Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Impact of occupants in building performance: extracting information from building data
(USC Thesis Other)
Impact of occupants in building performance: extracting information from building data
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Impact of Occupants in Building Performance:
Extracting Information from Building Data
By
Yun Kim Cheong
Presented to the
FACULTY OF THE
SCHOOL OF ARCHITECTURE
UNIVERSITY OF SOUTHERN CALIFORNIA
In partial fulfillment of the
Requirements of degree
MASTER OF BUILDING SCIENCE
MAY 2016
2
THESIS COMMITTEE
Joon-Ho Choi
Assistant Professor
USC School of Architecture
joonhoch@usc.edu
Marc Schiler
Professor
USC School of Architecture
marcs@usc.edu
Kyle Konis
Assistant Professor
USC School of Architecture
konis@usc.edu
3
ACKNOWLEDGEMENTS
The author is very grateful to Prof Joon-Ho Choi for his advice and guidance. Much gratitude is
given to Prof Marc Schiler and Prof Kyle Konis for sharing their experience in research and
academia. Thanks are due as well to the MBS 2016 corner comprising of Mohammad Al-jammaz,
Shinjini Battacharjee, Kelly Burkhart, Edna Catumbela, Dennis Chow, Lisha Deng, Aditya
Dharane, Qian Qian Fan, Joyce Hahn, Yin Kai, Brittany Moffett, Jehyun Moon, Maral Rahmani,
and Illaria Toldo. All were instrumental in the development of this thesis.
4
LIST OF ABBREVIATIONS
ASHRAE American Society of Heating, Refrigerating and Air-Conditioning Engineers
AHU Air Handling Unit
BMS Building Management System
CFM Cubic Feet per Minute
BTU British thermal unit
CDD Cooling Degree Days
DOE Department of Energy
DV Dependent Variable
EPA Environmental Protection Agency
EUI Energy Use Intensity
GJ Gigajoule
HDD Heating Degree Days
HVAC Heating, Ventilation, and Air-Conditioning
IEQ Indoor Environmental Quality
IV Independent Variable
kWh kilowatt-hour
MLR Multiple Linear Regression
POE Post Occupancy Evaluation
Quad Quadrillion Btu (10^15 Btu)
SCOPES Smart Cameras Object Position Estimation System
THSW Temperature/Humidity/Sun/Wind
TMY2 Typical Meteorological Year 2 (TMY2)
USA United States of America
UV Ultra Violet
W Watt
Wh Watt-hour
5
TABLE OF CONTENTS
Acknowledgements ......................................................................................................................... 3
List of Abbreviations ...................................................................................................................... 4
Table of Contents ............................................................................................................................ 5
List of Figures ................................................................................................................................. 8
List of Tables ................................................................................................................................ 11
Abstract ......................................................................................................................................... 12
Chapter 1 : INTRODUCTION...................................................................................................... 13
1.1 An Introduction to Building Energy Consumption in the United States ............................ 13
1.2 Factors affecting Building Energy Consumption ................................................................ 14
1.2.1 Climate.......................................................................................................................... 15
1.2.2 Building-related characteristics .................................................................................... 15
1.2.3 Building Systems Service and Operations .................................................................... 17
1.2.4 User Presence (Occupants) ........................................................................................... 17
1.3 Building Energy Performance vs. Occupancy (human factors) .......................................... 18
1.4 Why occupancy should be identified in building energy simulation .................................. 19
1.5 Data-driven Approach to Energy Use Estimation ............................................................... 20
1.6 Objectives of this research .................................................................................................. 21
1.7 Hypothesis Statement .......................................................................................................... 22
Chapter Summary ...................................................................................................................... 22
Chapter 2 : BACKGROUND AND LITERATURE REVIEW .................................................... 23
2.1 Background on Occupancy affecting Energy Performance ................................................ 23
2.2 Background on Energy Simulation ..................................................................................... 24
2.3 Background on Mining Building-related Data .................................................................... 25
2.3.1 Regression Model for Energy Performance Calculation .............................................. 26
2.3.2 Other Statistical Methods for analyzing data for energy consumption ........................ 29
Chapter Summary ...................................................................................................................... 30
Chapter 3 : METHODOLOGY ..................................................................................................... 31
Overview ................................................................................................................................... 31
3.1 Methodology ....................................................................................................................... 31
3.1.1 Problem Definition and Objective Setting.................................................................... 32
6
3.1.2 Data Source Selection ................................................................................................... 32
3.1.3 Installing Data Collection Tools ................................................................................... 33
3.1.4 Data Monitoring and Collection ................................................................................... 39
3.1.5 Data Pre-Processing ...................................................................................................... 40
3.1.6 Data Mining .................................................................................................................. 40
3.1.7. Results and Document Presentation ............................................................................ 45
3.2 Gantt Chart .......................................................................................................................... 45
Chapter 4 : DATA COLLECTED ................................................................................................ 46
Overview ................................................................................................................................... 46
4.1 Climate Data ........................................................................................................................ 46
a) Outdoor Temperature in degree Celsius ............................................................................ 46
b) Outdoor Solar Radiation (Watt/m
2
)................................................................................... 47
c) Outdoor Wind Direction .................................................................................................... 50
4.2 Indoor Conditions – Building Management System Data & Loggers ................................ 51
a) Zonal Temperature from 17 zones .................................................................................... 51
b) Indoor Spot Measurement of Temperature and Relative Humidity .................................. 52
