charging patterns. Furthermore, there has yet to be any analysis on the effects on emissions from the coupling of the charging patterns along with changes in the power content. With the largest vehicle population and home to one of the stringent environmental regulations in the nation, LA 38 serves as an ideal urban/suburban environment for a detailed study that integrated all of these major factors. 2.7 Models Projecting Future Transportation Emissions Recent transportation studies assess the viability of alternative technology vehicles as potential solutions to reducing transportation-related emissions. Table 9 highlights some published works that attempt to project future emissions based on advanced vehicle technology adoption and carbon policies. Researchers at the Institute of Transportation Studies at UC Davis created the Long-term Evaluation of Vehicle Emissions Reduction Strategies (LEVERS) to assess the emission reduction potential of advanced vehicle technologies in California’s transportation sector. Adoption scenarios were developed based on existing advanced vehicle technologies such as biofuel, hydrogen, and EVs. A key finding from the study is that an intensive all-electric adoption yields the greatest GHG emission reduction but fails to meet the reduction targets (Yang et al., 2009). The authors further present other scenarios with combinations of technologies and reductions in travel demand to meet reduction targets (Yang et al., 2009). These calculations, however, do not model the significant changes in California’s energy portfolio with greater adoption of renewable energy. The study makes fixed assumptions about the efficiencies of future technologies without any justification. This study was broadened in scope to the entire U.S. by McCollum and Yang (2009). The results were similar to the previous study but the deficiency in modeling remained. Researchers at MIT attempted to project the emission reductions from changes in vehicle efficiency, increase in adoption of alternative vehicles and fuels, changes to vehicle size and weight, and restraints on vehicle travel (Bandivadekar et al., 2008b). The study assumes a simplistic linear increase in the market share of alternative vehicles to make the projections. A key finding is the importance of fuel reduction over performance improvements for emissions 39 reductions (Bandivadekar et al., 2008b). The study, however, suffers from deficiencies in renewable energy projections and alternative vehicle adoption scenarios. Researchers at the Pacific Northwest National Laboratory developed a MiniCAM model to assess the emission mitigation potential of EVs with significant adoption of carbon-capture and storage (CCS) technology (Wise et al., 2010). The authors created scenarios of EV adoption with carbon pricing on electricity generation (Wise et al., 2010). Some of the main conclusions are that carbon pricing itself does not substantially boost the penetration of EVs and the transportation sector is last to be economical to decarbonize in all scenarios (Wise et al., 2010). Numerous other studies took different approaches to measure the mitigation potential in the transportation sector in the long run. Yeh et al. used an economic model (MARKAL) to assess the potential of policy-oriented mitigation strategies in the transportation sector (2008). Grimes-Casey et al. used IPCC’s MAGICC climate change modeling tool to assess the potential carbon reduction targets in the light-duty vehicle sector to reach IPCC’s mitigation goals (2008). Table 9. Major published works related to emission mitigation potential of transportation technologies C. Yang et al. A. Bandivadekar et al. M. Wise et al. Year 2009 2008a 2010 Base model LEVERS US LDV fleet model MiniCAM Region California U.S. U.S. Technologies considered: Biofuels, electric, hydrogen, high-efficient internal combustion vehicles Plug-in electric, hybrid, ethanol vehicles; efficiency gains, reductions in vehicle size and weight Plug-in Hybrid Electric Key finding(s): Electric vehicles have highest emission mitigation potential Reducing fuel consumption (versus performance) can greatly reduce the required penetration rates for advanced vehicle technologies for emission reductions Carbon pricing itself does not substantially boost electric vehicle penetration; transportation sector is the last to be economical to decarbonize Key assumption: Population doubles in 2050 GHG emission factors of future vehicle are higher because of advanced materials Two different pricing for carbon permits for electricity generation Modeling: Static (spread-sheet model) Static Dynamic, partial-equilibrium Renewable energy adoption: Not modeled; assumes grid electricity carbon intensity is 94% below 1990 level in 2050 Assumes U.S. grid energy Models changes in energy sources in response to carbon pricing; significant adoption of 40 carbon capture and storage technology Battery technology improvement: None None Assumes improvement from 0.28 to 0.25 kWh/mi in 2050 Charging profiles: None Assumes enough baseline load to meet generation requirements for new electric vehicles Assumes greater but negligible electricity demand Electric vehicle adoption: 77% of all vehicle miles Linear increase for all vehicle types Assumes two cases of market share from 2020 to 2050 41 Table 10. Overview of past works on future transportation emissions from electric-propulsion vehicle adoption Region Technologies considered Model type Spatial modeling Renewable energy adoption Battery technology improvement Charging profiles Electric vehicle adoption Proposed study Los Angeles EV, PHEV Dynamic Yes; mapping of generation sites Forecast of LADWP energy portfolios Modeling based on industry forecasts Modeling of peak generation emissions Projection based on modified Bass diffusion model Yang et al. (2009) CA Biofuel, EV, ICEV, FCHV Static N/A Assumes grid carbon intensity of 94% below 1990 level N/A N/A Assumes 77% of all vehicle miles Bandivadekar et al. (2008a) U.S. PHEV, ethanol, hybrid, ICEV Static N/A Assumes U.S. grid carbon intensity N/A Assumes enough baseline load for EV Linear increases for all vehicle types Wise el at. (2010) U.S. PHEV Dynamic N/A Models changes in carbon intensity from CCS technology Assumes improvement from 0.28 to 0.25 kWh/mi in 2050 Assumes negligible increases in electricity demand Assumes two cases of market share from 2020 to 2050 McCollum and Yang (2009) U.S. Biofuel, EV, ICEV, hydrogen Static N/A Assumes carbon intensity of 80% below 1990 levels N/A N/A Assumes 60% FCHV, 20% BEV, 20% PHEV 42 Grimes-Casey et al. (2008) U.S. PHEV, ethanol General equilibrium N/A Assumes carbon intensity equivalency of 100 g CO 2/mi N/A N/A 100% PHEV by 2015 Jansen et al. (2010) CA/ Western U.S. PHEV Static N/A Carbon intensity based on model output Assumes 0.312 kWh/mi for all PHEVs 2 different scenarios: off-peak charging and mixed charging 40% of LDV are PHEVs Kintner-Meyer et al. (2007) U.S. (divided into 12 regions) PHEV Static N/A Assumes U.S. grid energy projections embedded in GREET Assumes 0.28 and 0.42 kWh/mi for cars and SUVs, respectively Off-peak charging One PHEV per household (in economic assessment section) Peterson et al. (2011) NY PHEV Static N/A Assumes natural gas generation with 45% efficiency Assumes 0.12, 0.16, and 0.23 kWh/km for cars, vans, and light trucks in 2020 3 different charging scenarios 10% of LDV are PHEVs King and Webber (2009) U.S. ICEV, biofuel, FCHV, BEV, PHEV Static N/A Assumes U.S. grid energy Assumes 0.37 kWh/mi N/A N/A ICEV=internal combustion engine vehicle, FCHV=fuel-cell hydrogen vehicle, PHEV=plug-in electric vehicle, BEV=battery-electric vehicle, CCS=carbon capture and storage 43 3 SYSTEM MODELING In this section, a model predicting the hourly grid emissions of LADWP’s energy grid is provided. The model’s three major outputs are LADWP’s marginal carbon intensity, EV population projections for LA, and the EV charging load. The model is used to answer the major research questions in this dissertation. An overview of the modeling approach is shown in Figure 7. Although the model has multiple sub-models derived from multiple data source, it can be broadly segmented into the two parts: “supply” and “demand” side of the electric grid. The supply side is represents LADWP’s energy grid and its major output variable is the hourly marginal carbon intensity as a function of system load. The demand side represents the energy load from EV charging demands. It integrates multiple sub-models and datasets, namely the projected EV population from the Bass diffusion model and charging loads based on various scenarios. The following sections provides further details on each of the major sub-models. Figure 7. Model overview of estimating hourly GHG emissions from EV charging in LA Marginal CO 2e Intensity Renewable generation capacity Nonrenewable generation capacity Reported emissions data Emissions (LCA studies) Resource dispatching Supply Side Sales forecast (Bass model) LA’s vehicle fleet population Population projections LA’s trip data (SCAG) EV electric consumption rate Battery charge power EV Population EV Charging Load Demand Side Scrap rate (NHTSA model) Hourly system load profile 44 3.1 Modeling the Supply Side: LADWP’s Energy Grid The changes in the carbon intensity of LADWP’s grid energy is modeled based on the agency’s Power Integrated Resource Plan that contains information on the agency’s short and long- term system load projections as well as generation capacities. 3.1.1 LADWP’s generation sources The Los Angeles Department of Water and Power (LADWP) is a publicly-owned utility company that is responsible for the generation, transmission, and distribution of electricity to the City of LA. It is the largest municipal utility in the U.S. serving an area of 465 square miles and 3.9 million residents (LADWP, 2013). It is comprised of in-state natural gas, hydropower, and renewable generation sources and out-of-state coal and nuclear generation units. The generation capacities for LADWP are listed in Table 11. Note that the data shows the generation capacity, which does not imply the actual energy generated. For example, even though LADWP has a low cumulative renewable generation capacity (about 6.5%), the actual energy supplied from renewables was actually nearly 20% of retail sales in 2011 (LADWP, 2012). Renewable energy generation is significantly higher than capacity because of the state’s renewable portfolio standard (RPS) regulations. The utility agencies are required to reach certain renewable generation targets, which is prompting agencies to maximize the use of available renewable generation sources sometimes at higher marginal costs. As a result, the energy supplied by renewables make up a higher percentage of the total energy supplied than the actual capacity. 45 Table 11. List of LADWP's generation sources and net dependable capacity in 2012 (LADWP, 2012) Type Net Dependable Capacity (MW) % Non-renewables: Natural Gas 3,329 44.48% Coal 1,651 22.06% Nuclear 380 5.08% Large Hydro (>30MW) 1,643 21.95% Renewables: Small Hydro (<30MW) 174 2.33% Solar 15 0.20% Wind 218 2.92% Biomass 34 0.37% Other 45 0.61% Total: 7,484 3.1.2 LADWP’s system load profile The shape of LADWP’s system load varies daily but can be characterized by a typical seasonal profile. In the summer months (July to September), the energy demand is much higher mainly because of greater use of air conditioners in buildings. The energy demand is lower during other seasons and the differences between these periods are relatively minimal (see Figure 8). The shape of the load profile in the summer months is critical because LADWP plans its resources to handle the summer peak loads, which hit a record high of 6,142 MW in 2010 (LADWP, 2013). This implies that some of LADWP’s generation resources sit idle for much of the year and are only utilized at times of peak demand. These generation resources, often called marginal plants, play a critical role in meeting real-time demand. In this study, the focus is on the average energy load in the summer months when the average peak demand is highest. 46 Figure 8. LADWP's seasonal hourly energy load profile for the year 2011 (FERC, 2013) 3.1.3 Renewable generation sources LADWP’s ongoing and planned projects for new generation sources include those from wind and solar. Currently, combined nameplate generation capacity of wind and solar is approximately 1,013 MW (LADWP, 2012). The net dependable capacity, however, is far below at approximately 234 MW (LADWP, 2012). The lower net dependable capacity is due to the intermittent nature of wind and solar. In this study, the same assumption is made on the utilization rate (i.e., net dependable versus maximum capacity) as LADWP in its integrated resource planning. For wind generation, the utilization rate is assumed to be 10%. For solar, the utilization rates of 27% for photovoltaic (PV) generation and 68% for solar thermal, respectively. Aside from solar and wind, LADWP has also begun expansion projects for geothermal and biomass generation sources. Current projects already expand the capacity from these sources from the present 27.5 MW to 239 MW and 329 MW by the end 2020 and 2030, respectively (LADWP, 2012). 2000 2500 3000 3500 4000 4500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 ENERGY DEMAND (MW) TIME OF DAY Jan to Mar Apr to Jun Jul to Sep Oct to Dec 47 3.1.4 Non-renewable generation sources Two major changes in LADWP’s future non-renewable generation portfolio are the greater dependency on natural gas and removal of coal power. Natural gas is expected to make up 41% of the energy consumption by 2030 (LADWP, 2012). LADWP is phasing out its capacity stake in the coal-fired thermal plants in Arizona (Navajo Generation Station) by 2015 and Utah (Intermountain Power Project) by 2025. The complete phase out of these coal-fired generation sources is a sweeping change considering that coal generation accounted for 13,142 GWh or 39% of total energy consumption in 2010 (LADWP, 2012). 3.1.5 Modeling LADWP’s dispatching of generation sources LADWP meets real-time demand primarily using its own power plants and distribution resources. Each power plant operates differently because of various factors such as technology, generation capacity, type of energy resource, and economics. Large coal and nuclear plants typically are the baseload plants that operate continuously at the lowest cost. “Peaker” power plants do not operate continuously and are only dispatched when the system load is high, typically during the summer months. There have been different approaches to modeling the generation dispatching process. Parks et al. (2007) used a dispatching model based on least cost to calculate the costs and emissions of hybrid electric vehicle charging in the Xcel Energy Colorado service territory. Sioshansi and Denholm (2009) also used a least cost model in their analysis of grid operations in Texas. Unlike previous works, Jansen et al. (2010) created a grid model using the correlation in the historical data on system load and resource capacity factors for the Western U.S. region. McCarthy and Yang (2010) also analyzed the Western U.S. grid but used a dispatching model based on the total energy load. A similar method is created in this study by using the energy load rather than a least cost model or capacity factors to predict the dispatching of generation resources. 48 Every hour, the system load is checked against the available generation resource based on a list of resource priority. The types of generation resource are ranked in order of increasing marginal cost as well as consideration of renewable portfolio standards. If the system load exceeds the current resource’s net capacity, then the next generation resource is called upon to meet the demand. An overview of the dispatching model is shown in Figure 9. If excess system load remains after all generation sources have been exhausted, then the shortage is satisfied using energy imports from power plants within the WECC. Despite the elimination of the complex market interactions, this simple modeling approach yields results that closely resemble the actual resource deployment. In 2012, the actual annual generation for nuclear and coal were 11% and 41%, far above the generation capacities of 5% and 21%, respectively (LADWP, 2013). This result reflects the model’s dispatching scheme since the model dispatches the nuclear and coal generation sources before dispatching other generating sources. As a result, the model predicts a higher annual generation for these resources than actual capacity. The same reasoning applies to the dispatching of all other resources. For example, natural gas makes up 44% of LADWP’s net generation capacity but the actual generation in 2012 was 17% (LADWP, 2013) because it is one of the last resources to be utilized. 49 Figure 9. Overview of resource dispatching model 3.1.6 Marginal carbon intensity of grid energy Many studies assume a homogeneous, average grid energy emission factor when assessing GHG emissions (Bandivadekar et al., 2008a; Grimes-Casey et al., 2008; Hadley and Tsvetkova, 2009; Yang et al., 2009; McCollum and Yang, 2009; Shiau et al., 2009; Sioshansi and Denholm, 2009). As demonstrated in Chapter 2.4, however, the average grid intensity assumption has major drawbacks because of significant regional variances. Some studies have attempted to overcome the average grid intensity fallacy by calculating the “marginal” grid intensity in modeling the EV- grid interaction (Jansen et al., 2010; McCarthy and Yang, 2010). Marginal electricity is the additional electricity that must be supplied to meet the additional demand directly attributable to EV charging. Then, the marginal grid intensity is the emission intensity as a result of additional power generation sources of various types contributing to the grid supply to meet the additional 50 demand from EV charging. This study takes a similar approach and models the marginal emission factor for LADWP’s power grid assuming the resource dispatch method described in Chapter 3.1.5. The hourly marginal GHG emission factor or “marginal carbon intensity” is calculated using the LCA-based power generation emission factors and the resource dispatch model. The emission factor for each type of energy source used in the study is listed in Table 12. As previously mentioned in Chapter 2.4, the high variability in the LCA-based emission factors of some of the energy generation sources can be problematic in estimating the GHG emission factor for a specific power plant. Therefore, some of the results from the LCA harmonization studies by NREL were utilized as the best estimate for specific energy generation sources. The hourly marginal carbon intensity is calculated as the following: 1) For a given time (t), the total system load (L) to the LADWP energy grid system is calculated. 2) The resource dispatching model uses (L) as the input to compute the type (i) and level (X) of generation of each of the N energy generation sources (i.e., coal, nuclear, solar, etc.). Therefore, each X i represents the generation of each type of energy source. 3) The average emission factor (E t) is calculated by calculating the weighted average of the generation source and its associated LCA emissions (e i). That is, 𝐸 𝑡 = ∑ 𝑋 𝑖 𝑒 𝑖 𝑁 𝑖 =1 𝐿 This formulation relies on an implicit assumption that the additional load is distributed across all of the available energy generation sources. Since electricity generated from a particular power plant cannot be assigned to a specific load (such as EV charging), the same grid emission intensity is attributed to every load that is operating during a given hour. Based on the formulation, the additional load from EV charging directly affects the hourly marginal carbon intensity. 51 Table 12. Lifecycle emission rates of power generation plants used in the study Energy Source g CO2e/kWh Source Coal 1050 Whitaker et al. (2012) Natural Gas 506 O’Donoughue et al. (2014) Nuclear 18 Warner and Heath (2012) Solar 50 Fthenakis et al. (2008) Wind 14 Dones et al. (2007) Small Hydro 11 Bergerson and Lave (2002) Geothermal 120 ECW (2009) Biomass and Waste 31 Spitzley and Keoleian (2005) Large Hydro 240 Pacca (2007) 3.2 Modeling the Demand Side: EV Charging Scenarios The emission impacts from EVs depend on the daily electric load patterns resulting from EV charging. Recent studies have made different assumptions on the daily EV charging patterns. Many studies assume an ideal scenario where charging primarily occurs during the night or the “off-peak” hours (EPRI, 2007; Kintner-Meyer et al., 2007; Stephan and Sullivan, 2008). This ideal scenario is often referred to as “valley-filling” because EV charging increases the energy load of the off-peak hours closer to peak-hour levels resulting in a more leveled energy load profile. A major advantage of this scenario is that electricity costs are lowest during these hours since electricity is generated from baseload plants that would otherwise be underutilized. A number of studies use survey travel data to create charging scenarios. Sioshansi and Denholm (2009) considered the “wherever, whenever” charging case where charging occurs whenever vehicles are parked based on travel survey data of the St. Louis metropolitan area. Kang and Recker (2009) used the 2000-2001 California Statewide Household Travel Survey to consider four different scenarios including the “whenever, wherever” case as well as controlled off-peak 52 and public charging cases. Axsen et al. (2011) used a sample of travel survey data of new vehicle buyers in California to consider three different charging scenarios and also incorporated the availability of an electrical outlet within 25 feet as a prerequisite to charging. Weiller (2011) used the National Household Travel Survey to create uncontrolled and delayed charging scenarios where restrictions are placed on certain charging times. Kelly et al. (2012) expanded on Weiller’s recharging scenarios by incorporating demographic influences on charging behavior. 3.2.1 Los Angeles travel pattern and charging scenarios In this study, different charging scenarios for LA are modeled. First, the off-peak or the “smooth valley-filling” scenario is modeled where charging occurs exclusive at home during off- peak hours (night time). Then the “controlled peak” scenario is considered where charging occur primarily during the day. In both cases, the total EV load remains constant to keep the ramp up from the original energy load profile to a minimal. Lastly, a charging scenario is considered based on survey results by the Southern California Association of Governments (SCAG) that explicitly measured the travel distribution of LA (Figure 10). SCAG’s 56,428 data points show a clear concentrated peak travel period in the morning at 8 a.m. A second concentrated travel period occurs in the late afternoon hours between 4 and 6 p.m. The implicit assumption is that LA’s fleet of EVs will also exhibit the same travel behavior. Based on SCAG’s data, a charging profile for LA is created with an assumption that each EV is charged twice in any 24-hour period. SCAG’s travel pattern shows that the first 50% of all trips occur between 12 a.m. and 2 p.m. Previous SCAG studies also show that the duration of a single vehicle trip for work in Los Angeles is 29 minutes (SCAG, 2008). Therefore, the “first” charging of all EVs is assumed to occur between 1 a.m. and 3 p.m. distributed based on the travel distribution. Therefore, the peak charging occurs at 9 a.m. after the peak trip occurrence at 8 a.m. The “second” 53 charging occurs between 4 p.m. and 12 a.m. following trips generating between the hours 3 p.m. to 12 a.m. The charge duration is determined by the EV’s electric consumption rate and the EV’s onboard charger. A summary of the charging scenarios are listed in Table 13. Figure 10. Trip distribution by time of the day in Los Angeles (Fehr, 2009) Table 13. Summary of charging scenarios Scenario Available Charging Hours Location Case 1 - smooth off-peak 8 p.m. to 8 a.m. Primarily home Case 2 - controlled peak 9 a.m. to 7 p.m. Primarily work Case 3a - uncontrolled (3.3 kW) Anytime Both home and work Case 3b - uncontrolled (6.6 kW) Anytime Both home and work Case 3c - uncontrolled (10 kW) Anytime Both home and work 3.2.2 EV electric consumption rate The EV electric consumption rate is the amount of energy exhausted per mile traveled. Rates vary in the literature because the electric consumption depends on many factors such as vehicle size, weight, body shape, driving habit, and climate conditions. The U.S. EPA’s urban dynamometer driving schedule (UDDS) and the highway fuel economy test (HWFET) rate the 0 1000 2000 3000 4000 5000 6000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Frequencey of Trips Trip Departure Time (24 Hours) 54 typical energy consumption of a mid-size vehicle as 0.27 kWh/mi for urban and 0.22 kWh/mi for highway driving (Young et al., 2013). EPA’s US06 standard, which assumes more aggressive driving, rates the energy consumption as 0.40 kWh per mile (Young et al., 2013). Some other values reported in the literature are 0.21 kWh/mi (Kang and Recker, 2009), 0.30 kWh/mi (Sioshani and Denholm, 2009), and 0.41 kWh/mi (Stephan and Sullivan, 2008). In this study, the electric consumption rate is assumed to be 0.35 kWh per mile. Then, the average electric consumption rate is 11.55 kWh per day based on 33 miles traveled, which is the average for a driver in LA based on previous studies (SCAG, 2008). 3.2.3 Charge power The charge power refers to the rate of charging for an EV. This value is dependent on the type and level of charger. The different types and levels of charging are listed in Table 14. Recent survey of California's EV owners (the "first adopters") shows that vast majority of the charging is occurring at home using a Level-2 AC charger during off-peak hours (CSE, 2013). Even though the Level-2 AC charger today can supply up to 7.2 kW (240V, 30A), the actual charging rates are limited by the EV's on-board charger. The most popular EVs currently have on-board AC chargers with a maximum power rating of only 3.3 kW (e.g., Nissan LEAF, Chevy Volt) but higher level chargers are available (e.g., 6.6 and 10 kW AC). In this study, the onboard charging rates are set at three settings: 3.3, 6.6, and 10 kW. An efficiency rating of 90% is assumed for these chargers (EPRI, 2009). Table 14. Specifications of electric vehicle chargers (SAE, 2011) Max Voltage (V) Max Current (Amp) Max Charge Power (kW) Charge Time* (h) AC Level 1 120 16 1.92 21 AC Level 2 240 80 19.2 8 DC Level 1 450 80 36 0.74 *Charging time estimated based on Nissan LEAF with a 24 kWh battery pack and 3.3 kW onboard charger. 55 3.3 Modeling the EV Fleet Size The EV fleet size in LA is modeled by estimating new EV sales, scrap rate, and changes in LA’s overall passenger vehicle fleet size. An overview of the model is shown in Figure 11. The EV sales projections are created using working-age population as a surrogate measure to predict future sales. These projections are then used in the overall fleet estimates accounting for the vehicle scrap/survival rates from previous studies. Figure 11. Overview of LA's EV fleet projection model 3.3.1 Vehicle sales projection Past new vehicle registration data at the city level is limited and rarely accessible. Since past sales data is necessary to estimate changes in LA’s vehicle fleet, a model based on surrogate data is created by using sales data at the multi-county level that are more readily available. California Motor Car Dealers Association (CMCDA) publishes sales data for the entire state as well as Los Angeles and Orange County. CMCDA’s data categorizes vehicle sales broadly into cars and trucks as well as specific vehicle types such as compact sedan, midsize, sports-utility- vehicle, etc. (2013). CMCDA’s published reports indicate that Los Angeles and Orange County 56 typically have approximately 600,000 new vehicle sales annually. Passenger vehicle sales have risen in proportion from about 50% in 2005 to over 65% of all sales in 2012. Vehicles sales in LA is estimated by disaggregating this sales data based on the working-age population. “Working-age” population is defined as residents between the ages 18 and 64. The main assumption is that vehicle sale is a function of the changes in the working-age population. Therefore, as the region’s working- age population increases, new vehicle sales also increase. Detailed census data and population projections for each county are provided by the California Department of Finance (CADF, 2013). Data specific to the City of LA is obtained from studies published by SCAG that provides detailed census data and population forecasts (SCAG, 2012). Since city population projections are based on 10-year or longer increments, a linear regression method is used to estimate all other years (see Figure 12). Figure 12. LA's projected population using linear regression The working-age population for the City of LA is projected based on that of LA County. First, the proportion of LA County’s working-age population to the entire population is calculated. y = 20,590.066x - 37,585,020.637 R² = 0.999 3,700,000 3,800,000 3,900,000 4,000,000 4,100,000 4,200,000 4,300,000 4,400,000 2005 2010 2015 2020 2025 2030 2035 2040 City of Los Angeles Popluation Year 57 City of LA’s working-age population is assumed to have the same proportion relative the city’s entire population that was previously estimated using linear regression. That is, the same proportion is assumed to hold for the working-age populations of LA County and LA City. Based on these assumptions and census data projections, LA’s working-age population with respect to the County is shown in Figure 13. Figure 13. Projected working age population in LA and Orange County The vehicles sales of the City of LA is estimated based on vehicle sales for LA and Orange County. By applying the working-age population proportions, LA’s vehicle sales for cars and trucks are estimated for past years. For example, if the City of LA’s working age population makes up 60% of the County’s working age population, then the City’s vehicle sales is also 60% of the County’s. The results are shown in Figure 14. 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000 2010 2015 2020 2025 2030 2035 WORKING AGE (18-64) POPULATION (IN 1000S) YEAR City of LA LA County Orange County 58 Figure 14. Estimated vehicle sales in LA based on County sales data Future vehicle sales are estimated using the changes in working-age population as the growth rate. First, the growth rate in the working-age population of LA County is calculated. Then, LA County’s vehicle sales changes depending on the growth rate of its working-age population. The working-age population proportion is applied to the Country vehicle sales projections to derive the City’s future vehicle sales estimates. Based on these assumptions, the projected populations and new vehicle sales forecast are shown in Table 15. The sales projection serves as the upper bound of EV sales projection presented in the following section. Table 15. Projected population changes and associated new car sales for Los Angeles Year Total Population Working Age (18-64) New Passenger Vehicle Sales 2010 3,795,781 2,458,680 70,870 2012 3,857,799 2,505,648 102,619 2020 3,991,700 2,518,767 103,157 2030 4,212,813 2,484,844 101,767 2040 4,418,714 2,536,959 103,902 2050 4,624,615 2,584,145 105,834 0 100,000 200,000 300,000 400,000 500,000 600,000 700,000 800,000 2005 2006 2007 2008 2009 2010 2011 2012 VEHICLE SALES PER YEAR YEAR Estimated LA Cars LA & Orange County Cars Estimated LA Trucks LA & Orange County Trucks 59 3.3.2 EV sales forecast based on Bass diffusion model Previous studies have made an explicit assumption on the level of EV adoption for a particular year (Bandivadekar et al., 2008a; Samaras and Meisterling, 2008). In these studies, the EV market projections, however, are fairly simplistic often with an arbitrary market share percentage or a fixed number of vehicles (e.g., 1 million PEVs) without much justification. Therefore, the resulting emission forecasts may be greatly overestimated or underestimated for any given time forecast. In this dissertation, the Bass technology diffusion model is used to project EV adoption. Using the Bass model, the number of EVs operating in LA is forecasted over the next 30 years. In the “baseline” case, the maximum market penetration is estimated as 75% of new vehicle sales in LA, which captures all historical LA vehicle sales in the light duty vehicle (LDV) category except for large trucks, vans, and sports utility vehicles (SUVs). In the “low” case, the maximum market penetration is estimated at 65%, which is approximately the historical percentage of passenger vehicle sales in LA (CNCDA, 2013). In the “high” case, the EV market extends the entire LDV category except for full size trucks. The model parameters for the three adoption cases are shown in Table 16. Table 16. Model parameters for three adoption cases Low Base High m 117,000 135,000 162,000 p 0.004 0.007 0.01 q 0.15 0.2 0.25 The forecast for the three EV adoption scenarios are shown in Figure 15. In the baseline case, EV sales reach 11% and 38% of all vehicles sales in LA by 2020 and 2030, respectively. In the high case, EV market share reach 22% and 70% by 2020 and 2030, respectively. EV sales remain minimal in the low case until 2030 when sales exceed 20,000 units. These forecasts depend 60 on a time scale that is inherently uncertain; mass adoption of new technology (if successful) do not follow a common timeline. Zoepf (2011) has shown that the time scales of new technologies from market introduction to maximum market penetration has decreased in recent decades to about 10 years. Some factors to the shorter time scale can be attributed to fewer supply constraints, better communication (e.g., increasing consumer awareness, and marketing), and regulation (Zoepf, 2011). Vehicle technologies, however, tend to have much longer time scales (Zoepf, 2011). One estimate shows that the major fleet penetration of EVs is less than 15 years away (i.e., before year 2030) if one assumes that EVs have already achieved significant new vehicle production (Bandivadekar et al., 2008b). Therefore, actual EV adoption may be higher than the proposed scenarios but with high uncertainty. Figure 15. EV sales scenarios in Los Angeles based on Bass model 3.3.3 Vehicle survival rate Increasing sales of EVs will undoubtedly change the composition of LA’s vehicle fleet population. To assess the impact of annual EV sales on the region’s entire EV population, vehicle scrap rates must be integrated into the model. Unfortunately, there is yet to be any existing data 0 20 40 60 80 100 120 140 160 180 2012 2020 2030 2040 New EV Sales (1000s) Year Low Baseline High 61 available on the actual scrap rates of EVs. Therefore, past studies and published data on scrap rates of all motor vehicles are used to estimate the scrap rate. Two main data sources are considered to estimate the scrap rate of EVs:  The Oak Ridge National Laboratory (ORNL) calculated the survival rate (i.e., 1 minus the scrap rate) of vehicle models sold after 1990 based on the vehicle scrap model by Greenspan and Cohen (Davis et al., 2012).  National Highway Traffic Safety Administration (NHTSA) (2006) used data from the National Vehicle Population Profile (NVPP) to create a linear regression model estimating the vehicle survival rate based on vehicle age: 𝑆𝑢𝑟𝑣𝑖𝑣𝑎𝑙 𝑟𝑎𝑡𝑒 = 1 − exp [− exp(𝐴 + 𝐵 ∗ 𝑎𝑔𝑒 )] o If vehicle age is less than 10 years, then A = 1.64905, B = -0.12143 o If vehicle age is greater than 10 years, then A = 3.38136, B = -0.28623 o The NHTSA model predicts a more rapid decline in the survival rates, which implies a faster fleet turnover. Comparison of the two vehicle survival rates are shown in Figure 16. EVs are dependent on advanced battery technology with potentially limited lifespan. Most manufactures (e.g., Toyota) rate the battery lifespan of their popular PEV models to 100,000 miles (about 10 years). Actual field tests, however, suggest that EV owners can use their vehicles beyond the rated lifespan without experiencing any significant drop in performance (Smith et al., 2011). Therefore, the EV survival rate for later years (beyond 10 years) is still significant, albeit possibly lower than ICEVs. Based on these characteristics of the two models, the results from the NHTSA study is integrated into the model based on current status of advanced vehicle battery technology. 62 Figure 16. Vehicle survival rate of two different models (NHTSA, 2006; Davis et al., 2012) 3.3.4 Los Angeles vehicle fleet projection LA’s overall fleet population is modeled using the working-age population as a surrogate to forecast changes in the fleet population. Using vehicle fleet size data at the county level (i.e., LA County), the vehicle fleet population is estimated at the city level because vehicle sales data is accumulated at the county level. The initial passenger vehicle fleet size at base year 2010 is calculated based on the proportion of the city’s working age population with respect to that of the entire LA County. Population data from the state and local agencies are used for calculations (CADF, 2013; SCAG, 2012). Based on these assumptions, the number of passenger vehicles in LA County at base year 2010 is estimated at 5,859,407. The corresponding number of vehicles in the City of LA is estimated at 2,263,739. The fleet size changes with respect to changes to the working-age population. Within the fleet population, the number of EVs in operation increase as the sales increase. However, the diffusion of EVs into the entire fleet is not entirely cumulative because of scrap rates. By incorporating the scrap rate model in the previous sections, the total number of EVs with LA’s vehicle fleet increases but at a slower rate than the annual sales. The 0% 20% 40% 60% 80% 100% 1 6 11 16 21 SURVIVAL RATE AGE OF VEHICLE (YEARS) NHTSA ORNL 63 results for the three EV sale scenarios are shown in Figure 17. Note that conventional ICEVs remain dominant well into the year 2030 even in the baseline and high EV adoption scenarios. Figure 17. EV fleet size with respect to total vehicle fleet in Los Angeles 3.4 GHG Emissions of Gasoline Vehicles In this section, the future GHG emissions of a gasoline vehicle (i.e., ICEV) is estimated. The ICEV’s GHG emissions serve as the baseline for assessing the GHG mitigation potential of the major EV parameters considered in this study. Future ICEV GHG emission rate is estimated based on fuel efficiency standards and environmental regulatory laws already established for future years. Figure 18 shows an overview of the calculation for the estimated GHG emissions per vehicle per day. The “energy content” measured in megajoules per gallon is roughly a measure of the amount of energy stored in a gallon of gasoline and is assumed remain constant at 122.475 MJ per gallon based on fuel input specification in Argonne’s GREET model (AFDC, 2014). The “carbon intensity” measured in grams of CO 2e per megajoule is projected based on California’s Low Carbon Fuel Standard (LCFS) mandate (CARB, 2012b). “Fuel efficiency” measured in miles per 0 500 1,000 1,500 2,000 2,500 2012 2015 2018 2021 2024 2027 2030 2033 2036 2039 2042 2045 Number of Vehicles (1000s) Year Low Baseline High Total Fleet Size 64 gallon is projected based on the federal CAFE standards (DOT, 2012). For travel intensity, the average distance driven by a driver in LA is estimated to be 33 miles per day based on previous studies by SCAG (SCAG, 2008). Figure 18. Overview of the calculation for ICEV GHG emissions per day 3.4.1 Fuel efficiency The fuel efficiency standard is calculated based on the CAFE standards already established by the U.S. Department of Transportation (DOT) and Environmental Protection Agency (EPA). The current legislation calls for approximately 4% annual rate of increase in fuel efficiency for between the years 2017 and 2025 (DOT, 2012). The combined average fleet-wide fuel economy are assumed to represent the ICEV’s fuel efficiency in future years 2020 and 2030. Since there is yet to be a standard for the year 2030, the fuel economy standard is assumed to equivalent to the value in the year 2025, which is currently the highest value. Therefore, the fuel efficiency for an ICEV is 38.3 and 48.7 miles per gallon in years 2020 and 2030, respectively. 