c) Indoor Spot Measurement of Lighting Intensity (lux) ...................................................... 52
d) Indoor Spot Measurement of Carbon Dioxide Levels (ppm) ............................................ 53
4.3 Sub-metering Data............................................................................................................... 54
a) Pie Chart for distribution of total energy use by end-use category ................................... 54
b) Cumulative Energy Consumption ..................................................................................... 54
c) Sub-metering Data – AHU Fan ......................................................................................... 55
d) Sub-metering Data – Water Heating ................................................................................. 56
e) Sub-metering Data – Lighting ........................................................................................... 57
f) Sub-metering Data – Kitchen Equipment .......................................................................... 58
g) Sub-metering Data – Refrigerator ..................................................................................... 59
h) Sub-metering Data – Office Equipment ............................................................................ 60
i) Sub-metering Data – Computer ......................................................................................... 61
j) Sub-metering Data – Total Energy Consumption .............................................................. 62
4.4 Occupancy Data .................................................................................................................. 62
a) Occupant Data from Office Administration ...................................................................... 62
7
b) Occupant Data from Occupant Sensor .............................................................................. 63
Chapter Summary ...................................................................................................................... 63
Chapter 5 : DATA ANALYSIS .................................................................................................... 64
5.1 Box Plot ............................................................................................................................... 64
5.2 Aggregated Data .................................................................................................................. 66
5.3 Correlation Analysis ............................................................................................................ 68
a) Correlation between Occupant and Total Energy Consumption ....................................... 68
b) Correlation between Occupants and Energy Use by End-Use Category .......................... 69
c) Correlation between Climate and Energy Consumption ................................................... 70
5.4 Simple Linear Regression ................................................................................................... 70
a) Occupants vs. Total Energy Use ....................................................................................... 71
b) Outdoor Temperature and Energy Use .............................................................................. 73
c) Solar Radiation and Energy Use ........................................................................................ 73
d) Outdoor Relative Humidity and Energy Use .................................................................... 74
5.5 Multiple Linear Regression ................................................................................................. 75
5.5.1 Stepwise Regression Example ...................................................................................... 75
5.5.2 Stepwise Regression Results Summary ........................................................................ 79
5.5.3 Stepwise Regression Results with Occupant Time Lag ............................................... 83
Chapter 6 : CONCLUSIONS ........................................................................................................ 85
6.1 Study Limitations ................................................................................................................ 87
APPENDIX A: STEPWISE RESULTS WITH P-VALUES........................................................ 88
APPENDIX B: STEPWISE RESULTS FROM TIME LAG WITH P-VALUES ....................... 90
Bibliography ................................................................................................................................. 92
8
LIST OF FIGURES
Figure 1.1 (left): Commercial Building Floorspace, Energy Consumption and Energy Intensity
by Building Activity (Energy 2012) ...................................................................................... 13
Figure 1.2 (right): Commercial Buildings Site Energy Consumption by End Use [DOE 2012].. 13
Figure 1.3: Graphical illustration of energy use influencing factors ............................................ 14
Figure 1.4: U.S. Census Regions and Divisions (U.S. Energy Information Administration 2012)
............................................................................................................................................... 15
Figure 1.5: Consumption by end use for different building types (U.S. Energy Information
Administration 2012)............................................................................................................. 17
Figure 1.6: Acceptable operative temperature ranges (ASHRAE Standard 55 2013) .................. 18
Figure 1.7 : Assumed occupancy schedules for office spaces used in energy simulation programs
............................................................................................................................................... 19
Figure 1.8: Classification of energy estimation models in data-driven approach (Fumo, 2013).. 20
Figure 2.1: The occupancy patterns (a) single-square curve (b) one-valley curve (c) two-valley
curve (d) variable curve (e) flat curve ................................................................................... 23
Figure 2.2: Boxplot of annual EUI of various end-use loads (Yu, Benjamin, et al. 2011)........... 29
Figure 3.1: Flow diagram showing methodology ......................................................................... 31
Figure 3.2: List of measurable parameters for energy influencing factors for space in study. ..... 32
Figure 3.3: Picture of Open Office Layout of space in study ....................................................... 33
Figure 3.4: Location of sensors installed in study space .............................................................. 33
Figure 3.5: Integrated Sensor Suite used to collect weather data (Davis Instruments n.d.) ......... 34
Figure 3.6: Screenshot of Building Management System displaying zones and its temperature. 35
Figure 3.7: HOBO Temperature and Humidity Data Logger and Carbon Dioxide Sensor .......... 36
Figure 3.8: Wireless Circuit Level Clip on sensors for collecting sub-metered data ................... 36
Figure 3.9: USB Based Infra-red People Counter ........................................................................ 38
Figure 3.10: Entryway location of infrared bi-directional occupant sensor ................................. 38
Figure 3.11: Illustrated relationship between independent and dependent variables ................... 41
Figure 3.12: Illustrated difference between simple linear regression and multiple regression .... 42
Figure 3.13: Final structure for multiple regression analysis ....................................................... 43
Figure 3.14: Illustration of relationship of variables for Climate and Occupants with Energy. ... 43
9
Figure 3.15: Illustration of final structure of variables for Multiple Linear Regression Analysis.