3.4.2 Fuel carbon intensity The fuel carbon intensity is calculated based on California’s LCFS mandate established the California Air Resources Board. The current mandate calls for an immediate reduction in the carbon intensity in transportation fuels and an overall reduction to 89.06 average CO 2e per megajoule by the year 2020 and beyond (CARB, 2012b). Therefore, the average carbon intensity for gasoline is assumed to be 89.06 CO 2e per megajoule. 65 4 RESULTS AND DISCUSSION For each charging scenario, the following variables are considered: daily EV energy load, system load profile, and average marginal carbon intensity. As described in Chapter 3.1.2, utility agencies build and plan for the summer months when energy demand is highest. Therefore, all modeling results are presented with respect to LADWP’s load profile for the summer months (i.e., July to September). 4.1 EV Energy Loads in Year 2020 4.1.1 Daily EV energy load The net energy load requirements for LA’s EV fleet in the year 2020 for the three charging scenarios are shown in Figure 19. All cases show the baseline adoption scenario for LA’s EV adoption. Both Case 1 (off-peak) and 2 (peak) show the idealized scenario where the EV charging load remains leveled. Case 3 scenarios show different energy loads depending on the onboard battery charger in the EV. For the 3.3 kW case, the time duration of charge increases causing large spikes in energy load, especially at 4 p.m. (hour 16). For certain periods of the day, the 3.3 kW case actually results in greater hourly energy load requirements than the 6.6 kW case. Overall, total EV charging load is modest at 100 to 180 MW at any given hour. 66 Figure 19. Hourly energy load requirement from EV charging in LA in 2020 4.1.2 System energy load profile Modeling results for 2020 show that EV charging loads in LA have minimal effect on LADWP’s daily load profile. Figure 20 shows the off-peak or the smooth valley-filling scenario (Case 1) for the three EV adoption forecasts (i.e., low, baseline, and high). Even in the high adoption scenario, the incremental EV energy load has negligible effect on the overall profile. This result verifies previous studies’ finding that EV adoption does not require any major additional electric generation sources (Parks et al., 2007; Kintner-Meyer et al., 2007; Schneider et al., 2008; Stephan and Sullivan, 2008). In the ideal peak charging scenario (Case 2), the three EV adoption forecasts also yield negligible impacts on the overall energy load profile (Figure 21). Figure 22 shows the results for the Case 3, baseline EV adoption scenario. For all levels of chargers, there is a negligible impact of EV charging on the overall load profile (Figure 22). 0 20 40 60 80 100 120 140 160 180 200 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hourly EV Charging Load (MW) Time of Day Case 1: Smooth Valley-Filling Case 2: Controlled Peak Case 3A: Uncontrolled (3.3 kW) Case 3B: Uncontrolled (6.6 kW) Case 3C: Uncontrolled (10 kW) 67 Figure 20. LADWP’s hourly energy load profile during summer months for Case 1 (off-peak) Figure 21. LADWP's hourly energy load profile during summer months for Case 2 (controlled peak) 2000 2500 3000 3500 4000 4500 5000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energy Load (MW) Time of Day System load without EV Low Baseline High 2000 2500 3000 3500 4000 4500 5000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energy Load (MW) Time of Day System load without EV Low Baseline High 68 Figure 22. LADWP's hourly energy load profile during summer months for Case 3 (uncontrolled) 4.1.3 EV-related GHG emissions The EV marginal GHG emission rate was calculated by determining the hourly energy load on LADWP’s grid then calculating the hourly average grid carbon emission intensity. Varying hourly energy load changes the type and amount of energy generation mix that causes the changes in the grid carbon intensity. Since electricity generated from a particular power plant cannot be assigned to a specific load, the same grid emission intensity is attributed to every load that is operating during a given hour. Therefore, the additional demand from EV charging increases the overall emissions but can either increase or decrease the average grid emission intensity for a given hour. In the off-peak charging scenario (Case 1), additional EV energy load lowers the hourly marginal carbon intensity relative to the reference load during the charging hours (Figure 23). However, the marginal carbon intensity is higher for these hours relative to the peak hours. In the controlled peak charging scenario (Case 2), the reverse is true (Figure 24). This implies that EV charging during peak hours results in a lower emissions since the incremental generation has lower 2000 2500 3000 3500 4000 4500 5000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energy Load (MW) Time of Day System load without EV Case 3a: 3.3 kW Case 3b: 6.6 kW Case 3c: 10 kW 69 marginal carbon intensity. The same results are shown in the uncontrolled charging scenario (Case 3) (Figure 25). Results for Case 3 also show that the type of onboard charger has negligible effect on the hourly margin carbon intensity (Figure 25). Figure 23. LADWP's marginal carbon intensity for Case 1 charging (off-peak) Figure 24. LADWP's marginal carbon intensity for Case 2 charging (controlled peak) 510 511 512 513 514 515 516 517 518 519 520 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Marginal Carbon Intensity (kg CO 2e /MWh) Time of Day Low Baseline High 510 511 512 513 514 515 516 517 518 519 520 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Marginal Carbon Intensity (kg CO 2e /MWh) Time of Day Low Baseline High 70 Figure 25. LADWP’s marginal carbon intensity for Case 3 and baseline EV adoption (uncontrolled) 4.2 EV energy load requirement scenarios in year 2030 4.2.1 Daily EV energy load In 2030, the energy load requirements from EV charging increases from the rising number of EVs in LA’s vehicle fleet (Figure 26). The three charging cases are shown in Figure 26 with each case assuming the baseline EV adoption scenario. The overall hourly load profiles do change considerably from 2020, as there is greater “ramping up” and “ramping down” in the energy load profile. The peak energy load for the idealized off-peak (Case 1) and peak (Case 2) scenarios are between 600 and 735 MW, significantly higher than 2020. In Case 3, the hourly EV energy load can exceed 1,100 MW during peak charging times. 510 511 512 513 514 515 516 517 518 519 520 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Marginal Carbon Intensity (kg CO 2e /MWh) Time of Day Case 3a: 3.3kW Case 3b: 6.6 kW Case 3c: 10 kW 71 Figure 26. Hourly energy load requirement from EV charging in LA in 2030 4.2.2 System energy load profile In 2030, the energy load profile shows a sharp increase during the charging periods due to the greater number of EVs in operation. Results show that in all scenarios, one of the more pressing concerns is the significant “ramping up” and “ramping down” of the entire system due to the additional EV load. In the off-peak charging scenario (Case 1), the system load may require ramping up and down upwards to 900 MW in an hour in the high adoption scenario (Figure 27). In the ideal peak charging case (Case 2), the ramping up and down requirements can range from 550 to 1,600 MW in an hour depending on adoption (Figure 28). The uncontrolled charging scenario (Case 3) actually has the least ramping up and down requirements at 400 to 700 MW in an hour (Figure 29). There are also significant generation capacity concerns for the restrictive charging scenarios. Figure 27 shows the off-peak charging case (Case 1) for the three EV adoption forecasts (i.e., low, baseline, and high). In the high adoption scenario, the incremental EV energy 0 200 400 600 800 1000 1200 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hourly PEV Charging Load (MW) Time of Day Case 1: Smooth Valley-Filling Case 2: Controlled Peak Case 3A: Uncontrolled (3.3 kW) Case 3B: Uncontrolled (6.6 kW) Case 3C: Uncontrolled (10 kW) 72 load may be problematic during the off-peak hours especially in the summer. The sharp spike at 8 p.m. from vehicles arriving home and charging may exceed the upper generation capacity limits. In the ideal peak charging scenario (Case 2), the three EV adoption forecasts show that EV charging pushes the energy load to significantly higher levels (Figure 28). For 2030, LADWP’s planned dependable generation capacity is approximately 6,482 MW so there is enough capacity to meet demand on an average day. However, the energy load from EV may push the energy load beyond the limits during extremely hot days. For example, during some of hottest summer days, LADWP’s peak demand can reach over 6,000 MW. With additional load from EV charging during peak hours, there may simply be not enough generation capacity causing blackouts or orders of energy imports that may be worse in carbon intensities. For Case 3, the effects of EV charging increases the overall load profile. Case 3 also shared the same potential problems as the other scenarios in terms of dependable generation capacity during extreme weather conditions. The result also shows that the highest onboard charger (10 kW) exerts rather modest incremental pressure on the overall system load profile relative to other chargers than previously expected (Figure 29). 73 Figure 27. LADWP's hourly energy load profile during summer months for Case 1 (off-peak) Figure 28. LADWP's hourly energy load profile during summer months for Case 2 (controlled peak) 2000 2500 3000 3500 4000 4500 5000 5500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energy Load (MW) Time of Day System load without EV Low Baseline High 2000 2500 3000 3500 4000 4500 5000 5500 6000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energy Load (MW) Time of Day System load without EV Low Baseline High 74 Figure 29. LADWP's hourly energy load profile during summer months for Case 3 (uncontrolled) 4.2.3 EV-related GHG emissions In 2030, the marginal carbon intensity changes significantly from 2020. In the ideal off- peak charging scenario (Case 1), the average marginal carbon intensity is lower for the off-peak charging hours (Figure 30). In the controlled peak charging scenario (Case 2), the rate is even lower for the off-peak charging hours (Figure 31). The lowering of the marginal carbon intensity can be attributed to the complete phase out of the coal power plants from LADWP’s generation sources. Since coal has the highest carbon intensity, the additional EV load during off-peak hours is now supplied by a less carbon intensive energy source. During peak hours, the marginal carbon intensity decreases with greater EV adoption. With greater adoption, the additional EV load is being supplied with more hydro sources that lowers the overall margin carbon intensity. In the uncontrolled charging scenario (Case 3), EV charging has a similar effect on the hourly marginal carbon intensity as in Case 2 (Figure 32). 2000 2500 3000 3500 4000 4500 5000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energy Load (MW) Time of Day System load without EV Case 3a: 3.3 kW Case 3b: 6.6 kW Case 3c: 10 kW 75 Figure 30. LADWP's marginal carbon intensity for Case 1 charging (off-peak) Figure 31. LADWP's marginal carbon intensity for Case 2 charging (controlled peak) 200 220 240 260 280 300 320 340 360 380 400 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Marginal Carbon Intensity (kg CO 2e /MWh) Hour Low Baseline High 200 220 240 260 280 300 320 340 360 380 400 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Marginal Carbon Intensity (kg CO 2e /MWh) Hour Low Baseline High 76 Figure 32. LADWP’s marginal carbon intensity for Case 3 and baseline EV adoption (uncontrolled) 4.3 Discussion on EV Energy Loads and GHG Emissions As EV adoption increases, there seems to be a persistent concern over the EV charging effects on peak demand of power in LA. In fact, under certain cases, the additional load from uncontrolled EV charging may indeed push the overall system to a level beyond the grid’s capacity. The negative results would be potential blackouts, damages to grid components (e.g., transformers), higher supply costs, energy imports with greater emissions, etc. Modeling results, however, indicate that prior to the year 2020, these concerns are inconsequential even in the most optimistic EV adoption scenario. On an average summer day, the EV energy load makes up less than 5% of the entire system load. Across all charging scenarios, the effects from EVs on the overall system load profile remain minimal and are unlikely to cause major supply disruptions. In the year 2030, however, these concerns become relevant as EV charging requirements increase significantly beyond the grid’s capacity. The EV energy load can exceed 1,100 MW or almost 17% of the entire generation capacity in 2030. The impact of EV energy load is also 200 220 240 260 280 300 320 340 360 380 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Marginal Carbon Intensity (kg CO 2e /MWh) Hour Case 3a (3.3kW) Case 3b (6.6kW) Case 3c (10kW) 77 exacerbated by the fact that the removal of coal generation sources in 2025 diminishes the region’s maximum generation capacity by approximately 1,000 MW. Furthermore, the additional EV energy load may have a significant effect on the stability of the overall system load profile. During the summer months, the system load profile already exhibit significant ramping up and down behavior, as the electric demand rises from the off-peak to peak hours. The additional EV load magnifies the ramping up and down requirements significantly. Under certain cases, the system load demand can require a rapid ramping up of over 1,600 MW within an hour. Such requirements may not only be unstable but also destructive to grid components leading to accidents and costly repairs. Many have proposed off-peak charging as the primary means to power EVs because of the apparent minimal impacts on peak demand. For LADWP, however, there is a clear tradeoff between off-peak charging for peak demand management and GHG emissions. In 2020, the high percentage of coal power in LADWP’s base load results in a higher hourly marginal carbon intensity during off-peak hours. Therefore, additional EV load during the off-peak period results in higher marginal emissions compared to charging during peak hours. In other words, in the short run, peak charging is preferable to off-peak charging in terms of emissions. This results runs counter to previous marginal emission studies for California that showed significantly higher carbon intensity during peak periods. Unlike the rest of California, LADWP has a high percentage of coal power and relies on hydro power for peak power supply rather than natural gas. These differences in generation resources cause the divergence in marginal carbon intensity between California and LA. For these reasons, the current incentives to encourage off-peak charging may not be optimal in terms of GHG emission reductions. In terms of policy, the current “time-of-use” (TOU) pricing may be more economical for the power agency but results in greater GHG emissions. 78 Another significant policy implication is emerging from these results is that the focus on off-peak charging may actually hinder greater EV adoption because of reinforcing perception of the unavailability and high costs of daytime charging. Considering its minimal impacts on peak demand in the short run, a greater incentive for availability and better pricing for peak charging may create greater GHG emission reductions and encourage further behavioral change for early adopters. In the long run, however, LADWP’s marginal carbon intensity becomes similar to the rest of California’s power systems. The complete removal of coal in 2025 changes the hourly marginal carbon intensity significantly and makes off-peak EV charging more preferable. In such a scenario, off-peak charging lead to benefits for both peak demand management and lower marginal carbon intensity. Modeling results show that a greater understanding of the region’s specific electricity generation and dispatching of resources are imperative to creating the best strategies for EV adoption and charging behavior that would maximize emissions mitigation. The results of the study raise some major questions about effects of EV energy load on future generation capacity and emissions. Figure 33 and Figure 34 show the type and scale of generation sources in the general dispatching order with respect to the system loads in 2020 and 2030, respectively. For all three charging cases, the baseline EV adoption scenario is shown. The all-time peak demand for LADWP’s system is also shown (occurred in 2006). In 2020, there is enough generation capacity to meet increasing demand even on the hottest summer days with the highest energy demands. The expected generation capacity is sufficient to meet the anticipated energy loads even in the worst scenario ever recorded (i.e., all-time peak in 2006). The actual marginal carbon intensity is lower during the peak periods because of LADWP’s dependency on hydro resources for peak demand management. As pointed out previously, the marginal carbon 79 intensity during off-peak hours are worse because of the high proportion of coal in LADWP’s generation mix. Figure 33. System load scenarios in 2020 with respect to generation sources in general dispatching order Conditions change significantly in 2030 because of the removal of coal generation sources and increasing load from EV charging. The elimination of coal implies that LADWP will increasingly rely more heavily on natural gas to fulfill energy demand, which is in line with the rest of California. As a result, the marginal carbon intensity decreases for all hours because of the complete removal of coal. Furthermore, LADWP’s utilization of hydropower resource increases significantly during peak demand hours (see Figure 34). The increasing reliance on hydropower raises major concerns. First, hydro power inherently suffers from high variability throughout the year. Consequently, LADWP may not have the anticipated hydro resources to fulfill the additional energy load demands from EV charging. Secondly, the unavailability of hydropower will require greater energy imports. The energy imports may be from power plants within the state. In such a case, the type of generation would be natural gas which would have significantly higher marginal 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 System Load (MW) Hour Nuclear Coal Renewables DG Natural Gas Hydro All-time Peak Off-Peak Peak Uncon. (6.6kW) 80 carbon intensity than hydropower. The required energy resources may also be imported from other nearby states. In such a case, the average marginal carbon intensity would be far worse since other states have higher proportion of coal energy in their generation mix. In other words, the elimination of coal from LADWP’s generation mix may actually not reduce the consumption of coal energy. Rather, the additional energy loads from EV charging may be fulfilled by energy imports derived from coal power plants. The result would be significantly higher carbon intensities associated with EV charging. Figure 34. System load scenarios in 2030 with respect to generation sources in general dispatching order 4.4 Emissions Mitigation Potential of EV Policy Levers In this section, further analysis on the modeling results are performed to provide some strategic insight into potential policy options for GHG emission reductions. The main “policy levers” correspond to the main EV-related parameters in the model that may have a significant effect on the overall GHG emissions mitigation potential. 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 System Load (MW) Hour Nuclear Renewables DG Natural Gas Hydro All-time Peak Off-Peak Peak Uncon. (6.6kW) 81 4.4.1 Baseline case First, the baseline case is established to assess the impacts of EV parameters in GHG emissions reduction potential. The average GHG emissions for an ICEV are estimated for the years 2020 and 2030 based on the calculation described in Chapter 3.4. Table 17 shows the average GHG emissions per day for an ICEV in LA in the year 2020 and 2030, respectively. The results show a significant decline in the emissions due to the projected increase in the average fuel efficiency according to the CAFE standards. Table 17. Daily GHG emissions from ICEVs in LA in year 2020 and 2030 Unit 2020 2030 Energy Intensity MJ per gallon 122.475 122.475 Carbon Intensity g CO 2e per MJ 89.06 89.06 Vehicle Efficiency Miles per gallon 38.3 48.7 Travel Intensity Miles per day 33 33 Emissions kg CO 2e per day 9.398 7.391 The average marginal carbon intensity of LADWP’s grid without EV charging is calculated for years 2020 and 2030, respectively. The results of the seasonal averages for the year 2020 are shown in Figure 35. The peak carbon intensity values typically occur during the early morning hours for all seasons (see Figure 35). One notable result is the fact that the lowest peak actually occurs during the summer season (July to September) when the average temperatures and energy demand are at the highest. The months from April to June as well as October to December yield significantly higher marginal carbon intensity at all hours. Therefore, the addition of an EV charging load in the non-summer months actually leads to greater GHG emissions. This result also implies that the mismatch between TOU incentives and GHG emission reduction (described in Chapter 4.3) is exacerbated during the non-summer months. 82 Figure 35. Average seasonal marginal carbon intensity in year 2020 The seasonal average marginal carbon intensity for LADWP’s energy grid without EV charging for the year 2030 is shown in Figure 36. The carbon intensity values now more closely follow the overall system profile pattern (see Figure 8). Unlike the 2020 emission profile, the summer months (July to September) have the highest emission rate throughout the day. Relative to the levels in 2020, the overall carbon intensity is significantly lower throughout the day. The drop in overall intensity can be directly attributed to the removal of coal generation sources in 2025. The range of values for the marginal carbon intensity, however, is much greater than that of 2020 with the values ranging from 200 to 350 kgCO2e MWh -1 (Figure 36). 500 510 520 530 540 550 560 570 580 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 kg CO 2e /MWh HOUR Jan to Mar Apr to Jun Jul to Sep Oct to Dec 83 Figure 36. Average seasonal marginal carbon intensity in year 2030 4.4.2 EV technology adoption rate The mitigation potential of EV adoption can measured by assessing the amount of GHG emissions each EV would displace from an ICEV. The baseline GHG emissions for a future ICEV in 2020 and 2030 are calculated in Chapter 4.4.1. Since the EV-related GHG emissions are directly related to the daily charging load on the electric grid, a comparison of emissions is made for a 24- hour day averaged across all four seasons. The off-peak (Case 1) and peak (Case 3) charging scenarios for the year 2020 is shown in Figure 37. In Case 1 at low EV adoption, EV-related GHG emissions is 261,803 kgCO2e per day with a mitigation potential of 147,453 kgCO 2e per day. In other words, the resulting GHG emissions for an equivalent number of ICEVs would be 409,257 kgCO2e per day (261,803 + 147,453) without any EVs. The emissions mitigation rises to 357,835 kgCO2e per day for the baseline EV adoption scenario. The potential rises even further to 720,794 kgCO2e per day in the high EV adoption scenario. In Case 2, the low EV adoption scenario has roughly the same emissions mitigation potential (150,296 kgCO 2e per day) as that of Case 1. For 150 200 250 300 350 400 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 kg CO 2e /MWh HOUR Jan to Mar Apr to Jun Jul to Sep Oct to Dec 84 the other EV adoption scenarios, however, the emissions mitigation potential is significantly higher at 452,759 and 982,395 kgCO2e per day. Overall, higher EV adoption offers significantly greater GHG emissions mitigation potential. As the adoption goes from low to baseline and baseline to high, the emissions mitigation potential more than doubles. Results indicate that the GHG emissions mitigation is the greatest in the peak charging, high EV adoption case. Figure 37. LA's EV fleet GHG emissions per day and potential mitigation of ICEV GHG emissions per day in 2020 The GHG emissions mitigation potentials for 2030 is calculated with respect to the ICEV emissions in 2030 that decreases from the rising fuel efficiency standards (i.e., CAFE). The off- peak (Case 1) and peak (Case 3) charging scenarios is shown in Figure 38. In Case 1 at low EV adoption, EV-related GHG emissions is 838,446 kgCO2e per day with a mitigation potential of 1,110,728 kgCO2e per day. In other words, the resulting GHG emissions for an equivalent number of ICEVs would be 1,949,175 kgCO2e per day (838,446 + 1,110,728). The emissions mitigation rises to 2,498,197 and 4,119,693 kgCO2e per day for the baseline and high EV adoption scenario, respectively. In Case 2, the low EV adoption scenario is 973,903 kgCO 2e per day, which is slightly Low Baseline High 0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 1 2 3 4 5 6 7 8 kg CO 2e per day Off-peak Mitigation (Off-peak) Peak Mitigation (Peak) 85 lower than that of Case 1. The emissions mitigation potentials are significantly lower at 2,231,972 and 3,836,791 kgCO 2e per day for the baseline and high EV adoption scenarios, respectively. Unlike in 2020, the peak charging cases (Case 2) result in higher emissions mitigation potential. Overall, higher EV adoption offers significantly greater GHG emissions mitigation potential (see Figure 38). Results indicate that the GHG emissions mitigation is the greatest in the off-peak charging, high EV adoption case. Figure 38. LA's EV fleet GHG emissions per day and potential mitigation of ICEV GHG emissions per day in 2030 4.4.3 Renewable energy adoption The mitigation effectiveness of LADWP’s renewable energy adoption can be best demonstrated by examining the difference in the mitigation potential in 2020 and 2030. Results for the off-peak (Case 1) and peak (Case 2) charging cases are shown in Figure 39. In the off-peak, low EV adoption case, the GHG emissions mitigation potentials rises by 963,275 kgCO 2e per day from 2020 to 2030. The rise is higher for the other cases: 2,140,363 kgCO2e per day for baseline and 3,398,899 kg CO2e per day for high EV adoption. In the peak charging case, the rise in Low Baseline High 0 1,000,000 2,000,000 3,000,000 4,000,000 5,000,000 6,000,000 7,000,000 8,000,000 9,000,000 1 2 3 4 5 6 7 8 kg CO 2e per day Off-peak Mitigation (Off-peak) Peak Mitigation (Peak) 86 mitigation from 2020 to 2030 is significantly higher for the baseline and high EV adoption cases. The potential for the baseline case is nearly equal to the potential of the high adoption scenario in the off-peak case. The potential for the high, peak charging case is nearly double the high, off- peak case. These results indicate that LADWP’s greater renewable energy adoption yields the greatest GHG emissions mitigation for the peak charging cases especially at baseline and high EV adoption scenarios. Figure 39. Change in GHG emissions mitigation potential from 2020 to 2030 4.4.4 EV charging behavior The emissions mitigation potential based on charging behavior is examined by assessing the daily EV-related GHG emissions as a function of charging start time at three different charger rates. The results shown in Figure 40 represent the GHG emissions per EV for a single charging session in 2020. That is, the hourly GHG emissions rate is calculated assuming a single charging event that begins at a given hour and ends when an EV has charged its batteries to the maximum state. The amount of depleted energy in the EV battery at the start of a charging event is assumed 0 1,000,000 2,000,000 3,000,000 4,000,000 5,000,000 6,000,000 7,000,000 Off-peak Peak kg CO 2e / per day Low Baseline High 87 constant since every EV is assumed to have the same travel intensity as other vehicles in LA. Based on the charger level (3.3, 6.6. or 10 kW), the required charging time changes. Therefore, a 3.3 kW charger would require a longer charge time than 6.6 kW or10 kW charger. The highest emission rates occur for charging sessions beginning in the early mornings (1 to 4 a.m.). The emission rates remain relatively constant during most of the day (9 a.m. to 7 p.m.). The highest emission rates for the 6.6 and 10 kW cases occur for charging starting at 3 a.m. For the 3.3 kW charger, the highest emission rate occurs for charging starting at 2 a.m. These results indicate that the highest GHG emission rates occur in the off-peak hours with some differences between the levels of chargers. Figure 40. GHG emissions per EV based on start time of charging event in 2020 The GHG emissions per EV at each charging event start time for 2030 is shown in Figure 41. Now the GHG emission rates for charging sessions at each hour is significantly lower than the rates in 2020. The highest charging rates also occur during the peak rather than off-peak hours. Furthermore, there is negligible difference between the levels of chargers in emission rates for the 5.70 5.80 5.90 6.00 6.10 6.20 6.30 6.40 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 GHG emissions per EV (kg CO 2e ) Start Time of EV Charging 3.3 kW Onboard Charger 6.6 kW Onboard Charger 10 kW Onboard Charger 88 majority of the day. These results imply that the addition of an EV charging load is preferable during the early morning hours (2 to 3 a.m.) to minimize GHG emissions per EV. Figure 41. GHG emissions per EV based on start time of charging event in 2030 4.4.5 Policy Implications The mitigation potential of the EV parameters have broad policy implications. For EV adoption, the high adoption case result in greater emissions mitigation potential in every scenario. However, the type of charging has a significant impact on the scale of mitigation. In the short run, the off-peaking charging case (Case 1) yields lower mitigation potential versus the peak charging case (Case 2) in all levels of adoption. Therefore, incentivizing off-peak charging behavior may actually be more detrimental in maximizing GHG mitigation efforts. In the long run, the mitigation potential is greater for the off-peak charging case but the difference between peak and off-peak cases is smaller than that of 2020. For renewable energy adoption, LADWP’s current trajectory of increasing use of renewable energy sources from 2020 to 2030 yields higher mitigation of GHG emissions (see Figure 39). The increase in mitigation potential from 2020 to 2030 is greater when EV charging 0 0.5 1 1.5 2 2.5 3 3.5 4 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 GHG emissions per EV (kg CO 2e ) Start Time of EV Charging 3.3 kW Onboard Charger 6.6 kW Onboard Charger 10 kW Onboard Charger 89 occurs during peak hours. These results demonstrate the importance of California’s Renewable Portfolio Standard (RPS) that mandates increasing levels of renewables in the utilities’ energy portfolios. Furthermore, the result underlines the significance of LADWP’s removal of coal generation sources in 2025. The removal of coal from its generation would inevitably create greater need of renewable generation sources that are significantly lower in carbon intensity. Results seem to reinforce the need for the acceleration in removal of coal power and adoption of renewable energy sources at utility scale. The results for EV charging behavior reinforce some of the main points described in Chapter 4.3. In the short-term, off-peaking charging yield higher emissions for all levels of chargers (see Figure 40). Off-peak charging become preferably in terms of GHG emissions in 2030 (see Figure 41). These results demonstrate that the charging decision in terms of the time of day matters in GHG emissions mitigation efforts. The short-term incentives for off-peak charging may not only result in greater emissions but also lower technology adoption which would lower the mitigation potential. Encouraging restrictive charging behavior in the short-run may be counterproductive to GHG emissions reduction policies. 5 SUMMARY AND CONCLUSION This study presents a methodology to assess the environmental effects, namely GHG emissions, of a large-scale adoption of EV technology for a particular region. The study’s area of focus is the City of LA. An extensive literature review is conducted in Chapter 2 to compare alternate approaches to assessing vehicle emissions, technology adoption, travel patterns, and grid-related GHG emissions. Different datasets from government agencies, technical reports, and academic journal 90 articles are consulted for relevant data on grid energy emissions. Local agency data is used for population data and regional travel patterns. Chapter 3 develops the model to estimate the hourly GHG emissions for the LADWP region. The model is developed using LADWP’s generation sources and recent data on system load profiles. Renewable and nonrenewable generation sources are identified and their generation capacities are quantified. A resource dispatch algorithm is created to model the generation type and scale of sources based on the system energy load. The hourly average marginal carbon intensity is modeled using the lifecycle emissions data of various generation sources. Charging scenarios are created based on off-peak versus peak charging and the region’s travel pattern data. The region’s EV sales is modeled using the Bass technology diffusion framework. The region’s working-age population is used as a surrogate measure to create the upper bound on vehicle sales. The region’s EV fleet size is estimated using the Bass output and vehicle scrap rates. The GHG emissions of ICEVs are also modeled using current regulation on fuel efficiency and carbon intensity in 2020 and 2030. Model results are demonstrated in Chapter 4. The daily EV energy load from charging requirements are calculated for 2020 and 2030. The overall system profile as a result of EV charging loads are calculated and demonstrated. The EV-related GHG emissions on an hourly marginal basis is calculated and demonstrated for 2020 and 2030. The results reveal a dramatic change in the marginal carbon intensity profile from 2020 to 2030. Emissions mitigation potentials for the three EV adoption cases, renewable energy adoption, and charging scenarios are calculated with further analysis on their importance in mitigation strategies. Some of key findings of the study highlighted the impacts of EVs on the area’s energy load and GHG emissions in the near future. A number of conclusions can be drawn from this study: 91 1) Previous concerns about EV’s cumulative charging loads on the energy grid remain modest in 2020 even in the highest EV adoption case. 2) The EV loads do pose some significant challenges in 2030 because of the greater number of EV operating in the regions and the removal of major energy generation sources by 2025. 3) The EV-related GHG emissions has a significantly different profile in 2020 compared to 2030. In 2020, the average marginal carbon intensity during off-peak hours is higher than that of peak hours. This implies that EV charging during off-peak hours results in greater GHG emissions. More importantly, the current economic incentives to encourage off- peak charging (i.e., TOU pricing) may actually be more detrimental. 4) In 2030, the average marginal carbon intensity during off-peak charging is lower than that of peak hours. Therefore, TOU pricing aligns with emission mitigation objectives. 5) Average marginal carbon intensity changes across seasons with the non-summer months having the lowest intensity in 2020 but the highest in 2030. 6) Higher EV adoption leads to greater GHG emissions mitigation potential with differences between 2020 and 2030. In 2020, peak charging leads to greater emissions mitigation. In 2030, off-peak charging leads to greater emissions mitigation. 7) The difference in GHG emissions mitigation from 2020 to 2030 is highest in the peak charging case. That is, the mitigation gains are greatest when EV charging occurs during peak hours from 2020 to 2030. 8) The time of charging matters in GHG emissions with early hours having the highest emission rates in 2020 but the lowest in 2030. There are some difference between the levels of chargers in terms of emissions depending when charging begins. 92 6 DIRECTIONS FOR FUTURE RESEARCH Figure 42. An overview of the new energy system As demonstrated in this study, increasing EV charging loads may lead to major grid capacity issues in 2030. The constrained capacity may also lead to worse emissions depending on the type and scale of imported energy and peak generation sources. Furthermore, the growth in renewable generation may amplify the intermittency issues that may lead to greater on-demand generation sources with worse emission profiles. Therefore, more research in the EV’s interaction with the energy grid is imperative in understanding the complex relationships to create the best possible strategies to reduce emissions and costs. In general, research in the related area of optimal distribution of renewable energy generation and EV-grid interactions offer some of the greatest potential in reducing EV-related emissions and grid capacity concerns. 93 More broadly, the “new” energy system offers a unique, complex challenge as more EVs are integrated into the system (see Figure 42). The energy system is changing rapidly; due to the intensifying challenges from climate change and depletion of energy resources, the expansion of renewable energy sources such as solar and wind is accelerating. In the new energy system, both renewable and nonrenewable energy sources must coexist and operate seamlessly as a single system. More EVs will put greater stress on the system. With more renewables in the energy mix, resource intermittency poses new challenges from uncertainties in reliability and quality. To remedy the problem, energy storage capacity will be expanded to store excess energy when renewable generation is plentiful. The stored energy can quickly be deployed to meet demand when renewable generation is unavailable. Then, the first major research question arises from society’s choice in the type and scale of storage capacity: utility-scale or localized. Secondly, the optimal interaction of energy loads such as EVs with these resources to reduce emissions and costs is a significant area of research. Utility-scale storage is usually conventional pumped-hydro technology that is prevalent throughout the world. However, it has limited capacity for expansion because of the necessary natural landscapes and stringent environmental regulations. An alternative method is to create localized energy storage capacity at the “micro-grid” level. This decentralized configuration requires less storage capacity and offers greater operational flexibility. 94 7 REFERENCES AFDC. 2014. Fuel Properties Comparison. Alternative Fuels Data Center, US Department of Energy. < http://www.afdc.energy.gov/fuels/fuel_properties.php> (accessed May 2014). Axsen, J. and Kurani, K. 2008. The Early U.S. Market for PHEVs: Anticipating Consumer Awareness, Recharge Potential, Design Priorities and Energy Impacts (Report: UCD -ITS-RR-08-22). Davis, CA: Institute of Transportation Studies, University of California Davis. Axsen, J. and Kurani, K. 2010. Anticipating plug-in hybrid vehicle energy impacts in California: Constructing consumer-informed recharge profiles. Transportation Research Part D 15, 212-219. Axsen, J., Kurani, K., McCarthy, R. and Yang, C. 2011. Plug-in hybrid vehicle GHG impacts in California: Integrating consumer-informed recharge profiles with an electricity-dispatch model. Energy Policy 39, 1617-1629. Bandivadekar, A., Cheah, L., Evans, C., Groode, T., Heywood, J., Kasseris, E., Kromer, M. and Malcolm, W. 2008a. Reducing the fuel use and greenhouse gas emissions of the US vehicle fleet. Energy Policy 36, 2754-2760. Bandivadekar, A., Bodek, K., Cheah, L., Evans, C., Groode, T., Heywood, J., Kasseris, E., Kromer, M., Weiss, M., 2008b. On the road in 2035: reducing transportation’s petroleum consumption and GHG emissions. Technical Report LFEE 2008-05 RP, Laboratory for Energy and the Environment, MIT, Cambridge, Massachusetts. Bass, F. 1969. A New Product Growth for Model Consumer Durables. Management Science 15(5), 215-227. Bayus, B. 1987. Forecasting Sales of New Contingent Products: An Application to the Compact Disk Market. Journal of Product Innovation Management 4(4), 243-255. Bergerson, J. and Lave, L. 2002. A Life Cycle Analysis of Electricity Generation Technologies: Health and Environmental Implications of Alternative Fuels and Technologies. Pittsburgh, PA: Carnegie Mellon Electricity Industry Center. Bretschneider, S. and Mahajan, V. 1980. Adaptive Technological Substitution Models. Technological Forecasting and Social Change 18(2), 129-139. British Energy. 