............................................................................................................................................... 44
Figure 4.1: Collected data for Outdoor Temperature .................................................................... 46
Figure 4.2 (a) & (b): Collected data for Outdoor Solar Radiation ................................................ 47
Figure 4.3: Outdoor Condition for (a) Week 1 and 2, (b) Week 3 and 4, and (c) Week 5 and 6. 48
Figure 4.4: Comparison of recorded for hottest day and average TMY2 values for October. ..... 49
Figure 4.5: Comparison of recorded for coldest day and average TMY2 values for November . 49
Figure 4.6 (a) and below (b): (a) Snapshot of Wind Rose from Los Angeles TMY2 Weather file
and (b) Collected Data for Cumulative Wind Direction ....................................................... 50
Figure 4.7: Zonal Temperature from Building Management System ........................................... 51
Figure 4.8: Screenshot of Zones in Office Layout ........................................................................ 51
Figure 4.9: Indoor Spot Measurement of Indoor Temperature and Relative Humidity ............... 52
Figure 4.10: Indoor Spot Measurement – Lighting Intensity (lux) ............................................... 52
Figure 4.11: Average Lighting Levels for a typical workday, 6am to 6pm .................................. 53
Figure 4.12: Indoor Spot measurement – Indoor Carbon Dioxide Levels (ppm) ......................... 53
Figure 4.13: Pie Chart summarizing energy consumption by end uses ........................................ 54
Figure 4.14: Sub-metering data for AHU Fan Category ............................................................... 55
Figure 4.15: Day profile for AHU Fan Energy Use ...................................................................... 55
Figure 4.16: Sub-metering data for Water Heating Category ....................................................... 56
Figure 4.17: Day profiles for day of highest Water Heating Energy Use ..................................... 56
Figure 4.18: Collected sub-metering data for lighting category ................................................... 57
Figure 4.19: Day profile for day of highest Lighting Energy Use ................................................ 57
Figure 4.20: Sub-metering data for Kitchen Equipment Category ............................................... 58
Figure 4.21: Average day profile for Kitchen Equipment Energy Use ........................................ 58
Figure 4.22: Sub-metering data for Refrigerator Category ........................................................... 59
Figure 4.23: Average day profile for refrigerator energy use ....................................................... 59
Figure 4.24: Sub-metering data for Office Equipment Category ................................................. 60
Figure 4.25: Average daily profile of Energy Use Consumption for Office Equipment .............. 60
Figure 4.26: Sub-metering data for Computer Category .............................................................. 61
Figure 4.27: Average daily profile of Energy Use Consumption for Computer Category ........... 61
Figure 4.28: Sub-metering data for Total Office Energy Consumption ....................................... 62
10
Figure 4.29: Energy Consumption for 30 Working Days, 6am to 6pm ........................................ 62
Figure 4.30: Occupancy Schedule collected from Occupant Sensors .......................................... 63
Figure 4.31: Graphical summary for Occupants ........................................................................... 63
Figure 5.1: Box Plot of Energy Consumption of whole study period .......................................... 64
Figure 5.2: Box Plot of Energy Consumption of 30 working days (6am to 6pm) ........................ 65
Figure 5.3: Average Energy Use by End-Use Category ............................................................... 66
Figure 5.4: Interval Plot for Energy Use ....................................................................................... 67
Figure 5.5: Scatterplot of Total Energy Consumption (kWh) vs. Occupants (No.) ..................... 69
Figure 5.6: Simple Linear Regression between Independent Variables and Dependent Variables
............................................................................................................................................... 70
Figure 5.7: Fitted Line Plot for Total Energy Use against Occupants .......................................... 71
Figure 5.8 (a) & (b): Fitted line plot for (a) 0 to 22 and (b) 23 to 44 number of occupants ......... 72
Figure 5.9: Fitted Line Plot for Energy Consumption vs. Outdoor Temperature ........................ 73
Figure 5.10: Fitted Line Plot for Energy Consumption vs Solar Radiation.................................. 73
Figure 5.11: Fitted Line Plot for Energy Consumption vs. Outdoor Relative Humidity ............. 74
Figure 5.12: Illustration of final structure of variables for Multiple Linear Regression Analysis.