2005. Environmental Product Declaration of Electricity from Torness Nuclear Power Station: Summary of Results. London, U.K.: British Energy Group. 95 BSI. 2008. PAS 2050: Specification for the Assessment of the Life-cycle Greenhouse Gas Emissions of Goods and Services. London, U.K.: British Standards Institute. Bucklin, L. and Sengupta, S. 1993. The Co-Diffusion of Complementary Innovations: Supermarket Scanners and UPC Symbols. Journal of Product Innovation Management 10(2), 148-160. Burkhardt III, J.J., Heath, G. and Cohen, E. 2012. Life cycle greenhouse gas emissions of trough and tower concentrating solar power electricity generation: Systemic review and harmonization. Journal of Industrial Ecology 16(S1), S93-S109. CADF. 2013. State and County Total Population Projections by Race/Ethnicity and Detailed Age (Report P-3). California Department of Finance, Demographic Research Unit: Sacramento, CA. Campbell, J.E., Lobell, D.B. and Field, C.B. 2009. Greater transportation energy and GHG offsets from biolelectricity than ethanol. Science 324 (5930), 1055-1057. CARB. 2010. 2008 Estimated Annual Average Emissions. Sacramento, CA: California Air Resources Board. CARB. 2011. EMFAC2011 Technical Documentation. Sacramento, CA: California Air Resources Board. CARB. 2012a. California Air Resources Board Releases Proposed Advanced Clean Car Rules. California Air Resources Board. <http://www.arb.ca.gov/newsrel/newsrelease.php?id=267> (accessed March 2012). CARB. 2012b. Final Regulation Order (Unofficial Electronic Version), Low Carbon Fuel Standard. California Air Resources Board. <http://www.arb.ca.gov/fuels/lcfs/lu_tables_11282012.pdf> (accessed May 2014). CARB. 2014. Clean Vehicle Rebate Project. California Air Resources Board. < http://www.afdc.energy.gov/laws/laws/CA/incentive/3237> (accessed March 2014). CNCDA. 2013. California Auto Outlook, Volume 9, Number 1. California New Car Dealers Association, <http://www.cncda.org/>. CSE. 2013. California Plug-in Electric Vehicle Owner Survey. Center for Sustainable Energy: San Diego, CA. <http://energycenter.org/sites/default/files/docs/nav/transportation/ cvrp/survey-results/California_Plugin_Electric_Vehicle_Driver_Survey_Results -May_2013.pdf >. Curran, M.A. 1996. Environmental Life-Cycle Assessment. New York: McGraw Hill. 96 Davis, S.C., Diegel, S.W. and Boundy, R.G. 2012. Transportation Energy Data Book Edition 31 (ORNL-6987). Oak Ridge National Laboratory: Oak Ridge, TN. Denholm, P. and Short, W. 2006. An Evaluation of Utility System Impacts and Benefits of Optimally Dispatched Plug-In Hybrid Electric Vehicles (Technical Report: NREL/TP -620-40293). Golden, CO: National Renewable Energy Laboratory, U.S. Department of Energy. DOE. 2010. The Recovery Act: Transforming America’s Transportation Sector. United States Department of Energy. <http://www.whitehouse.gov/files/documents/Battery-and -Electric-Vehicle-Report-FINAL.pdf> (accessed April 2014). DOE. 2013. United States Department of Energy. Federal Tax Credits for Electric Vehicles Purchased in or after 2010. <http://www.fueleconomy.gov/feg/taxevb.shtml> (accessed November 2013). Dolan, S.L. and Heath, G.A. 2012. Life cycle greenhouse gas emissions of utility-scale wind power electricity generation: Systemic review and harmonization. Journal of Industrial Ecology 16(S1), S136-S154. Dones, R., Bauer, C., Bolliger, R., Burger, B., Faist Emmenegger, M., Frischknecht, R., Heck, T., Jungbluth, N., Roder, A. and Tuchschmid, M. 2007. Life cycle inventories of energy systems: results for current systems in Switzerland and other UCTE countries. Ecoinvent report No. 5. Paul Scherrer Institut Villigen, Swiss Centre for Life Cycle Inventories, Dubendorf, CH. DOT. 2012. 2017 and Later Model Year Light-Duty Vehicle Greenhouse Gas Emissions and Corporate Average Fuel Economy Standards. Environmental Protection Agency and US Department of Transportation. Federal Registrar Vol.77, No. 199, 62623- 63200. Washington D.C.: National Archives and Records Administration. Easingwood, C., Mahajan, V. and Muller, E. 1983. A Nonuniform Influence Innovation Model of New Product Acceptance. Marketing Science 2(3), 273-295. ECW. 2009. Ten-year update: emissions and economic analysis of geothermal heat pumps in Wisconsin: update of geothermal analysis prepared in 2000. Energy Center of Wisconsin: Madison, WI. EIA. 2007. Voluntary reporting of greenhouse gases: Appendix F. Electricity Emission Factors (Form EIA-1605). U.S. Energy Information Administration. < http://www.eia.doe.gov/ oiaf/1605/pdf/Appendix%20F_r071023.pdf>. 97 EIA. 2012a. California Renewable Electricity Profiles 2010. Washington, DC: U.S. Energy Information Administration. <http://www.eia.gov/renewable/state/california/> (accessed March 2012). EIA. 2012b. State Electricity Profiles 2010. Report: DOE/EIA-0348(01)/2. Washington, DC: U.S. Energy Information Administration. <http://www.eia.gov/electricity/state/> (accessed March 20120). Elgowainy, A., Zhou, Y., Vyas, A.D., Mahalik, M., Santini, D. and Wang, M. 2012. Impacts of charging choices for plug-in hybrid electric vehicles in 2030 scenario. Transportation Research Record: Journal of the Transportation Research Board No. 2298, 9-17. EPA. 2010a. Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2008. Washington, DC: U.S. Environmental Protection Agency. EPA. 2011. Emissions & generation resource integrated database (eGRID) Version 1.1 Year 2007. <http://www.epa.gov/cleanenergy/energy-resources/egrid/index.html> (accessed March 2012). EPA. 2012a. Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2009. Washington, DC: U.S. Environmental Protection Agency. EPA. 2012b. Motor Vehicle Emission Simulator (MOVES): User Guide for MOVES 2010a (EPA-420-B-10-036). Washington, DC: U.S. Environmental Protection Agency. EPRI. 2007. Environmental Assessment of Plug-In Hybrid Electric Vehicles, Volume 1: Nationwide Greenhouse Gas Emissions. Energy Power Research Institute, Palo Alto, CA. EPRI. 2009. Summary of PHEV plug-in hybrid electric vehicle system analysis. Technical Report 1017676. Electric Power Research Institute, Palo Alto, CA. Fehr & Peers (Fehr). 2009. Transportation Strategic Plan – City of Los Angeles Travel Demand Model Improvement Program/Methodolgy. In: Southern California Association of Government Regional Travel Demand Model Report. FERC. 2013. Form 714 - Annual Electric Balancing Authority Area and Planning Area Report (Part 3 Schedule2). 2006-2012 Form 714 Database. Federal Energy Regulatory Commission. Fourt, L. and Woodlock, J. 1960. Early Prediction or Market Success for New Grocery Products. Journal of Marketing 25(2), 31-38. 98 Fthenakis, V., Kim, H. and Alsema, E. 2008. Emissions from photovoltaic life-cycles. Environmental Science and Technology 42(6), 2168-2174. Garland, J. and Tomashefsky, S. 1996. Electric Vehicle and Power Plant Emissions Update: South Coast Air Basin. Sacramento, CA: California Energy Commission Energy Information & Analysis Division. Grimes-Casey, H., Keoleian, G. and Willcox, B. 2008. Carbon Emission Targets for Driving Sustainable Mobility with US Light-Duty Vehicles. Environmental Science and Technology 43(3), 585-590. Hadley, S.W. and Tsvetkova, A. 2008. Potential Impacts of Plug-in Hybrid Electric Vehicles on Regional Power Generation (ORNL/TM-2007/150). Oak Ridge, TN: Oak Ridge National Laboratory. Hadley, S.W. and Tsvetkova, A. 2009. Potential Impacts of Plug-in Hybrid Electric Vehicles on Regional Power Generation. The Electricity Journal 22(10), 56-68. Hondo, H. 2005. Life-cycle GHG emission analysis of power generation systems. Energy 30 (11-12), 2042-2056. Horsky, D. 1990. A Diffusion Model Incorporating Product Benefits, Price, Income, and Information. Marketing Science 9(4), 342-365. Hsu, D.D., O’Donoughue, P., Fthenakis, V., Heath, G.A., Kim, H.C., Sawyer, P., Choi, J.K. and Turney, D.E. 2012. Life cycle greenhouse gas emissions of crystalline silicon photovoltaic electricity generation: Systemic review and harmonization. Journal of Industrial Ecology 16(S1), S122-S135. Hyman, M. 1988. The Timeliness Problem in the Application of Bass-Type New Product Growth Models to Durables Sales Forecasting. Journal of Business Research 16(1), 31-47. IER. 1997. ExternE National Implementation Germany (Contract: JOS3-CT95-0010, JOU2-CT- 0264). Berlin, Germany: European Commission. IPCC. 2003. Radiative Forcing of Climate Change. In: Third Assessment Report: Climate Change 2001.Geneva, Switzerland: United Nations Intergovernmental Panel on Climate Change. Jain, D., Mahajan, V. and Muller, E. 1991. Innovation Diffusion in the Presence of Supply Restrictions. Marketing Science 10(1), 83-90. Jansen, K., Brown, T. and Samuelsen, G.S. 2010. Emissions impacts of plug-in hybrid electric vehicle deployment on the U.S. western grid. Journal of Power Sources 195, 5409-5416. 99 Jimenez-Gonzalez, C., Kim, S. and Overcash, M. 2000. Methodology for developing gate-to- gate life cycle inventory information. International Journal of Life Cycle Assessment 5(3), 153-159. Kahn Ribeiro, S., Kobayashi, S., Beuthe, M., Gasca, J., Greene, D., Lee, D.S., Muromachi, P.J., Newton, S., Plotkin, S., Sperling, D., Wit, R. and Zhou, P.J. 2007. Transport and its infrastructure. In Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, U.K. and New York, NY. Kalish, S. 1985. A New Product Adoption Model with Price, Advertising, and Uncertainty. Management Science 31(12), 1569-1585. Kalish, S. and Lilien, G. 1986. A Market Entry Timing Model for New Technologies. Management Science 32(2), 194-205. Kamakura, W. and Balasubramanian, S. 1988. Long-term view of the diffusion of durables: A study of the role of price and adoption influence processes via tests of nested models. International Journal of Research in Marketing 5(1), 1-13. Kang, J. and Recker, W. 2009. An activity-based assessment of the potential impacts of plug-in hybrid electric vehicles on energy and emissions using 1-day travel data. Transportation Research Part D 14, 541-556. Kelly, J., MacDonald, J. and Keoleian, G. 2012. Time-dependent plug-in hybrid electric vehicle charging based on national driving patterns and demographics. Applied Energy 94, 395-405. Keoleian, G. and Spitzley, D. 2006. Life cycle based sustainability metrics. In: Abraham MA (ed.), Sustainability Science and Engineering, Vol. 1: Defining Principles. Amsterdam, The Netherlands: Elsevier, 237-159. Kim, H.C., Fthenakis, V., Choi, J.K. and Turney, D.E. 2012. Life cycle greenhouse gas emissions of thin-film photovoltaic electricity generation: Systemic review and harmonization. Journal of Industrial Ecology 16(S1), S110-S121. Kim, J.D., Rahimi, M. and Newell, J. 2012. Life-Cycle Emissions from Port Electrification: A Case Study of Cargo Handling Tractors at the Port of Los Angeles. International Journal of Sustainable Transportation 6(6), 321-337. Kintner-Meyer, M., Schneider, K. and Pratt, R. 2007. Impacts assessment of plug-in hybrid vehicles on electric utilities and regional U.S. power grids. Part 1: technical analysis. Richland, WA: Pacific Northwest National Laboratory. 100 LADWP. 2010. Green Power Annual Report. Los Angeles, CA: Los Angeles Department of Water and Power. LADWP. 2012. Power Integrated Resource Plan. Los Angeles, CA: Los Angeles Department of Water and Power. LADWP. 2013. The Los Angeles Department of Water and Power website: Facts & Figures < https://www.ladwp.com/ladwp/faces/ladwp/aboutus/a-water/a-w- factandfigures?_adf.ctrl-state=197l5ssn25_4&_afrLoop=118331583952000>. Mahajan, V. and Muller, E. 1979. Innovation Diffusion and New Product Growth Models in Marketing. Journal of Marketing 43(4), 55-68. Mahajan, V. and Peterson. R. 1979. First-Purchase Diffusion Models of New-Product Acceptance. Technological Forecasting and Social Change 15(2), 127-146. Mahajan, V. and Peterson. R. 1985. Models for Innovation Diffusion. Beverly Hills: Sage Publications. Mahajan, V., Muller, E. and Srivastava, R. 1990a. Determination of Adopter Categories by Using Innovation Diffusion Models. Journal of Marketing Research 27(1), 37-50. Mahajan, V., Muller, E. and Bass, F. 1990b. New Product Diffusion Models in Marketing: A Review and Directions for Research. Journal of Marketing 54(1), 1-26. Mansfield, E. 1961. Technical Change and the Rate of Imitation. Econometrica 29(4), 741-766. Marnay, C., Fisher, S., Murtishaw, S., Phadke, A., Price, L. and Sathaye, J. 2002. Estimating Carbon Dioxide Emissions Factors for the California Electric Power Sector. Berkeley, CA: Environmental Energy Technologies Division. McCarthy, R. and Yang, C. 2010. Determining marginal electricity for near-term plug-in and fuel cell vehicle demands in California: Impacts on vehicle greenhouse gas emissions. Journal of Power Sources 195, 2099-2109. McCollum, D. and Yang, C. 2009. Achieving deep reductions in US transport greenhouse gas emissions: Scenario analysis and policy implications. Energy Policy 37, 5580-5596. Meier, P. 2002. Life-Cycle Assessment of Electricity Generation Systems and Applications for Climate Change Policy Analysis. PhD dissertation, University of Wisconsin Madison. MIT. 2010. Electrification of the Transportation System. In: An MIT Energy Initiative Symposium. Cambridge, Massachusetts: Massachusetts Institute of Technology. <https://mitei.mit.edu/system/files/electrification-transportation-system.pdf>. 101 NAP. 2010. Electricity from Renewable Resources: Status, Prospects, and Impediments. National Academies Press. <http://www.nap.edu/openbook.php?record_id= 12619&page=67> (accessed March 2012). NHTSA. 2006. Vehicle Survivability and Travel Mileage Schedules. Technical Report DOT HS 809 952, National Highway Transportation Safety Agency. <http://www -nrd.nhtsa.dot.gov/Pubs/809952.pdf>. O’Donoughue, P.R., Heath, G.A., Dolan, S.L. and Vorum, M. 2014. Life cycle greenhouse gas emissions of electricity generated from conventionally produced natural gas: Systemic review and harmonization. Journal of Industrial Ecology 18(1), 125-144. O’Keefe, W. 2010. Electric Cars: Not Ready for Prime Time. George C. Marshall Institute Policy Outlook, December 2010. <http://www.marshall.org/pdf/materials/922.pdf> (accessed March 2012). Pacca, S. 2007. Impacts from decommissioning of hydroelectric dams: A life-cycle perspective. Climate Change 84, 281-294. Pacca, S. and Horvath, A. 2002. Greenhouse Gas emissions from building and operating electric power plants. Environmental Science and Technology 36(14), 3194-3200. Plevin, R.J. 2009. Modeling Corn Ethanol and Climate: A Climate Comparison of the BESS and GREET Models. Journal of Industrial Ecology 13(4), 495-507. Parks, K., Denholm, P. and Markel, T. 2007. Costs and emissions associated with plug-in hybrid electric vehicle charging in the Xcel Energy Colorado service territory. Techical Report NREL/TP-641-41410. National Renewable Energy Laboratory: Golden, CO. Rogers, E. 1995. Diffusion of Innovations, 4 th Ed. New York: Free Press. SAE. 2011. SAE Charging Configurations and Ratings Terminology. Society of Automotive Engineers International. < http://www.sae.org/smartgrid/chargingspeeds.pdf>. Samaras, C. and Meisterling, K. 2008. Life Cycle Assessment of Greenhouse Gas Emissions from Plug-In Hybrid Vehicles: Implications for Policy. Environmental Science and Technology 42(9), 3170-3176. SCAG. 2008. Regional Travel Demand Model and 2008 Validation, Chapter 5: Trip Distribution. Southern California Association of Governments. < http://www.scag.ca.gov/modeling/pdf/MVS08/MVS08_Chap05.pdf>. SCAG. 2012. Adopted 2008 Growth Forecast, by City. Southern California Association of Governments. < http://www.scag.ca.gov/forecast/adoptedgrowth.htm>. 102 Schmittlein, D. and Mahajan, V. 1982. Maximum Likelihood Estimation for an Innovation Diffusion Model of New Product Acceptance. Marketing Science 1(1), 57-78. Schneider, K., Gerkensmeyer, C., Kintner-Meyer, M. and Fletcher, R. 2008. Impact Assessment of Plug-In Hybrid Vehicles on Pacific Northwest Distribution Systems. IEEE Power and Energy Society General Meeting. Shiau, C., Samaras, C., Hauffe, R. and Michalek, J.J. 2009. Impact of battery weight and charging patterns on the economic and environmental benefits of plug-in hybrid vehicles. Energy Policy 37, 2653-2663. Sioshansi, R. and Denholm, P. 2009. Emissions Impacts and Benefits of Plug-In Hybrid Electric Vehicles and Vehicle-to-Grid Services. Environmental Science and Technology 43(4), 1199-1204. Slezak, L. 2013. Transportation Electrification US DOE Activities. United States Department of Energy. In: Presentation at Lehigh University on September 6, Bethlehem, Pennsylvania. < http://www.lehigh.edu/engineering/news/events/ine/pdf/Slezak.pdf>. Smith, K., Earleywine, M., Wood, E. and Pesaran, A. 2011. Comparison of Battery Life Across Real-World Automotive Drive-Cycles. In: Presentation at the “7 th Lithium Battery Power Conference”, Las Vegas, Nevada. Spath, P., Mann, M. and Kerr, D. 1999. Life-cycle Assessment of Coal-fired Power Production. Golden, CO: National Renewable Energy Laboratory, U.S. Department of Energy. Spitzley, D. and Keoleian, G. 2005. Life-cycle Environmental and Economic Assessment of Willow Biomass Electricity: A Comparison with Other Renewable and Non-Renewable Sources (Report No. CSS04-05 R). Ann Arbor, MI: Center for Sustainable Systems, University of Michigan. Srinivasan, V. and Mason, C. 1986. Nonlinear Least Square Estimation of New Product Diffusion Models. Marketing Science 5(2), 169-178. Stephan, C. and Sullivan, J. 2008. Environmental and Energy Implications of Plug-In Hybrid- Electric Vehicles. Environmental Science and Technology 42(4):1185-1190 Storm van Leeuwen, J.K. and Smith. P. 2002. Can nuclear power provide energy for the future; would it solve the CO2 emission problem? Gorningen, Netherlands: University of Groningen. Sultan, F., Farley, J. and Lehmann, D. 1990. A Meta-Analysis of Applications of Diffusion Models. Journal of Marketing Research 27(1), 70-77. 103 Tigert, D. and Farivar, B. 1981. The Bass New Product Growth Model: A Sensitivity Analysis for a High Technology Product. Journal of Marketing 45(4), 81-90. UK DTI. 2006. The Energy Challenge Energy Review Report 2006. Chapter 5: Electricity Generation. United Kingdom Department of Trade and Industry. UK SDC. 2006. The role of nuclear power in a low carbon economy. London, U.K.: Sustainable Development Commission. UNFCCC (United Nations Framework Convention on Climate Change). 2009. Methodological Tool: Tool to calculate the emission factor for an electricity system, Version 2. EB50 Report, Annex 14. United Nations. Vattenfall. 1999. Vattenfall’s Life Cycle Studies of Electricity. Stockholm, Sweden: Vattenfall. Wang, M. 2002. Fuel Choices for fuel-cell vehicles: well-to-wheels energy and emission impacts. Journal of Power Sources 112(1), 310-321. Warner, E.S. and Heath, GA. 2012. Life cycle greenhouse gas emissions of nuclear electricity generation: Systemic review and harmonization. Journal of Industrial Ecology 16 (S1), S73-S92. Weber, C., Jaramillo, P., Marriot, J. and Samaras, C. 2010. LCA and Grid Electricity: What do we know and what can we know? Environmental Science and Technology 44(6), 1895-1901. Weiller, C. 2011. Plug-in hybrid electric vehicle impacts on hourly electricity demand in the United States. Energy Policy 39, 3766-3778. Whitaker, M. and Heath, G.A., O’Donoughue, P. and Vorum, M. 2012. Life cycle greenhouse gas emissions of coal-fired electricity generation: Systematic review and harmonization. Journal of Industrial Ecology 16(S1), S53-S72. Who Killed the Electric Car? Directed by Chris Paine. 2006. Los Angeles, CA: Electric Entertainment, 2006. DVD. Williams, B., Martin, B., Lipman, T. and Kammen, D. 2011. Plug-In Hybrid Vehicle Use, Energy Consumption, and Greenhouse Emissions: An Analysis of Household Vehicle Placements in Northern California. Energies 2011(4), 435-457. Wise, M., Page Kyle, G., Dooley, J.J. and Kim, S.H. 2010. The impact of electric passenger transport technology under an economy-wide climate policy in the United States: Carbon dioxide emissions, coal use, and carbon dioxide capture and storage. International Journal of Greenhouse Gas Control 4, 301-308. 104 Wu, M., Wang, M., Liu, J. and Huo, H. 2008. Assessment of Potential Life-Cycle Energy and Greenhouse Gas Emission Effects from Using Corn-Based Butanol as a Transportation Fuel. Biotechnology Progress 24(6), 1204-1214. Yang, C., McCollum, D., McCarthy, R. and Leighty, W. 2009. Meeting an 80% reduction in greenhouse gas emissions from transportation by 2050: A case study in California. Transportation Research Part D 14, 147-156. Yeh, S., Farrell, A., Plevin, R., Sanstad, A. and Weyant, J. 2008. Optimizing U.S. Mitigation Strategies for the Light-Duty Transportation Sector: What We Learn from a Bottom-Up Model. Environmental Science and Technology 42(22), 8202-8210. Young, K., Wang, C. and Strunz, K. 2013. Electric Vehicle Battery Technology. In: Electric Vehicle Integration into Modern Power Networks. New York: Springer Science+Business Media. Ch. 2 Zoepf, S.M. 2011. Automotive Features: Mass Impact and Deployment Characterization. Master Thesis. Massachusetts Institute of Technology, Cambridge, Massachusetts. 105 8 APPENDIX 8.1 MATLAB Scripts ”PEV_LACity_rev2.m” %% Population Estimates close all clear all % Projectiong until 2060 upper = 2060; year = [2010:1:upper]; % Orange County working age (18-64) population projection from California % Department of Finance Demographic Research Unit (Report P-3) OC_WorkingAge = [1930149.25787831 1954930.45475925 1969827.64741563 1982742.81462390 1995612.61977880 2006466.75665010 2011557.59497806 ... 2014340.01657265 2013992.09645276 2011252.77386657 1999773.28054930 1993880.07804192 1986328.64254443 1977576.71657135 ... 1967787.70106248 1955938.75604841... 1943367.43950303 1929703.07483880 1916003.74713991 1902633.57905480 1892497.24892954 1884682.77083859 1877714.65810823... 1871202.27506104 1864336.25421002 1855962.32093076 1849215.99630556 1845428.83981782 1843893.39330243 1843328.81112876... 1842530.45193843 1843680.53779416 1845649.00145206 1847740.23957791 1848965.07077066 1849751.07084412 1850231.20734389... 1849715.31520809 1848675.17256102 1847746.53218306 1846035.44060912 1844718.47407159 1843407.86372329 1841204.17599427 ... 1837578.14479231 1832838.59737747 1827500.51647931 1821563.02290805 1816205.05440591 1811461.45383412 1807491.14458862]'; % Los Angeles County working age (18-64) and total population projection from California % Department of Finance Demographic Research Unit (Report P-3) LACo_WorkingAge = [6363987.99614322 6405223.48382971 6437644.63618259 6470403.19963837... 6497777.43717404 6516279.27744486 6532199.23811855 6540722.46330393 6547332.75424849... 6547263.09968394 6588561.50625007 6577986.04658102 6565919.48148966 6554059.12255589 6543083.07440009... 6528510.16212122 6516344.58487466 6501020.13527837 6483122.07139542 6467502.03307879 6458836.27177287... 6456712.80041774 6456774.15300101 6456333.70643192 6452691.09549258 6443210.78881274 6440073.47847938... 6442621.55012656 6447889.09486007 6453405.13007666 6455065.19289561 6460259.40491455 6463991.07672872... 6465732.57340444 6464332.06084355 6454322.71798099 6444377.41183582 6432277.14037938 6419296.92189089... 6406565.80443435 6389414.15374781 6372665.55203542 6357349.40325810 6341627.08362112 6324447.87341453... 6303716.59494540 6282064.90605421 6262105.54526297 6245471.99501994 6231976.53243152 6220911.51186029]; 106 LACo_Pop = [9824906.00000000 9860835.99999999 9911665.00000001 9969762.09472442... 10025578.61473420 10081144.00072320... 10137924.97714400 10190448.56631290 10246313.49846540 10302862.02391310 10441441.15318460... 10497592.03197200 10551148.85148690 10606824.71671360 10664208.49409150 10718808.62526880... 10770284.44196300 10818573.41865820 10865863.74662120 10909681.26712960 10950334.63731680... 10988115.18284590 11023910.40056690 11058335.70767670 11090308.91858810 11120283.80871660... 11147989.78691060 11173889.71488110 11198284.63690260 11221311.03652920 11243022.31656330... 11264061.86253860 11284223.01846710 11303671.12659970 11322421.23836670 11342948.15237280... 11362778.88477760 11382058.71355240 11400337.87512060 11417787.24717480 11434565.41354270... 11450614.17393570 11466219.45236510 11481080.64325800 11494999.39194820 11508036.67699250... 11520084.54435670 11531647.42391400 11542618.75756230 11552915.85155220 11562720.37693120]; Total_OC_LACo = OC_WorkingAge + LACo_WorkingAge'; OC_LAPopWorkAgeGrowth = []; LACoWorkAgeGrowth =[]; for i=2:length(Total_OC_LACo) OC_LAPopWorkAgeGrowth(i-1)= (Total_OC_LACo(i)- Total_OC_LACo(i- 1))/Total_OC_LACo(i-1); LACoWorkAgeGrowth(i-1) = (LACo_WorkingAge(i)-LACo_WorkingAge(i- 1))/LACo_WorkingAge(i-1); % Used in City of LA Fleet modeling end OC_LAPopWorkAgeGrowth = OC_LAPopWorkAgeGrowth'; %Use this growth rate for vehicle sales projections clear i % City of LA population based on SCAG (2020 and 2035) and U.S. Census % (2010 and 2012) LAPop = [3795781 3857799 3991700 4320600]; LAYear = [2010 2012 2020 2035]; % Linear fit to predict other years P_LA = polyfit(LAYear, LAPop, 1); LAPop3 = P_LA(1).*2011 + P_LA(2); LAYear2 = [2010:1:upper]; LAPop2 = P_LA(1).*LAYear2 + P_LA(2); LAPop = [LAPop(1) LAPop2(2) LAPop(2) LAPop2(4:10) LAPop(3) LAPop2(12:25) LAPop(4) LAPop2(27:end)]; LAPop = LAPop'; % Estimate City of LA's working age population assuming the percentage is % the same as LA County LA_WorkAge = (LACo_WorkingAge' ./ LACo_Pop') .* LAPop; %% CNCDA Vehicle Sales Data % The following section is only inserted for reference (further analysis) CNCDAYear = [2005:1:2012]; 107 US_Total = [17444329 17048981 16460315 13493165 10601368 11772219 13040613 14785936]; CA_Total = [1760444 1661974 1505234 1151543 821626 912074 1035756 1310720]; CA_Cars = [872252 875452 816841 692533 503879 552277 625526 823084]; CA_LightTrucks = [888192 786522 688393 459010 317747 359797 410230 487636]; OC_LACo_Total = [669916 638089 586794 467641 337035 373711 419258 522356]; OC_LACo_Cars = [358086 360836 339543 293989 216728 239074 266129 344330]; OC_LACo_LightTrucks = [311830 277253 247251 173652 120307 134637 153129 178026]; %% Total Vehicle Sales Projections based on Working Age Population Changes SalesYear = [2012:1:upper]; Sales_OC_LACo_Cars = OC_LACo_Cars(end); Sales_OC_LACo_LightTrucks = OC_LACo_LightTrucks(end); % Use "OC_LAPopWorkAgeGrowth" for projections Sales_OC_LAPopWorkAgeGrowth = OC_LAPopWorkAgeGrowth(3:end); for i=1:length(SalesYear)-1 % Year 2013 to 2060 Sales_OC_LACo_Cars(i+1) = (1+Sales_OC_LAPopWorkAgeGrowth(i))*Sales_OC_LACo_Cars(i); Sales_OC_LACo_LightTrucks(i+1) = (1+Sales_OC_LAPopWorkAgeGrowth(i))*Sales_OC_LACo_LightTrucks(i); end clear i Sales_OC_LACo_Cars = [OC_LACo_Cars(end-2:end-1) Sales_OC_LACo_Cars]'; Sales_OC_LACo_LightTrucks = [OC_LACo_LightTrucks(end-2:end-1) Sales_OC_LACo_LightTrucks]'; % Use City of LA's work age proportion to calculate vehicle sales estimate LA_WorkAge_vs_Total = LA_WorkAge./Total_OC_LACo; LA_WorkAge_vs_LACo = LA_WorkAge./LACo_WorkingAge'; %LA City's work age percentage relative to LA County (used in LA's Fleet modeling) for i=1:length(LA_WorkAge) Sales_LA_Cars = Sales_OC_LACo_Cars.*LA_WorkAge_vs_Total; Sales_LA_LightTrucks = Sales_OC_LACo_LightTrucks.*LA_WorkAge_vs_Total; end %% Projection of EV sales based on Bass diffusion modeling upper = 60; time = [0:1:upper]; %years a0 = 0; % Bass parameters m = .3.*[.65*600000 .75*600000 .9*600000]; % Vehicle sales in LA and Orange County range from 400K to 600K each year, 30% are in City of LA (by population) and 65% is in passenger cars p = [.004 0.007 0.01]; % Based on hybrid sales q = [0.15 0.2 0.25]; % Based on hybrid sales 108 fhand = @BEVAdoptLA; for i=1:length(time) y1(i) = feval(fhand,time(i),m(1),p(1),q(1)); y2(i) = feval(fhand,time(i),m(2),p(2),q(2)); y3(i) = feval(fhand,time(i),m(3),p(3),q(3)); end clear i year = [2009:1:(2009+upper)]; figure(1) plot(year,y1,'-',year,y2,'g',year,y3,'r') legend('1','2','3') y3 = y3'; y1 = y1'; y2 = y2'; y=[y1 y2 y3]; %% LA's Fleet Size Changes LACo_Cars = [5859407]; % Cars in LA County in 2008 LACo_Trucks = [1152856]; % Trukcs in LA County in 2008 for i=2:length(LACo_Pop) LACo_Cars(i) = LACo_Cars(i-1)*(1+LACoWorkAgeGrowth(i-1)); LACo_Trucks(i) = LACo_Trucks(i-1)*(1+LACoWorkAgeGrowth(i-1)); end LACo_Cars = LACo_Cars'; LACo_Trucks = LACo_Trucks'; LA_Cars = LA_WorkAge_vs_LACo.*LACo_Cars; LA_Trucks = LA_WorkAge_vs_LACo.*LACo_Trucks; %% Survival rate % Survival rate = 1-exp(-exp(a+b*age)) % Passenger Cars: % For Age <=10 %A1 = 1.64905; %B1 = -0.12143; % For Age >=0 %A2 = 3.38136; %B2 = -0.28623; life = 30; % max life expectancy of car for i=1:length(y) for j=1:life inFleet1(i,j) = (survivalCar(j))*y1(i); inFleet2(i,j) = (survivalCar(j))*y2(i); inFleet3(i,j) = (survivalCar(j))*y3(i); end end inFleet1 = [inFleet1 zeros(61,31)]; inFleet2 = [inFleet2 zeros(61,31)]; inFleet3 = [inFleet3 zeros(61,31)]; Fleet1 = []; Fleet2 = []; 109 Fleet3 = []; for k=1:length(y) temp1 = inFleet1(1:k,1:k); temp1 = fliplr(temp1); Fleet1(k) = trace(temp1); temp2 = inFleet2(1:k,1:k); temp2 = fliplr(temp2); Fleet2(k) = trace(temp2); temp3 = inFleet3(1:k,1:k); temp3 = fliplr(temp3); Fleet3(k) = trace(temp3); end for m=1:length(y)-1 FleetSize1(m) = Fleet1(m)+y1(m+1); FleetSize2(m) = Fleet2(m)+y2(m+1); FleetSize3(m) = Fleet3(m)+y3(m+1); end FleetSize1 = FleetSize1'; FleetSize2 = FleetSize2'; FleetSize3 = FleetSize3'; FleetSize = [FleetSize1 FleetSize2 FleetSize3]; %% Baseline % ICEV GHG emission for the year 2020 and 2030 EnergyContent = 122.475; %MJ/gallon CarbonIntensity = 89.06; %gCO2e/MJ FuelEfficiency = [38.3 48.7]; %MPG TravelIntensity = 33; % miles per day ICEV = EnergyContent*CarbonIntensity.*(1./FuelEfficiency)*TravelIntensity; ICEV = ICEV./1000; %kgCO2e/day % Total daily emissions from ICEV for the year 2020 and 2030 ICEV20 = FleetSize(10,:); ICEV30 = FleetSize(20,:); ICEV20_Emiss = ICEV(1).*ICEV20; ICEV30_Emiss = ICEV(2).*ICEV30; % LADWP Load Profile setyear = 2020; % set value to change the energy portfolio hour = [0:1:23]; Jan2Mar_avg = [2512.658111 2383.369222 2309.227416 2273.104222 2306.146222... 2445.614556 2678.520889 2817.615667 2976.288111 3105.388889 3193.546667... 3231.802111 3235.368111 3216.667333 3189.348333 3152.178444 3142.058778... 110 3276.506556 3389.077 3387.948222 3319.249556 3169.181 2949.022222 2711.026889]; Apr2Jun_avg = [2495.035604 2341.874505 2241.530989 2191.444286 2219.784615... 2327.91978 2447.098901 2663.122637 2873.974615 3061.751099 3219.828022... 3313.41022 3353.648462 3391.746923 3412.915275 3412.381209 3390.596374... 3321.342637 3208.067912 3231.652308 3344.786264 3223.681429 3005.123736 2726.23022]; Jul2Sep_avg = [2772.784674 2593.879457 2465.761522 2396.87337 2403.454891... 2509.0925 2649.67 2846.733913 3099.03337 3348.741413 3577.011848 3766.794565... 3911.221848 4049.197174 4164.83837 4231.860761 4232.61413 4125.59337 3918.4275... 3812.787174 3788.372717 3609.800978 3323.637609 3021.828152]; Oct2Dec_avg = [2473.300435 2344.769348 2262.04087 2223.93163 2248.163913 2368.117717... 2587.297283 2724.535109 2881.944565 3019.962609 3117.922065 3174.41587... 3195.509348 3206.286413 3208.240652 3196.87413 3230.018804 3381.388261... 3393.767935 3368.104022 3284.239022 3132.370109 2904.906196 2671.523587]; for i=1:length(Jul2Sep_avg) x1 = LADWPrev2(1000*Jan2Mar_avg(i),setyear); Jan2Mar_avg_CO2(i) = x1(1); Jan2Mar_avg_NOx(i) = x1(2); Jan2Mar_avg_SOx(i) = x1(3); x1 = LADWPrev2(1000*Apr2Jun_avg(i),setyear); Apr2Jun_avg_CO2(i) = x1(1); Apr2Jun_avg_NOx(i) = x1(2); Apr2Jun_avg_SOx(i) = x1(3); x1 = LADWPrev2(1000*Jul2Sep_avg(i),setyear); Jul2Sep_avg_CO2(i) = x1(1); Jul2Sep_avg_NOx(i) = x1(2); Jul2Sep_avg_SOx(i) = x1(3); x1 = LADWPrev2(1000*Oct2Dec_avg(i),setyear); Oct2Dec_avg_CO2(i) = x1(1); Oct2Dec_avg_NOx(i) = x1(2); Oct2Dec_avg_SOx(i) = x1(3); end plot(hour,Jan2Mar_avg_CO2,hour,Apr2Jun_avg_CO2,hour,Jul2Sep_avg_CO2,hour,Oct2 Dec_avg_CO2) legend('Jan to Mar','Apr to Jun', 'Jul to Sep', 'Oct to Dec') SeasonalMarginal = [Jan2Mar_avg_CO2;Apr2Jun_avg_CO2;Jul2Sep_avg_CO2;Oct2Dec_avg_CO2]'; %% Energy discharge rates % Assumption: The energy discharge rate for the vehicle is approximately % 0.25 kWh per mile to 0.45 kWh per mile. EnergyDischarge = 0.35; AvgDailyMiles = 33; % Average miles driven in LA AvgDailyMilesOneway = AvgDailyMiles/2; DailyDischarge = AvgDailyMiles.*EnergyDischarge; 111 OnewayDischarge = AvgDailyMilesOneway.*EnergyDischarge; %% Recharge rates % Assumption: 90% efficiency EffRecharge = 0.9; % Level 1 charging 120V/230V: % On-board 1-phase; home or office; convenience outlet; up to 2kW % cost reported as $500 - $900 but usually integrated into vehicle % A:= 120V, 12 Amps % B:= 120V, 16 Amps Lev1A = 120*12/1000*EffRecharge; % in kW Lev1B = 120*16/1000*EffRecharge; % Level 2 charging 240V/400V: % On-board 1- or 3-phase; dedicated outlets; dedicated outlet; 4-20kW % cost reported as $1000 - $3000 % A:= 240V, 20 Amps % B:= 240V, 40 Amps % C:= 240V, 60 Amps % D:= 240V, 80 Amps Lev2A = 240*20/1000*EffRecharge; Lev2B = 240*40/1000*EffRecharge; Lev2C = 240*60/1000*EffRecharge; Lev2D = 240*80/1000*EffRecharge; % Level 3 (fast) charging 480-600V or direct DC): % Off-board 3-phase; commercial filling station; dedicated EVSE; 50-100kW % cost reported $30,000 - $160,000 % DC Level 1 - typically upto 80 Amps % DC Level 2 and 3 - typically upto 200 Amps % A:= 300V, 80 Amps % B:= 350V, 80 Amps % C:= 400V, 200 Amps % D:= 450V, 200 Amps Lev3A = 300*80/1000*EffRecharge; Lev3B = 350*80/1000*EffRecharge; Lev3C = 400*200/1000*EffRecharge; Lev3D = 450*200/1000*EffRecharge; Lev = [Lev1A; Lev1B; Lev2A; Lev2B; Lev2C; Lev2D]; %% On-Board vs. DC charging % Various types of on-board charging limits % Volt = 3.3 kW % LEAF = 6.6 kW % Tesla = 10 kW % BMW i3 = 3.3 kW % Prius Plug-in = 2 kw % Toyota Rav4 = 10 kW % AC1 = 3.3 kW max allowable % AC2 = 6.6 kW max allowable % AC3 = 10 kW max allowable % 95% efficiency EffOnBoard = 0.95; 112 AC1 = 3.3*EffOnBoard; % 3.3 kW max allowable AC2 = 6.6*EffOnBoard; AC3 = 10*EffOnBoard; OnBoardLimit = [AC1; AC2; AC3]; %% Charging Profile % Valley-filling scenario (off-peak 8 p.m. to 8 a.m.) (Peak 9 a.m. to 7 % p.m.) % Total energy requirements: DailyDischarge*Number of PEVs (in KWhs) % Adoption scenario 1 = S1 % Adoption scenario 2 = S2 % Adoption scenario 3 = S3 % Peak charing: 9 a.m. to 6 p.m. % "Peak" traveling: 8 a.m. to 7 p.m. implies charging takes place between 9 % a.m. and 6 p.m. % Off-peak traveling: 8 p.m. to 7 a.m. S1 = FleetSize(:,1).*DailyDischarge; S2 = FleetSize(:,2).*DailyDischarge; S3 = FleetSize(:,3).*DailyDischarge; clear i for i=1:length(Lev) % Equal distribution, equal charging S1ValleyLev(:,i) = S1.*Lev(i); S1ValleyLevHours(:,i) = DailyDischarge/Lev(i); S2ValleyLev(:,i) = S2.*Lev(i); S2ValleyLevHours(:,i) = DailyDischarge/Lev(i); S3ValleyLev(:,i) = S3.*Lev(i); S3ValleyLevHours(:,i) = DailyDischarge/Lev(i); end clear i for i=1:length(OnBoardLimit) S1ValleyLevOnBoard(:,i) = FleetSize(:,1).*OnBoardLimit(i); S2ValleyLevOnBoard(:,i) = FleetSize(:,2).*OnBoardLimit(i); S3ValleyLevOnBoard(:,i) = FleetSize(:,3).*OnBoardLimit(i); end %% % SET YEAR HERE: 2012: k=2, 2020: k=10, 2030: k=20, 2035: k=25; k = 10; % Off-peak charging: S1ValleySteady = S1(k)/12; S2ValleySteady = S2(k)/12; S3ValleySteady = S3(k)/12; % Assumes same load throughout the offpeak hours S1ValleySteadyProfile = [[Jul2Sep_avg(1:7).*1000+S1ValleySteady] Jul2Sep_avg(8:19).*1000 [Jul2Sep_avg(20:24).*1000+S1ValleySteady]]; S2ValleySteadyProfile = [[Jul2Sep_avg(1:7).*1000+S2ValleySteady] Jul2Sep_avg(8:19).*1000 [Jul2Sep_avg(20:24).*1000+S2ValleySteady]]; S3ValleySteadyProfile = [[Jul2Sep_avg(1:7).*1000+S3ValleySteady] Jul2Sep_avg(8:19).*1000 [Jul2Sep_avg(20:24).*1000+S3ValleySteady]]; ValleySteadyProfile = [S1ValleySteadyProfile' S2ValleySteadyProfile' S3ValleySteadyProfile']; figure(1) plot(ValleySteadyProfile) 113 legend('S1','S2','S3') for j=1:min(size(ValleySteadyProfile)) for i=1:length(ValleySteadyProfile) x = LADWPrev2(ValleySteadyProfile(i,j),2010+k); ValleySteadyProfile_CO2(i,j) = x(1); ValleySteadyProfile_NOx(i,j) = x(2); ValleySteadyProfile_SOx(i,j) = x(3); end end figure(2) subplot(3,1,1), plot(ValleySteadyProfile_CO2); subplot(3,1,2), plot(ValleySteadyProfile_NOx); subplot(3,1,3), plot(ValleySteadyProfile_SOx); %% Calculation of daily emissions from EV charging (off-peak charging case) %Jan to Mar S1ValleySteadyProfile_Jan2Mar = [[Jan2Mar_avg(1:7).*1000+S1ValleySteady] Jan2Mar_avg(8:19).*1000 [Jan2Mar_avg(20:24).*1000+S1ValleySteady]]; S2ValleySteadyProfile_Jan2Mar = [[Jan2Mar_avg(1:7).*1000+S2ValleySteady] Jan2Mar_avg(8:19).*1000 [Jan2Mar_avg(20:24).*1000+S2ValleySteady]]; S3ValleySteadyProfile_Jan2Mar = [[Jan2Mar_avg(1:7).*1000+S3ValleySteady] Jan2Mar_avg(8:19).*1000 [Jan2Mar_avg(20:24).*1000+S3ValleySteady]]; ValleySteadyProfile_Jan2Mar = [S1ValleySteadyProfile_Jan2Mar' S2ValleySteadyProfile_Jan2Mar' S3ValleySteadyProfile_Jan2Mar']; for j=1:min(size(ValleySteadyProfile_Jan2Mar)) for i=1:length(ValleySteadyProfile_Jan2Mar) x = LADWPrev2(ValleySteadyProfile_Jan2Mar(i,j),2010+k); ValleySteadyProfile_Jan2Mar_CO2(i,j) = x(1); ValleySteadyProfile_Jan2Mar_NOx(i,j) = x(2); ValleySteadyProfile_Jan2Mar_SOx(i,j) = x(3); end end S1ValleySteadyProfile_Jan2Mar_Emiss = ([ones(1,7).*S1ValleySteady zeros(1,12) ones(1,5).*S1ValleySteady]*ValleySteadyProfile_Jan2Mar_CO2(:,1))/1000; S2ValleySteadyProfile_Jan2Mar_Emiss = ([ones(1,7).*S2ValleySteady zeros(1,12) ones(1,5).*S2ValleySteady]*ValleySteadyProfile_Jan2Mar_CO2(:,2))/1000; S3ValleySteadyProfile_Jan2Mar_Emiss = ([ones(1,7).*S3ValleySteady zeros(1,12) ones(1,5).*S3ValleySteady]*ValleySteadyProfile_Jan2Mar_CO2(:,3))/1000; %Apr to Jun S1ValleySteadyProfile_Apr2Jun = [[Apr2Jun_avg(1:7).*1000+S1ValleySteady] Apr2Jun_avg(8:19).*1000 [Apr2Jun_avg(20:24).*1000+S1ValleySteady]]; S2ValleySteadyProfile_Apr2Jun = [[Apr2Jun_avg(1:7).*1000+S2ValleySteady] Apr2Jun_avg(8:19).*1000 [Apr2Jun_avg(20:24).*1000+S2ValleySteady]]; S3ValleySteadyProfile_Apr2Jun = [[Apr2Jun_avg(1:7).*1000+S3ValleySteady] Apr2Jun_avg(8:19).*1000 [Apr2Jun_avg(20:24).*1000+S3ValleySteady]]; ValleySteadyProfile_Apr2Jun = [S1ValleySteadyProfile_Apr2Jun' S2ValleySteadyProfile_Apr2Jun' S3ValleySteadyProfile_Apr2Jun']; for j=1:min(size(ValleySteadyProfile_Apr2Jun)) for i=1:length(ValleySteadyProfile_Apr2Jun) x = LADWPrev2(ValleySteadyProfile_Apr2Jun(i,j),2010+k); ValleySteadyProfile_Apr2Jun_CO2(i,j) = x(1); ValleySteadyProfile_Apr2Jun_NOx(i,j) = x(2); ValleySteadyProfile_Apr2Jun_SOx(i,j) = x(3); end end 114 S1ValleySteadyProfile_Apr2Jun_Emiss = ([ones(1,7).*S1ValleySteady zeros(1,12) ones(1,5).*S1ValleySteady]*ValleySteadyProfile_Apr2Jun_CO2(:,1))/1000; S2ValleySteadyProfile_Apr2Jun_Emiss = ([ones(1,7).*S2ValleySteady zeros(1,12) ones(1,5).*S2ValleySteady]*ValleySteadyProfile_Apr2Jun_CO2(:,2))/1000; S3ValleySteadyProfile_Apr2Jun_Emiss = ([ones(1,7).*S3ValleySteady zeros(1,12) ones(1,5).*S3ValleySteady]*ValleySteadyProfile_Apr2Jun_CO2(:,3))/1000; %Oct to Dec S1ValleySteadyProfile_Oct2Dec = [[Oct2Dec_avg(1:7).*1000+S1ValleySteady] Oct2Dec_avg(8:19).*1000 [Oct2Dec_avg(20:24).*1000+S1ValleySteady]]; S2ValleySteadyProfile_Oct2Dec = [[Oct2Dec_avg(1:7).*1000+S2ValleySteady] Oct2Dec_avg(8:19).*1000 [Oct2Dec_avg(20:24).*1000+S2ValleySteady]]; S3ValleySteadyProfile_Oct2Dec = [[Oct2Dec_avg(1:7).*1000+S3ValleySteady] Oct2Dec_avg(8:19).*1000 [Oct2Dec_avg(20:24).*1000+S3ValleySteady]]; ValleySteadyProfile_Oct2Dec = [S1ValleySteadyProfile_Oct2Dec' S2ValleySteadyProfile_Oct2Dec' S3ValleySteadyProfile_Oct2Dec']; for j=1:min(size(ValleySteadyProfile_Oct2Dec)) for i=1:length(ValleySteadyProfile_Oct2Dec) x = LADWPrev2(ValleySteadyProfile_Oct2Dec(i,j),2010+k); ValleySteadyProfile_Oct2Dec_CO2(i,j) = x(1); ValleySteadyProfile_Oct2Dec_NOx(i,j) = x(2); ValleySteadyProfile_Oct2Dec_SOx(i,j) = x(3); end end S1ValleySteadyProfile_Oct2Dec_Emiss = ([ones(1,7).*S1ValleySteady zeros(1,12) ones(1,5).*S1ValleySteady]*ValleySteadyProfile_Oct2Dec_CO2(:,1))/1000; S2ValleySteadyProfile_Oct2Dec_Emiss = ([ones(1,7).*S2ValleySteady zeros(1,12) ones(1,5).*S2ValleySteady]*ValleySteadyProfile_Oct2Dec_CO2(:,2))/1000; S3ValleySteadyProfile_Oct2Dec_Emiss = ([ones(1,7).*S3ValleySteady zeros(1,12) ones(1,5).*S3ValleySteady]*ValleySteadyProfile_Oct2Dec_CO2(:,3))/1000; % Jul to Sep S1ValleySteadyProfile_Emiss = [ones(1,7).*S1ValleySteady zeros(1,12) ones(1,5).*S1ValleySteady]; S2ValleySteadyProfile_Emiss = [ones(1,7).