............................................................................................................................................... 75
Figure 5.13: Stepwise Regression Analysis of AHU End-Use Category ..................................... 77
11
LIST OF TABLES
Table 1.1: Influencing factors of building energy consumption (Yu, Benjamin, et al. 2011) ...... 14
Table 1.2: Energy Intensity by Census Region (U.S. Energy Information Administration 2006) 15
Table 2.1: Classification of building related data (Yu, Fung and Fariborz 2013) ........................ 28
Table 3.1: Description of parameters measured for climate conditions (Davis Instruments n.d.) 34
Table 3.2: Indoor Parameters and Tools Used .............................................................................. 35
Table 3.3: Location of Wireless Clip-Ons and what they represent ............................................. 37
Table 3.4: Time Resolution collected for each parameter ............................................................ 39
Table 3.5: Breakdown of study period in weeks........................................................................... 40
Table 3.6: Gantt chart ................................................................................................................... 45
Table 4.1: Peak temperature values for by week .......................................................................... 46
Table 4.2: Cumulative Energy Consumption and Occupants for Random Weeks ....................... 54
Table 4.3: Cumulative Occupant Count for Random weeks ........................................................ 63
Table 5.1: Tabulated Occupancy Usage for Weeks A, B, and C .................................................. 67
Table 5.2: Tabulated Correlation between Total Energy Consumption and Predictors ............... 70
Table 5.3: Summary of Results from Stepwise Regression Analysis ........................................... 80
Table 5.4: Summary of Results from Stepwise Regression Analysis with Time Lag .................. 84
12
ABSTRACT
Many approaches have been used in establishing a relationship of energy consumption. Two main
types include engineering-based and statistics. Statistical approaches offer a range of methods to
draw links between energy consumption and a range of variables related to building characteristics,
its operations, the outdoor environment, and occupant presence. An advantage of statistical over
engineering-based is the capability of taking into account the behavior of occupants. Many past
studies use information from power suppliers but mostly referred to the residential sector or at a
whole building context. The big question left unanswered is, by how much sensitively does
occupants influence energy use at end-use consumption? One approach is to extract building-
related data as it already contains the full effects. With the advancement of sub-metering and cloud
storage technology, energy consumption data can now be broken down into end-use categories in
a commercial setting and occupant’s impact can be more refined. A study was undertaken to gain
insights into electrical energy profiles, climate parameters, indoor conditions and occupancy in an
office setting. Sensors are placed for sub-metering energy use at the circuit level, weather data is
collected from a weather station located on the site, and indoor conditions is gathered from data
loggers, occupants are counted using a bi-directional occupancy counter. Data collected at the
minute usage, to take full advantage of building operational data. This study aims at using statistics,
regression analysis in particular to understand the impact of occupants on end-use categories and
also the order of influencing factors on energy consumption. Upon analyzing, the relationship
between occupancy and energy use is found as a significant component in Lighting, Office
Equipment and Computers category. Results from the correlation between end-use category and
occupants show occupants effects in descending order first Computer, followed by Office
Equipment, Lighting, AHU, Kitchen Equipment and Refrigeration. It was found that climate
precedes occupancy in affecting total energy use. However, within the studied climate predictors
and study period, Outdoor Temperature affects the Air Handling Unit (AHU) energy consumption
during the warmer days of the study period, and when the temperature lowers, solar radiation
precedes outdoor temperature in affecting the energy consumption of AHU. This data is collected
between October and November within the Los Angeles climate. As office spaces take up the lot
of commercial spaces and energy consumption, the findings from this study hopes to give a deeper
insight into how energy is used and the energy savings potential.
13
CHAPTER 1 : INTRODUCTION
1.1 An Introduction to Building Energy Consumption in the United States
In 2003, the most recent year for which such data are available, office and retail buildings
represented– 17% and 16% respectively of commercial space – and 19% and 18% respectively, of
commercial sector energy consumption. (Figure 1.1) (Energy 2012). Buildings in the office
category represent the highest both in total floor space use and primary energy consumption.
Figure 1.1 (left): Commercial Building Floorspace, Energy Consumption and Energy Intensity
by Building Activity (Energy 2012)
Figure 1.2 (right): Commercial Buildings Site Energy Consumption by End Use [DOE 2012]
The breakdown of energy consumption in a commercial setting, in 2010: space heating takes up
27% of site energy in the commercial sector followed by 14% from lighting and 10% in space
cooling (
Figure 1.2 1.2). These three end uses top the consumption whether it is categorized by all buildings,
commercial buildings or office buildings and is close to half of commercial site energy
consumption. This work’s focus is in office buildings, since they comprise the largest category in
the commercial sector and are usually equipped with more actuation possibilities, which raise the
expectation of a greater energy savings potential.
14
1.2 Factors affecting Building Energy Consumption
Several factors influencing building energy consumption are shown in Table 1.1.
Table 1.1: Influencing factors of building energy consumption (Yu, Benjamin, et al. 2011)
No Influencing factor Example
1 Climate Temperature, Humidity, Solar Radiation..
2 Building-related characteristics Type, area, orientation, materials
3 Building service systems and operation Space cooling/heating, hot water supplying
4 User-related characteristics User presence
5 Building occupants’ behavior and activities Turn on/off lights, TVs
6 Social and economic factors Degree of education, energy cost
7 Indoor environmental quality required Preferred indoor air quality and comfort
Although broken down to seven factors, the last three are occupant related. Social and economic
factors will partly determine the occupants’ attitude toward energy consumption in their daily
activities, thereby influencing building energy consumption. Indoor environmental quality is
decided by building occupants, thereby also influencing building energy consumption. Both these
factors directly influence occupants’ behavior and indirectly the building energy consumption.
(Yu, Benjamin, et al. 2011). Their influences on building energy consumption are already
contained within the factor of user presence. The main influences are therefore a function of- a)
weather parameters, b) building characteristics, c) energy systems and its operations, and d)
occupants’ behavior. (Fumo, 2013).
Figure 1.3: Graphical illustration of energy use influencing factors
15
1.2.1 Climate
Climate affects building energy use as the indoor condition is related to the outdoor conditions.
The second law of thermodynamics states that heat flows naturally from a higher temperature to a
lower temperature. In order to add / remove the additional heat, building systems will have to work
harder to maintain the space condition. Energy use intensity is different by regions in the United
States, and can be explained by climatic conditions. Offices in the South region have a 5% higher
energy use intensity compared to offices in the Mid-West region (Table 1.2).
Figure 1.4: U.S. Census Regions and Divisions (U.S. Energy Information Administration 2012)
Table 1.2: Energy Intensity by Census Region (U.S. Energy Information Administration 2006)
Electricity Energy Intensity (kWh/square foot)
By Building Activity North-east Mid-West South West
Health Care 19.4 23.2 24.5 22.5
Retail 8.8 14.2 16.0 14.8
Office 16.5 17.9 18.8 15.0
1.2.2 Building-related characteristics
a) Building Type
Different types of commercial buildings have different energy consumption rates because of
various internal loads due to equipment and user presence. Food sales, food service, and healthcare
have high-energy use intensities; (almost twice or more than office space) they tend to occupy
16
more hours than a commercial office space and require higher energy intensive equipment (Energy
2012).
b) Building Area
In general, the larger the building area, the more energy is needed to condition the space. “Less
compact building forms increase a building's daylighting potential, but they also may magnify the
influence of outdoor climate fluctuations. Greater surface-to-volume ratios increase conductive
and convective heat transfer through the building envelope. It is critical to assess the daylighting
characteristics of the building form in combination with the heat transfer characteristics of the
building envelope in order to optimize building energy performance.” (ASHRAE Standard 90.1)
c) Orientation
In the northern hemisphere, buildings that have larger areas exposed to the south have more solar
gain throughout the year (Gronadzik, et al. 2010). Depending on how solar gain affects internal
condirions, building systems are used to add or remove heat in a space to achieve comfort. Allen’s
book “Architect’s Studio Companion” recommends orientation of the building and glazing along
the east-west axis to maximize natural lighting and design with daylight. It also optimizes Winter
solar gain and reduces Summer gains.
d) Façade / Materials
The façade, or the exterior walls of a building, is the physical separation between outside
environment and inside conditioned space, which could resist air, heat, water, light, noise transfer.
(Cleveland 2006) Façade design strategies are dependent on the location and climate of a building.
(Aksamija 2013). In a study done by Synnefa et al., the effect of roof properties in cooling loads
showed that increasing the solar reflectance by 0.65 from a base case of 0.2 could achieve savings
that vary between 10.7% and 27% according to the specific climatic conditions. Use of cool
coatings can contribute to the reductions of cooling loads and improvement of indoor thermal
comfort. (A, M and Akbari 2007).
There seem to be no “best” building parameters as it is a complex problem balancing multiple
variables of form, shape, orientation, envelope materials for the design of high-performance
buildings.
17
1.2.3 Building Systems Service and Operations
Building mechanical or control systems are designed so that they are able to maintain the space at
set conditions for occupant thermal comfort.
Figure 1.5: Consumption by end use for different building types (U.S. Energy Information
Administration 2012)
Heating, Ventilation and Air-Conditioning (HVAC) systems and their associated energy
consumption have become an unavoidable component accounting for almost 50% energy
consumed in buildings (Figure 1.5) (Perez-Lombard, Ortiz, and Pout, 2007). However, the
majority of building HVAC systems operate inefficiently, conditioning for space instead of
occupants.
Apart from building systems, the dependency on plug loads in offices has increased as the use of
computers, is crucial in all workspaces as compared to paper loads in the past. This contributes to
internal heat gains, which indirectly contributes to the electricity demand for conditioning the
spaces, Yagi in her thesis found a reduction of up to 150Wh/person if a desktop computer is
replaced with a laptop computer. Her study also found a discrepancy between measured and
modeled results using the information found in plug loads (Yagi-Kim 2013).
1.2.4 User Presence (Occupants)
Logically, the higher the number of occupants, the higher the energy use will be. According to
ASHRAE, six primary factors must be addressed for thermal comfort: metabolic rate, clothing
insulation, air temperature, radiant temperature, air speed and humidity. ASHRAE Standard 55
defines the comfort zone in terms of the range of operative temperatures that provide acceptable
thermal environmental conditions (Figure 1.6). The first two factors (metabolic rate and clothing
18
insulation) will vary according to occupant. The other four factors can be maintained by a building
system to provide thermal comfort. (ASHRAE Standard 55 2013).
Figure 1.6: Acceptable operative temperature ranges (ASHRAE Standard 55 2013)
As occupants produce internal gains, more ventilation will be required as the number increase.
ASHRAE Standard 62.1 standard requires buildings to provide ventilation rate using the equation:
Equation 1.1
𝑉 = 𝑅 𝑝 𝑃 𝑧 + 𝑅 𝑎 𝐴 𝑧
Where V is the ventilation rate, Rp is the minimum L/s.person, Pz is the number of people, Ra is
the minimum L/s.m
2
and Az is the floor area. Values change depending on the building type as
specified in the standard. For example for an office space, the minimum ventilation rate for Rp is
2.5 L/s.person. Therefore, as the number of occupants increase, more outdoor air is required to
maintain good indoor air quality.
1.3 Building Energy Performance vs. Occupancy (human factors)
Occupancy has always been investigated since the occupant’s presence is what is really necessary
for an action to occur. Unless in extreme climates, where systems are needed to operate to prevent
equipment malfunction, for example, pipes cannot freeze even when the building is unoccupied.