*S2ValleySteady zeros(1,12) ones(1,5).*S2ValleySteady]; S3ValleySteadyProfile_Emiss = [ones(1,7).*S3ValleySteady zeros(1,12) ones(1,5).*S3ValleySteady]; S1ValleySteadyProfile_Emiss = S1ValleySteadyProfile_Emiss*ValleySteadyProfile_CO2(:,1); S2ValleySteadyProfile_Emiss = S2ValleySteadyProfile_Emiss*ValleySteadyProfile_CO2(:,2); S3ValleySteadyProfile_Emiss = S3ValleySteadyProfile_Emiss*ValleySteadyProfile_CO2(:,3); S1ValleySteadyProfile_Emiss = S1ValleySteadyProfile_Emiss/1000; S2ValleySteadyProfile_Emiss = S2ValleySteadyProfile_Emiss/1000; S3ValleySteadyProfile_Emiss = S3ValleySteadyProfile_Emiss/1000; % Matrix of results [Low Baseline High], each row represents season (ICEV20 % and ICEV30 are the ICEV GHG emissions in 2020 and 2030) AdoptionPotential_ValleySteady = [S1ValleySteadyProfile_Jan2Mar_Emiss, S2ValleySteadyProfile_Jan2Mar_Emiss, S3ValleySteadyProfile_Jan2Mar_Emiss;... S1ValleySteadyProfile_Apr2Jun_Emiss, S2ValleySteadyProfile_Apr2Jun_Emiss, S3ValleySteadyProfile_Apr2Jun_Emiss;... 115 S1ValleySteadyProfile_Emiss, S2ValleySteadyProfile_Emiss, S3ValleySteadyProfile_Emiss;... S1ValleySteadyProfile_Oct2Dec_Emiss, S2ValleySteadyProfile_Oct2Dec_Emiss, S3ValleySteadyProfile_Oct2Dec_Emiss;... ICEV20_Emiss; ICEV30_Emiss]; %% Peak-charging % close all % Controlled case where all load is distributed evenly % SET YEAR HERE: 2012: k=2, 2020: k=10, 2030: k=20, 2035: k=25; k = 10; S1PeakCon = S1(k)/10; S2PeakCon = S2(k)/10; S3PeakCon = S3(k)/10; %Assumes same load throughout the peak hours S1PeakConProfile = [Jul2Sep_avg(1:8).*1000 [S1PeakCon+ Jul2Sep_avg(9:18).*1000] Jul2Sep_avg(19:24).*1000]; S2PeakConProfile = [Jul2Sep_avg(1:8).*1000 [S2PeakCon+ Jul2Sep_avg(9:18).*1000] Jul2Sep_avg(19:24).*1000]; S3PeakConProfile = [Jul2Sep_avg(1:8).*1000 [S3PeakCon+ Jul2Sep_avg(9:18).*1000] Jul2Sep_avg(19:24).*1000]; PeakConProfile = [S1PeakConProfile' S2PeakConProfile' S3PeakConProfile']; figure(3) plot(PeakConProfile) legend('S1','S2','S3') % Emission Calculation for controled peak charging case for j=1:min(size(PeakConProfile)) for i=1:length(PeakConProfile) x = LADWPrev2(PeakConProfile(i,j),2010+k); PeakConProfile_CO2(i,j) = x(1); PeakConProfile_NOx(i,j) = x(2); PeakConProfile_SOx(i,j) = x(3); end end figure(4) subplot(3,1,1), plot(PeakConProfile_CO2); subplot(3,1,2), plot(PeakConProfile_NOx); subplot(3,1,3), plot(PeakConProfile_SOx); %% Calculation of daily emissions from EV charging (peak charging case) %Jan to Mar S1PeakConProfile_Jan2Mar = [Jan2Mar_avg(1:8).*1000 [S1PeakCon+ Jan2Mar_avg(9:18).*1000] Jan2Mar_avg(19:24).*1000]; S2PeakConProfile_Jan2Mar = [Jan2Mar_avg(1:8).*1000 [S2PeakCon+ Jan2Mar_avg(9:18).*1000] Jan2Mar_avg(19:24).*1000]; S3PeakConProfile_Jan2Mar = [Jan2Mar_avg(1:8).*1000 [S3PeakCon+ Jan2Mar_avg(9:18).*1000] Jan2Mar_avg(19:24).*1000]; PeakConProfile_Jan2Mar = [S1PeakConProfile_Jan2Mar' S2PeakConProfile_Jan2Mar' S3PeakConProfile_Jan2Mar']; for j=1:min(size(PeakConProfile_Jan2Mar)) for i=1:length(PeakConProfile_Jan2Mar) 116 x = LADWPrev2(PeakConProfile_Jan2Mar(i,j),2010+k); PeakConProfile_Jan2Mar_CO2(i,j) = x(1); PeakConProfile_Jan2Mar_NOx(i,j) = x(2); PeakConProfile_Jan2Mar_SOx(i,j) = x(3); end end S1PeakConProfile_Jan2Mar_Emiss = ([zeros(1,8) ones(1,10).*S1PeakCon zeros(1,6)]*PeakConProfile_Jan2Mar_CO2(:,1))/1000; S2PeakConProfile_Jan2Mar_Emiss = ([zeros(1,8) ones(1,10).*S2PeakCon zeros(1,6)]*PeakConProfile_Jan2Mar_CO2(:,2))/1000; S3PeakConProfile_Jan2Mar_Emiss = ([zeros(1,8) ones(1,10).*S3PeakCon zeros(1,6)]*PeakConProfile_Jan2Mar_CO2(:,3))/1000; %Apr to Jun S1PeakConProfile_Apr2Jun = [Apr2Jun_avg(1:8).*1000 [S1PeakCon+ Apr2Jun_avg(9:18).*1000] Apr2Jun_avg(19:24).*1000]; S2PeakConProfile_Apr2Jun = [Apr2Jun_avg(1:8).*1000 [S2PeakCon+ Apr2Jun_avg(9:18).*1000] Apr2Jun_avg(19:24).*1000]; S3PeakConProfile_Apr2Jun = [Apr2Jun_avg(1:8).*1000 [S3PeakCon+ Apr2Jun_avg(9:18).*1000] Apr2Jun_avg(19:24).*1000]; PeakConProfile_Apr2Jun = [S1PeakConProfile_Apr2Jun' S2PeakConProfile_Apr2Jun' S3PeakConProfile_Apr2Jun']; for j=1:min(size(PeakConProfile_Apr2Jun)) for i=1:length(PeakConProfile_Apr2Jun) x = LADWPrev2(PeakConProfile_Apr2Jun(i,j),2010+k); PeakConProfile_Apr2Jun_CO2(i,j) = x(1); PeakConProfile_Apr2Jun_NOx(i,j) = x(2); PeakConProfile_Apr2Jun_SOx(i,j) = x(3); end end S1PeakConProfile_Apr2Jun_Emiss = ([zeros(1,8) ones(1,10).*S1PeakCon zeros(1,6)]*PeakConProfile_Apr2Jun_CO2(:,1))/1000; S2PeakConProfile_Apr2Jun_Emiss = ([zeros(1,8) ones(1,10).*S2PeakCon zeros(1,6)]*PeakConProfile_Apr2Jun_CO2(:,2))/1000; S3PeakConProfile_Apr2Jun_Emiss = ([zeros(1,8) ones(1,10).*S3PeakCon zeros(1,6)]*PeakConProfile_Apr2Jun_CO2(:,3))/1000; %Oct to Dec S1PeakConProfile_Oct2Dec = [Oct2Dec_avg(1:8).*1000 [S1PeakCon+ Oct2Dec_avg(9:18).*1000] Oct2Dec_avg(19:24).*1000]; S2PeakConProfile_Oct2Dec = [Oct2Dec_avg(1:8).*1000 [S2PeakCon+ Oct2Dec_avg(9:18).*1000] Oct2Dec_avg(19:24).*1000]; S3PeakConProfile_Oct2Dec = [Oct2Dec_avg(1:8).*1000 [S3PeakCon+ Oct2Dec_avg(9:18).*1000] Oct2Dec_avg(19:24).*1000]; PeakConProfile_Oct2Dec = [S1PeakConProfile_Oct2Dec' S2PeakConProfile_Oct2Dec' S3PeakConProfile_Oct2Dec']; for j=1:min(size(PeakConProfile_Oct2Dec)) for i=1:length(PeakConProfile_Oct2Dec) x = LADWPrev2(PeakConProfile_Oct2Dec(i,j),2010+k); PeakConProfile_Oct2Dec_CO2(i,j) = x(1); PeakConProfile_Oct2Dec_NOx(i,j) = x(2); PeakConProfile_Oct2Dec_SOx(i,j) = x(3); end end S1PeakConProfile_Oct2Dec_Emiss = ([zeros(1,8) ones(1,10).*S1PeakCon zeros(1,6)]*PeakConProfile_Oct2Dec_CO2(:,1))/1000; 117 S2PeakConProfile_Oct2Dec_Emiss = ([zeros(1,8) ones(1,10).*S2PeakCon zeros(1,6)]*PeakConProfile_Oct2Dec_CO2(:,2))/1000; S3PeakConProfile_Oct2Dec_Emiss = ([zeros(1,8) ones(1,10).*S3PeakCon zeros(1,6)]*PeakConProfile_Oct2Dec_CO2(:,3))/1000; % Jul to Sep S1PeakConProfile_Emiss = [zeros(1,8) ones(1,10).*S1PeakCon zeros(1,6)]; S2PeakConProfile_Emiss = [zeros(1,8) ones(1,10).*S2PeakCon zeros(1,6)]; S3PeakConProfile_Emiss = [zeros(1,8) ones(1,10).*S3PeakCon zeros(1,6)]; S1PeakConProfile_Emiss = S1PeakConProfile_Emiss*PeakConProfile_CO2(:,1); S2PeakConProfile_Emiss = S2PeakConProfile_Emiss*PeakConProfile_CO2(:,2); S3PeakConProfile_Emiss = S3PeakConProfile_Emiss*PeakConProfile_CO2(:,3); S1PeakConProfile_Emiss = S1PeakConProfile_Emiss/1000; S2PeakConProfile_Emiss = S2PeakConProfile_Emiss/1000; S3PeakConProfile_Emiss = S3PeakConProfile_Emiss/1000; % Matrix of results [Low Baseline High], each row represents season (ICEV20 % and ICEV30 are the ICEV GHG emissions in 2020 and 2030) AdoptionPotential_PeakCon = [S1PeakConProfile_Jan2Mar_Emiss, S2PeakConProfile_Jan2Mar_Emiss, S3PeakConProfile_Jan2Mar_Emiss;... S1PeakConProfile_Apr2Jun_Emiss, S2PeakConProfile_Apr2Jun_Emiss, S3PeakConProfile_Apr2Jun_Emiss;... S1PeakConProfile_Emiss, S2PeakConProfile_Emiss, S3PeakConProfile_Emiss;... S1PeakConProfile_Oct2Dec_Emiss, S2PeakConProfile_Oct2Dec_Emiss, S3PeakConProfile_Oct2Dec_Emiss;... ICEV20_Emiss; ICEV30_Emiss]; %% Peak UNCONTROLLED case close all clear s1 s2 s3 s1Charge s2Charge s3Charge s1FirstCharge s1SecondCharge s2FirstCharge s2SecondCharge s3FirstCharge s3SecondCharge] EnergyDischarge = 0.35; AvgDailyMiles = 33; % Average miles driven in LA AvgDailyMilesOneway = AvgDailyMiles/2; DailyDischarge = AvgDailyMiles.*EnergyDischarge; OnewayDischarge = AvgDailyMilesOneway.*EnergyDischarge; % SCAG's travel intensity data travel_count = [21 78 75 73 222 723 1893 5141 3643 2851 2966 3366 ... 3673 3459 4467 4763 4368 4490 3608 1879 1781 1404 903 581]; travel_hourly = travel_count./(sum(travel_count)); % SET YEAR HERE: 2012: k=2, 2020: k=10, 2030: k=20, 2035: k=25; k= 10; clear i for i=1:24 s1(i) = FleetSize(k,1)*travel_hourly(i); s2(i) = FleetSize(k,2)*travel_hourly(i); s3(i) = FleetSize(k,3)*travel_hourly(i); end 118 s1 = [s1(end) s1(1:end-1)]; s2 = [s2(end) s2(1:end-1)]; s3 = [s3(end) s3(1:end-1)]; % 3 types of onboard chargers (3.3, 6.6, and 10): "OnBoardLimit" % Trip to work occurs from 12 a.m. to 2 p.m. implies that 1st charging % happens from 1 a.m. to 3 p.m. % Trip to home occurs from 2 p.m. to 12 a.m. implies that 2nd charging % happens from 4 p.m. to 12 a.m. % One way discharge in kWH (results in duration) PeakUnCont = OnewayDischarge./OnBoardLimit; IntPeakUnCont = round(PeakUnCont); clear i j for i=1:15 for j=2:3 s1FirstCharge(i,j)= 2*s1(i).*IntPeakUnCont(j).*OnBoardLimit(j); s2FirstCharge(i,j)= 2*s2(i).*IntPeakUnCont(j).*OnBoardLimit(j); s3FirstCharge(i,j)= 2*s3(i).*IntPeakUnCont(j).*OnBoardLimit(j); end end % set 3.3kW case for i=1:15 s1FirstCharge(i+1,1) = 2*[s1(i)+s1(i+1)].*OnBoardLimit(1); s2FirstCharge(i+1,1) = 2*[s2(i)+s2(i+1)].*OnBoardLimit(1); s3FirstCharge(i+1,1) = 2*[s3(i)+s3(i+1)].*OnBoardLimit(1); end s1FirstCharge(1,1) = 2*s1(1)*OnBoardLimit(1); s2FirstCharge(1,1) = 2*s2(1)*OnBoardLimit(1); s3FirstCharge(1,1) = 2*s3(1)*OnBoardLimit(1); for i=1:9 for j=2:3 s1SecondCharge(i,j) = 2*s1(i+15).*IntPeakUnCont(j).*OnBoardLimit(j); s2SecondCharge(i,j) = 2*s2(i+15).*IntPeakUnCont(j).*OnBoardLimit(j); s3SecondCharge(i,j) = 2*s3(i+15).*IntPeakUnCont(j).*OnBoardLimit(j); end end % set 3.3kW case s1(25) = 2*s1(1); s2(25) = 2*s2(1); s3(25) = 2*s3(1); for i=1:9 s1SecondCharge(i+1,1) = 2*[s1(i+15)+s1(i+16)].*OnBoardLimit(1); s2SecondCharge(i+1,1) = 2*[s2(i+15)+s2(i+16)].*OnBoardLimit(1); s3SecondCharge(i+1,1) = 2*[s3(i+15)+s3(i+16)].*OnBoardLimit(1); end s1SecondCharge(1,1) = 2*s1(16)*OnBoardLimit(1); s2SecondCharge(1,1) = 2*s2(16)*OnBoardLimit(1); s3SecondCharge(1,1) = 2*s3(16)*OnBoardLimit(1); s1Charge = [s1FirstCharge(1:end-1,:); s1SecondCharge(1:end-1,:)]; s2Charge = [s2FirstCharge(1:end-1,:); s2SecondCharge(1:end-1,:)]; s3Charge = [s3FirstCharge(1:end-1,:); s3SecondCharge(1:end-1,:)]; 119 s1Charge(:,1) = [s1FirstCharge(1,1)+s1SecondCharge(end,1); s1FirstCharge(2:end-1,1); s1FirstCharge(end,1)+s1SecondCharge(1,1); s1SecondCharge(2:end-1,1)]; s2Charge(:,1) = [s2FirstCharge(1,1)+s2SecondCharge(end,1); s2FirstCharge(2:end-1,1); s2FirstCharge(end,1)+s2SecondCharge(1,1); s2SecondCharge(2:end-1,1)]; s3Charge(:,1) = [s3FirstCharge(1,1)+s3SecondCharge(end,1); s3FirstCharge(2:end-1,1); s3FirstCharge(end,1)+s3SecondCharge(1,1); s3SecondCharge(2:end-1,1)]; figure(1) subplot(3,1,1), plot(s1Charge) subplot(3,1,2), plot(s2Charge) subplot(3,1,3), plot(s3Charge) S1PeakUnConProfile = [1000*Jul2Sep_avg'+s1Charge(:,1) 1000*Jul2Sep_avg'+s1Charge(:,2) 1000*Jul2Sep_avg'+s1Charge(:,3)]; S2PeakUnConProfile = [1000*Jul2Sep_avg'+s2Charge(:,1) 1000*Jul2Sep_avg'+s2Charge(:,2) 1000*Jul2Sep_avg'+s2Charge(:,3)]; S3PeakUnConProfile = [1000*Jul2Sep_avg'+s3Charge(:,1) 1000*Jul2Sep_avg'+s2Charge(:,2) 1000*Jul2Sep_avg'+s3Charge(:,3)]; figure(2) plot(S1PeakUnConProfile) %Low adoption case figure(3) plot(S2PeakUnConProfile) %Baseline adoption case figure(4) plot(S3PeakUnConProfile) %High adoption case % Emission Calculation for UNCONTROLLED peak peak charging case for j=1:min(size(s1Charge)) for i=1:length(s1Charge) x = LADWPrev2(S1PeakUnConProfile(i,j),2010+k); S1PeakUnConProfile_CO2(i,j) = x(1); S1PeakUnConProfile_NOx(i,j) = x(2); S1PeakUnConProfile_SOx(i,j) = x(3); end end for j=1:min(size(s2Charge)) for i=1:length(s2Charge) x = LADWPrev2(S2PeakUnConProfile(i,j),2010+k); S2PeakUnConProfile_CO2(i,j) = x(1); S2PeakUnConProfile_NOx(i,j) = x(2); S2PeakUnConProfile_SOx(i,j) = x(3); end end for j=1:min(size(s3Charge)) for i=1:length(s3Charge) x = LADWPrev2(S3PeakUnConProfile(i,j),2010+k); S3PeakUnConProfile_CO2(i,j) = x(1); S3PeakUnConProfile_NOx(i,j) = x(2); S3PeakUnConProfile_SOx(i,j) = x(3); end 120 end figure(5) subplot(3,1,1), plot(S1PeakUnConProfile_CO2); subplot(3,1,2), plot(S2PeakUnConProfile_CO2); subplot(3,1,3), plot(S3PeakUnConProfile_CO2); %% Calculation of daily emissions from EV charging (uncontrolled charging) %Use variable SeasonalMarginal to estimate GHG emissions from adding a single EV to %the grid assuming a certain start time to charging % EnergyDischarge = 0.35; % AvgDailyMiles = 33; % Average miles driven in LA % DailyDischarge = AvgDailyMiles.*EnergyDischarge; % OnBoardLimit = [AC1; AC2; AC3]; ChargeTimeAC1 = DailyDischarge/OnBoardLimit(1); ChargeTimeAC2 = DailyDischarge/OnBoardLimit(2); ChargeTimeAC3 = DailyDischarge/OnBoardLimit(3); partialChargeAC1 = mod(ChargeTimeAC1,floor(ChargeTimeAC1)); partialChargeAC2 = mod(ChargeTimeAC2,floor(ChargeTimeAC2)); partialChargeAC3 = mod(ChargeTimeAC3,floor(ChargeTimeAC3)); ChargeAC1 = [ones(1,floor(ChargeTimeAC1)) partialChargeAC1 zeros(1, 24- (ceil(ChargeTimeAC1)))]; ChargeAC2 = [ones(1,floor(ChargeTimeAC2)) partialChargeAC2 zeros(1, 24- (ceil(ChargeTimeAC2)))]; ChargeAC3 = [ones(1,floor(ChargeTimeAC2)) partialChargeAC3 zeros(1, 24- (ceil(ChargeTimeAC3)))]; for i=1:(24-ceil(ChargeTimeAC1)) temp1 = [zeros(1,i) ones(1,floor(ChargeTimeAC1)) partialChargeAC1 zeros(1, 24-(ceil(ChargeTimeAC1)+i))]; ChargeAC1 = [ChargeAC1;temp1]; end for i=1:(24-ceil(ChargeTimeAC2)) temp2 = [zeros(1,i) ones(1,floor(ChargeTimeAC2)) partialChargeAC2 zeros(1, 24-(ceil(ChargeTimeAC2)+i))]; ChargeAC2 = [ChargeAC2;temp2]; end for i=1:(24-ceil(ChargeTimeAC3)) temp3 = [zeros(1,i) ones(1,floor(ChargeTimeAC3)) partialChargeAC3 zeros(1, 24-(ceil(ChargeTimeAC3)+i))]; ChargeAC3 = [ChargeAC3;temp3]; end ChargeAC1(22,:) = [ partialChargeAC1 zeros(1,20) 1 1 1]; ChargeAC1(23,:) = [1 partialChargeAC1 zeros(1,20) 1 1]; ChargeAC1(24,:) = [1 1 partialChargeAC1 zeros(1,20) 1]; ChargeAC1 = ChargeAC1.*OnBoardLimit(1); ChargeAC2(24,:) = [partialChargeAC2 zeros(1,22) 1]; ChargeAC2 = ChargeAC2.*OnBoardLimit(2); 121 ChargeAC3(24,:) = [partialChargeAC3 zeros(1,22) 1]; ChargeAC3 = ChargeAC3.*OnBoardLimit(3); for i=1:24 for j=1:4 ChargeAC1_Emiss(i,j) = ChargeAC1(i,:)*SeasonalMarginal(:,j); ChargeAC2_Emiss(i,j) = ChargeAC2(i,:)*SeasonalMarginal(:,j); ChargeAC3_Emiss(i,j) = ChargeAC3(i,:)*SeasonalMarginal(:,j); end end ChargeAC1_Emiss = ChargeAC1_Emiss./1000; ChargeAC2_Emiss = ChargeAC2_Emiss./1000; ChargeAC3_Emiss = ChargeAC3_Emiss./1000; “GenSourcesLADWPrev2.m” % LADWP Generation Sources (kW) %NaturalGas NG_HYGS = 1525000; NG_VGS = 556000; NG_HGS = 452000; NG_SGS = 796000; NG = NG_HYGS + NG_VGS + NG_HGS + NG_SGS; NG = [NG NG NG]; % Coal Coal_Navajo = [159000 159000 159000]; Coal_IPP = [401553 401553 36000 36000 149500 149500]; Coal = sum(Coal_Navajo) + sum(Coal_IPP); % Coal = [2013-2015 End of 2015 End of 2025] Coal = [Coal (Coal-sum(Coal_Navajo)) 0]; % Nuclear (ends in 2045) Nuclear_PV = [74727 74898 74784 51916 52034 51955]; Nuclear = sum(Nuclear_PV); Nuclear = [Nuclear Nuclear Nuclear]; % Large Hydro LH_CGP = 1175000; LH_Hoover = 468000; % Expires on Sept. 30, 2017 LH = LH_CGP + LH_Hoover; LH = [LH LH LH]; % Wind WindCap_2012 = [82200 72000 68695 120000 185000 262200 15000 50000 102000]; WindNetDep_2012 = [8220 36000 34000 12000 18500 73000 1500 25000 10200]; WindCap_2020 = [120000]; WindNetDep_2020 = [12000]; WindCap_2030 = 0; WindNetDep_2030 = 0; 122 WindCap = [sum(WindCap_2012) sum(WindCap_2012)+sum(WindCap_2020) sum(WindCap_2012)+sum(WindCap_2020)+sum(WindCap_2030)]; WindNet = [sum(WindNetDep_2012) sum(WindNetDep_2012)+sum(WindNetDep_2020) sum(WindNetDep_2012)+sum(WindNetDep_2020)+sum(WindNetDep_2030)]; % Solar SolarCap_2012 = [2100 54000]; SolarNetDep_2012 = [567 14580]; SolarCap_2020 = [337000 200000 200000 350000]; SolarNetDep_2020 = [90990 54000 54000 94500]; SolarCap_2030 = [159000]; SolarNetDep_2030 = [42930]; SolarCap = [sum(SolarCap_2012) sum(SolarCap_2012)+sum(SolarCap_2020) sum(SolarCap_2012)+sum(SolarCap_2020)+sum(SolarCap_2030)]; SolarNet = [sum(SolarNetDep_2012) sum(SolarNetDep_2012)+sum(SolarNetDep_2020) sum(SolarNetDep_2012)+sum(SolarNetDep_2020)+sum(SolarNetDep_2030)]; % Small hydro SH_SH = 134930; SH_MWD = 8540; SH_Castaic = 30000; SH_PSPP = 1000; SH = SH_SH + SH_MWD + SH_Castaic + SH_PSPP; SH = [SH SH SH]; % Biomass/Landfill/Biogas BLG_Hyperion = 16000; BLG_Lopez = 1500; BLG_WM = 6400; BLG_Toyon = 3600; BLG_LD2020 = 60000; BLG = BLG_Hyperion+BLG_Lopez+BLG_WM+BLG_Toyon; BLG = [BLG BLG+BLG_LD2020 BLG+BLG_LD2020]; % Geothermal GEO = 0; GEO_SSFrink2020 = 90000; GEO_Inyo2020 = 13500; GEO_Gen2020 = 47700; GEO2020 = GEO_SSFrink2020 + GEO_Inyo2020 + GEO_Gen2020; GEO_Inyo2030 = 45000; GEO_Gen2030 = 45000; GEO2030 = GEO2020 + GEO_Inyo2030 + GEO_Gen2030; GEO = [GEO GEO2020 GEO2030]; % DG Generation DG = 45000; DG = [DG DG DG]; % Nonrenewable and renewable NonRen = NG+Coal+Nuclear+LH; Ren = WindNet + SolarNet + SH + BLG + GEO + DG; 123 TotalGenCap = NonRen + Ren; % Generation Capacity in matrix form: % Generation = [SolarNet; WindNet; Nuclear; Coal; GEO; BLG; DG; NG; SH; LH]; Generation = [Nuclear; Coal; SolarNet; WindNet; GEO; BLG; DG; NG; SH; LH]; % Natural gas plants based on 2011 data Haynes_CO2_2011 = 1309826; Haynes_Gen = 3346902; Haynes_CO2 = Haynes_CO2_2011/Haynes_Gen; %in metric tons of CO2e/MWh Haynes_NOx = 95.76/4167111; Haynes_SOx = 10.65/4167111; ValleyGen_CO2_2011 = 453831; ValleyGen_Gen = 1233261; ValleyGen_CO2 = ValleyGen_CO2_2011/ValleyGen_Gen; ValleyGen_NOx = 72.51/2564342; ValleyGen_SOx = 5.6/2564342; Harbor_CO2_2011 = 56080; Harbor_Gen = 91290; Harbor_CO2 = Harbor_CO2_2011/Harbor_Gen; Harbor_NOx = 28.03/152309; Harbor_SOx = 0.53/152309; Scattergood_CO2_2011 = 505291; Scattergood_Gen = 890372; Scattergood_CO2 = Scattergood_CO2_2011/Scattergood_Gen; Scattergood_NOx = 15.27/1011980; Scattergood_SOx = 58.87/1011980; %NG_CO2 = sum([Haynes_CO2_2011 ValleyGen_CO2_2011 Harbor_CO2_2011 Scattergood_CO2_2011])/sum([Haynes_Gen ValleyGen_Gen Harbor_Gen Scattergood_Gen]); NG_CO2 = 506; % Based on harmonized LCA, NGCT and NGCC, mean value of 506 g/kWh (NREL, Journal of Industrial Ecology 2012) NG_NOx = (95.76+72.51+28.03+15.27)/(4167111+2564342+152309+1011980); % Based on specific plant data to EPA NG_SOx = (10.65+5.6+0.53+58.87)/(4167111+2564342+152309+1011980); % Based on specific plant data to EPA % Coal plants based on 2011 data Navajo_CO2_2011 = 16928813; Navajo_Gen = 16951775; Navajo_CO2 = Navajo_CO2_2011/Navajo_Gen; Navajo_NOx = 30500.62/16140683; Navajo_SOx = 4596.16/16140683; Intermountain_CO2_2011 = 11843842; Intermountain_Gen = 13002872; Intermountain_CO2 = Intermountain_CO2_2011/Intermountain_Gen; Intermountain_NOx = 24814.83/13555580; Intermountain_SOx = 5516.03/13555580; 124 % Coal_CO2 = sum([Intermountain_CO2_2011 Navajo_CO2_2011])/sum([Intermountain_Gen Navajo_Gen]); Coal_CO2 = 1050; % Based on harmonized LCA, all technologies (NREL, Journal of Industrial Ecology 2012) Coal_NOx = (30500.62+24814.83)/(16140683+13555580); Coal_SOx = (4596.16+5514.03)/(16140683+13555580); % Solar plant based on LCA studies Solar_CO2 = 0.05; Solar_NOx = 0.000180; Solar_SOx = 0.00037; DG_CO2 = 0.05; DG_NOx = 0.000180; DG_SOx = 0.00037; % Wind plant based on LCA studies Wind_CO2 = 0.014; Wind_NOx = 0.000048; Wind_SOx = 0.00032; % Small hydro based on LCA studies SH_CO2 = 0.011; SH_NOx = 0.000074; SH_SOx = 0.000027; % Geothermal based on LCA studies Geo_CO2 = .12; Geo_NOx = 0; Geo_SOx = 0; % Biomass/landfill/biogas based on LCA studies BLG_CO2 = .031; BLG_NOx = 0.00065; BLG_SOx = 0.00037; % Nuclear based on LCA studies Nuclear_CO2 = .018; % Based on harmonized LCA, LWR (NREL, Journal of Industrial Ecology 2012) Nuclear_NOx = 0.00007; Nuclear_SOx = 0.000032; % Large hydro based on LCA studies LH_CO2 = 0.24; LH_NOx = 0.00006; LH_SOx = 0.00004; % Emissions in matrix form: CO2, NOx, SOx % Nuclear, Coal, GEO, BLG, Wind, Solar, DG, SH, LH, NG % EmissFactMat = [Solar_CO2 Solar_NOx Solar_SOx; Wind_CO2 Wind_NOx Wind_SOx; Nuclear_CO2 Nuclear_NOx Nuclear_SOx; Coal_CO2 Coal_NOx Coal_SOx; Geo_CO2 Geo_NOx Geo_SOx; BLG_CO2 BLG_NOx BLG_SOx; ... % DG_CO2 DG_NOx DG_SOx; NG_CO2 NG_NOx NG_SOx; SH_CO2 SH_NOx SH_SOx; LH_CO2 LH_NOx LH_SOx]; 125 EmissFactMat = [Nuclear_CO2 Nuclear_NOx Nuclear_SOx; Coal_CO2 Coal_NOx Coal_SOx; Solar_CO2 Solar_NOx Solar_SOx; Wind_CO2 Wind_NOx Wind_SOx; Geo_CO2 Geo_NOx Geo_SOx; BLG_CO2 BLG_NOx BLG_SOx; ... DG_CO2 DG_NOx DG_SOx; NG_CO2 NG_NOx NG_SOx; SH_CO2 SH_NOx SH_SOx; LH_CO2 LH_NOx LH_SOx]; 8.2 MATLAB Functions “BEVAdoptLA .m” function y = BEVAdoptLA(t,m,p,q) y = m*(1-exp(-t*(p+q)))/(1+(q/p)*exp(-t*(p+q))); “LADWPrev2.m ” function EmissionFactor = LADWPrev2(EnerLoad, year) run('GenSourcesLADWPrev2') if year == 2012 GenAug = Generation(:,1); elseif year == 2020; GenAug = Generation(:,2); else GenAug = Generation(:,3); end GenProf = zeros(size(GenAug))'; temp = EnerLoad; i = 0; while temp >= 0 i = i+1; if (temp - GenAug(i)) < 0 GenProf(i) = temp/EnerLoad; temp = temp-GenAug(i); elseif i==length(GenProf) GenProf(i) = temp/EnerLoad; temp = -1; else GenProf(i) = GenAug(i)/EnerLoad; temp = temp-GenAug(i); end end EmissionFactor = GenProf*EmissFactMat; 126

Abstract (if available)

Abstract Greenhouse gas (GHG) emissions reduction has become an important component of policy decisions on transportation systems design, research and development, and implementation. Particularly in major urban centers, increasing use of electric vehicles (EVs) is being encouraged to support the overall objective of reduction in transportation emissions. This encouragement ranges from consumer tax credits to research and development grants for advanced EV technologies. Assessing the net effect EVs on actual emission mitigation potential, however, depends on three main factors: 1) energy portfolio of power providers, 2) consumer adoption rate, and 3) battery charging patterns. Unfortunately, the current U.S. energy grid is predominantly composed of coal‐fired plants that emit high concentrations of GHGs. Therefore, EVs essentially push emissions upstream to the electricity generation sources. EVs represent a dramatic paradigm shift in transportation such that forecasting their adoption requires adaptations to the innovation diffusion models. The charging patterns also affect the emission mitigation potential because the use of “peak” versus “off-peak” power changes the grid energy emissions significantly. This study seeks to quantify the emissions mitigation potential of these three main influencing factors. In order to answer the main research questions, an integrated emissions model is developed for the City of Los Angeles. The model incorporates modules such as changes in population and mobility patterns, consumer technology adoption, vehicle charging patterns, and lifecycle emissions of GHGs from electricity generation. Some of the model’s main outputs are the daily EV energy loads, daily system load profile, hourly average marginal grid energy carbon intensity, and the types of energy generation dispatched at every hour. For 2020, model results show that the EV charging loads will be modest with negligible effects on the overall system load profile. Results indicate that high EV adoption results in greater emissions mitigation potential. However, the type of charging has a significant impact on the scale of mitigation at all levels of adoption. Contrary to previous study results, the average marginal carbon intensity is higher if EV charging occurs during off‐peak hours. These results demonstrate that the charging decision in terms of the time of day matters in GHG emissions mitigation efforts. The short‐term incentives for off‐peak charging may not only result in greater emissions but also deter EV technology adoption which would lower the overall emissions mitigation potential. Encouraging restrictive charging behavior in the short‐run may be counterproductive to GHG emissions reduction policies. Model results for 2030 show that EV charging loads increase significantly resulting in potential generation shortages. There are also significant grid operation challenges as the region’s energy grid is required to ramp up and down rapidly to meet EV loads. For 2030, the average marginal carbon intensity for off‐peak charging becomes lower than peak charging. Increasing use of renewable generation sources leads to greater GHG emissions mitigation but the greatest effect arises from the removal of coal generation sources. The study concludes with remarks on further research into the optimal distribution of renewable energy and EV‐grid interactions as major research areas to enhance understanding of the EV’s effectiveness as GHG emissions mitigating technology.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

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

PDF

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

PDF

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

PDF

Using demand-side management for decarbonization: developing methods to quantify the impact of altering electricity consumption patterns

PDF

A parametric study of the thermal performance of green roofs in different climates through energy modeling

PDF

Identifying and mitigating the effects of urban heat islands in California

Asset Metadata

Creator Kim, Jae Duk (author)

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

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Industrial and Systems Engineering

Publication Date 07/11/2014

Defense Date 06/06/2014

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

Tag electric grid,electric vehicle (EV),emissions mitigation,energy load,greenhouse gas (GHG) emissions,OAI-PMH Harvest,Transportation

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Rahimi, Mansour (committee chair), Moore, James Elliott, II (committee member), Schweitzer, Lisa (committee member)

Creator Email jaedkim@live.com,jaedkim@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-435882

Unique identifier UC11286948

Identifier etd-KimJaeDuk-2655.pdf (filename),usctheses-c3-435882 (legacy record id)

Legacy Identifier etd-KimJaeDuk-2655.pdf

Dmrecord 435882

Document Type Dissertation

Format application/pdf (imt)

Rights Kim, Jae Duk

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