Human factors or occupant behavior affects the building energy use directly and indirectly by
opening/closing windows, turning on/off lights, office equipment, HVAC systems, and setting of
indoor thermal, acoustic and visual comfort criteria (Hong 2014). In reality, a perfect prediction
of occupancy will never be achieved, as occupant behavior is stochastic in nature (Yu, Benjamin,
19
et al. 2011). Human behavior consists of universal reactions in human nature and actions done
conditioned by personal background and experiences (Bonte, Francoise and Berangere 2014).
Thus, actions could be very different depending on the individual.
Maier in his study investigating 22 identical houses over 2 years showed differences between
houses equipped with the same ventilation systems, concluding discrepancy was due to occupant’s
behavior (Meir, et al. 2009). If subjected to the same environmental conditions in buildings,
occupants reacting differently could lead to large discrepancies between energy performances of
the space.
1.4 Why occupancy should be identified in building energy simulation
Figure 1.7 : Assumed occupancy schedules for office spaces used in energy simulation programs
Building energy simulation provide an efficient, simple method for predicting the energy use of
new and existing buildings. Occupancy schedule assumptions are usually based on standards or
local code specified schedules for code compliance. Lighting, building systems and equipment
follow these schedules for operation. These schedules are static and usually represent what is
“thought’ of how a space is used by percentage. In Figure 1.7, during weekdays occupants are
assumed to trail in from 7 am and reach 100% occupancy by 10 am, and then 50% of the occupants
would leave for lunch and occupancy resumes to 100% again until 7pm. This rigid schedule is
repeated weekly but does not represent how offices are used today especially with the increase
20
occurrence of working “away” from the office. Results do not accurately represent “typical”
occupant’s nature, calling for new models for the building occupant. (Newsham, Mancini and Birt
2009) (Clevenger and Haymaker 2006) (Ryan and Sanquist 2012). More detailed occupancy
schedules can replace the current ones to generate results that are more realistic.
1.5 Data-driven Approach to Energy Use Estimation
For new buildings, computer simulations can provide alternative design solutions to compare
energy efficient and cost-effective options in a short time. For existing buildings, energy
consumption modeling approaches can be classified as statistical, hybrid or engineering (Fumo
and Biswas, 2015).
Figure 1.8: Classification of energy estimation models in data-driven approach (Fumo, 2013)
Fumo in (Fumo, 2013) compiled different approaches presented by various researchers (Swan and
Ugursal 2008) (Zhao and Magoules) (Pedersen) on this topic (Figure 1.8) and elaborated below:
1. Statistical methods:
i) Regression approach: Purely statistical technique that needs a large amount of historical
/ detailed data on energy consumption to identify the source of the energy consumption
base on statistical significance. An example of methods include correlation, simple linear,
multiple linear, stepwise regression.
ii) Intelligent approach: Machine learning methods or intelligent computer systems that
are developed based on machine learning algorithms that are capable of “making
Data-driven
Approach
Statistical
Regression
Intelligent
Hybrid Engineering
Forward
Calibrated
21
decisions” as an effect of interpretation of solving even small quantities of data. An
example of methods includes neural networks, support vector machine.
2. Engineering methods:
i) Forward approach: Uses physical specifications usually from equipment and systems
level to account for end-use energy consumption.
ii) Calibration approach: Uses energy simulation program that has been calibrated with
actual measured data, and upon changing parameters provide change or prediction in
energy consumption.
3. Hybrid models or “gray models”: Used when the information is partially known or due to data
uncertainty. This model combines elements from engineering and statistical.
Approaches for building energy studies depend on the information that is available for the purpose.
Statistical approaches use available measured data, while engineering approaches use building or
systems characteristics (Fumo and Biswas, 2015). For an existing building, measured data that
represents how the building operates can be used. Therefore for the type of information available
and in measurable amounts, the statistical approach is better suited to complete this study. As for
which type of statistical method, a comparison of statistical approaches between regression
analysis, decision tree and neural networks showed that small differences in terms of errors
indicating the three techniques are generally comparable in predicting energy consumption
(Mastrucci, et al. 2014)
1.6 Objectives of this research
The main goal of this research is to find the relationship between occupancy and energy use profile;
this function can be used to understand further how energy consumption in an office by end-use
categories are currently affected by occupants. This applied to fundamentals of energy simulation
can hopefully reduce energy gaps between simulated and real building performance.
The first objective is to explore the impact of energy influencing factors in an office space using
statistical methods, in particular using linear and multiple regression analysis. All parameters that
22
are thought to influence building energy consumption are measured in an office setting. Regression
coefficients can provide the percentage of accountability of each influencing factor.
The second objective is to investigate how occupancy patterns affect energy consumption in an
office setting by end-use categories. Many of recent studies have been done to identify the impacts
of occupant behavior on building energy consumption, where findings show occupants do affect
energy consumption. However, by how much and where does it affect?
The third objective is to estimate variations in energy use based on the change of outdoor
environment. Data is analyzed to obtain the sensitivity factor of building performance with
different outdoor parameters. The outcome function makes it beneficial for energy modelers or
energy companies to understand or predict energy usage throughout a year.
1.7 Hypothesis Statement
Occupant behavior can explain variation in office energy consumption. If so, how much could they
be responsible for?
To accomplish the objectives discussed above, the following research hypotheses are established:
1. The relationship between occupancy and energy use profile is significant,
2. Occupancy data significantly affect some of the end-use energy records,
3. The impact of occupancy on energy performance varies depending on climatic and end-use.
To gain insights into electrical energy profiles, sensors are placed for sub-metering at the circuit
level, weather data is collected from a weather station located on the site, and indoor conditions
are gathered from the building BMS system and data loggers. Occupants are accounted for using
a bi-directional occupancy counter. The effect of occupancy, climate, or indoor conditions on
energy consumption are analyzed; key trends and data patterns in energy use is identified.
Chapter Summary
In this chapter, we discussed why finding the impact of occupants in energy use of a building is
important and how it could contribute to high-performance buildings. In the next chapter, a
literature review in this area will be presented. Chapter Three shows the methodology and tools
used to collect data. Chapter Four shows the data collected. Results are analyzed and presented in
Chapter Five. Chapter Six include a conclusion, study limitations, and future work.
23
CHAPTER 2 : BACKGROUND AND LITERATURE REVIEW
2.1 Background on Occupancy affecting Energy Performance
Occupant behavior is one of the sources of uncertainty in the predictions of building energy use.
The relative impact of occupant behavior seems to differ in various studies; however researchers
agree that as thermal properties of buildings improve, the role of building characteristics will
decrease, making occupant behavior more important (Chen, Wang and Steemers 2013). Building
occupants affect building energy use through the temperature set points, fixed operation schedules,
manual operation actions indirectly affecting conditioning systems load on a building (Chang and
Tianzhen 2013).
A study was done by (Chang and Tianzhen 2013) where an occupancy schedule is created based
on lighting switches from 200 cubicle offices of a commercial office building. Occupancy levels
were identified by measuring cubilcle lighting-switch data. Profiles were classified into five
patterns showing variation of occupancy schedules in the same place (Figure 2.1). It is unclear
how much building energy use is affected by this variation of occupancy.
Figure 2.1: The occupancy patterns (a) single-square curve (b) one-valley curve (c) two-valley
curve (d) variable curve (e) flat curve
Much literature agrees on the reduction potential of HVAC systems that take account of occupancy
information. Erickson’s study uses SCOPES, a wireless camera sensor network to gather traces of
24
human mobility patterns in buildings (Erickson, et al. 2009). The data was used to create prediction
models for describing occupancy and movement behavior and used to compute energy savings
where controls were adjusted to the collected occupancy patterns. Results show HVAC energy
savings estimation of 14%, suggesting that knowing the occupancy and usage patterns will result
in significantly higher energy savings. However, the occupancy schedules were first modeled and
placed into an energy simulation and did not use a real HVAC system. There is a 20% occupancy
estimation error resulting in inaccurate savings estimation, but this is still better than using a static
schedule. (Erickson, et al. 2009) Other studies maintained that large savings in HVAC related
energy if controls are adopted or adjusted for occupancy (Yang and Becerik-Gerber 2014)
(Brandemuehl and Braun 1999).
Oldewurtel et. al maintains that taking occupancy information into building control has significant
energy savings potential. The study shows total energy savings potential of up to 34%. The study
was done for Switzerland offices where the climate could have also played a role. (Oldewurtel,
Sturzenegger and Morari 2013)
In reality, a perfect prediction of occupancy will never be achieved as occupant behavior is
stochastic in nature (Yu, Benjamin, et al. 2011). As building simulation tools use occupancy
patterns in a static way; which does not include occupants’ complexities; the reasons for
discrepancy often relates to occupant behavior. (Yu, Benjamin, et al. 2011). Results could be more
realistic if based on the better-assessed impact of occupancy patterns on building energy
performances.
2.2 Background on Energy Simulation
Building energy modeling has gained importance and is required in some locations in the United
States. During a building design phase, building systems are designed based on maximum input
parameters based of handbooks or codes. This is unavoidable as occupancy may vary widely from
one company to another upon occupancy, a general schedule is as best as it gets. While loads vary
throughout the year based on weather files applied, building systems will be selected based on
peak days as the building is required to perform in the worst-case scenario through building energy
modeling. This is also unavoidable.
Current simulation software is capable of simulating different situations based on four factors;
which are climate, construction characteristics, occupancy schedules and building systems (Chang
25
and Tianzhen 2013). Designers often observe considerable energy consumption differences
between the designed/simulated and the actual building. It's hard to identify the influences of
occupant behavior and activities completely through simulation due to users’ behavior diversity
and complexity. (Yu, Benjamin, et al. 2011) While building systems are designed to condition for
occupants, however, often building systems operate when occupants are away, resulting in high-
energy waste. In a typical office, space is usually divided into workstations, meeting rooms, pantry.
Although the spaces are never used 100% of the time by 100% of the occupants, because the people
gathering in the meeting rooms are usually the same individuals from the workstations, each person
can only occupy a space at a time. The same goes for plug load devices. (Sanchez, et al. 2007)
(Yagi-Kim 2013). Many loads are left powered regardless of actual occupancy or needs. Monitors
and display screens are in standby mode 24/7 even when they are not actively used.
This study can provide a more precise understanding of energy use in an office setting based on
real data and maybe help to fine tune energy models in the future. Sub-metering data contains
useful information about the interactions between energy use by end-use category and influencing
factors.
2.3 Background on Mining Building-related Data
Overview
While the above studies have been done to find a reliable occupancy model, other researchers
focus on mining available building–related data. Data mining techniques can be used to analyze
huge amounts of data, extracting useful knowledge from data. Today, many industries collect data
to either understand or predict patterns of future trends. The same data mining techniques can be
applied to data collected from building related parameters.
Building related data can include quantitative and qualitative data such as measurements and
monitoring, and methods such as walk-throughs, observations, and user satisfaction questionnaires
(Post Occupancy Evaluations) (Meir, et al. 2009). Building automation systems (BAS) in buildings
now can produce vast amounts of data; providing a more robust understanding of energy use. They
contain rich information but rarely translated into useful knowledge due to the lack of public
sharing, complexity; and also lack effective data analysis personnel. For the residential sector, a
vast amount of billing information stored by energy suppliers worldwide provides a substantial
data source for energy modeling. Researchers have applied a variety of statistical methods to draw
26
conclusions of energy consumption as a function of residential characteristics (Swan and Ugursal
2008). Other ways to collect data are on-site spot measurements. The on-site spot measurements
include indoor thermal, relative humidity, lighting intensity, and carbon dioxide levels.
The three well-documented techniques in the residential sectors are Regression, Conditional
Demand Analysis, and Neural Network. Fumo explained of the three; regression technique is
popular because:
- It is relatively easy to apply,
- Requires less computational power than other statistical approaches such as Neural Networks,
- It provides satisfactory prediction results,
- Availability of more measured data through smart metering and cloud storage. Patterns can be
drawn regardless the size of data unlike Conditional Demand Analysis where large amount of data
is required for reliable results (Fumo and Biswas, 2015).
2.3.1 Regression Model for Energy Performance Calculation
In general, the Regression technique is used to evaluate data base on a “line-of-best-fit”, providing
coefficients corresponding to the input parameters. These coefficients determine the correlation
between multiple input parameters. The response, or dependent variable, is usually presented by
the “y”-axis, and other variables are called the predictor or independent variables and usually
presented byx
1
, x
2
, etc. (Samprit and Jeffrey 2013). When dealing with one response, the
regression is known as simple linear regression, while two or more responses is called multiple
linear regression. Linear regression analysis attempts to model the relationship by fitting a linear
equation to the data. Although this is useful when two variables are of a different metric and scale
are compared, it should be noted that the relationship does not necessarily imply that one is the
causation of the other variable. However, there could be a significant association based on the
value of the coefficient.
Simple linear regression has an equation form:
Equation 2.1
Y = β
0
+ β
1
X + ε
Where Y here is the response variable and X the predictor variable, β0 and β1 are regression
coefficients and ε is the error accounted for.
27
While multiple linear regression (MLR) is an extension of the simple linear regression model with
an equation form of:
Equation 2.2
Y = β
0
+ β
1
X
1
+ β
2
X
2
+ ⋯ + β
n
X
n
+ ε
MLR allows more than one predictor variable. Here, Y here is the response variable and X0,
X1, ...Xn are the predictor variables, β0, β1 … βn are regression coefficients, and ε is the error to
account for the discrepancy between predicted data and the observed data. The difference between
MLR and several simple linear regressions (Rudolf Jakob and William J 2006) is that in MLR each
coefficient indicates the change in the response associated with changes in that independent
variable while others remain fixed. This advantage could help to explain the accountability of a
single variable without change in other predictor variables. In this research, multiple linear
regression models used to predict the response can be developed using multiple indicators of the
energy influencing factors analyzed. Another reason for using multiple regression analysis is to
study the impact of various predictors on building energy performance.
Stepwise regression refers to the process of developing a regression model by adding or removing
variables and re-computing the coefficient (Castree, Kitchin, and Rogers 2013). It could be used
to deal with a lot of datasets and determine the most significant predictors after dropping out less
important variables, a step at a time based on the statistical significance. Some examples of using
the approach of regression analysis to predicting estimation on whole building energy consumption
are below:
Yang Chao in his thesis applied regression models using building façade features as predictors to
estimate energy use intensity as a response. The study aimed at setting a more accurate energy use
benchmark. He found that façade features that are important in Los Angeles were the built year,
volume, volume-to-façade area and cooling degree-days. (Chao 2015)
(Chen, Wang and Steemers 2013) found from studying residences in Hangzhou through regression
from survey data, that socio-economic characteristics and behavior could explain 28.8% variation
in heating / cooling energy consumption. It was not certain what extent of building characteristics
could account for the variation. However, it seemed to play a smaller role than occupancy behavior
based on socio-economic standards. The study concluded that elderly residents exhibited a more
frugal behavior pattern than the younger ones based on a negative correlation between occupants
age and energy consumption.
28
Multiple linear regression was used by to predict annual energy consumption of a bank as a
function of its construction characteristics, climate, and energy performance of 55 banks. The
resulting model allowed for detection of inefficient bank branches in terms of energy use.
(Korolija, et al. 2012)
Catalina used regression techniques on a simulation database and proposed a model based on the
main factors that influence a building’s heat consumption to predict heating energy demand.
Factors were heat loss coefficient, south equivalent surface and the difference between the indoor
set-point temperature and the sol-air temperature. (Catalina, Iordache and Caracaleanu 2013).
However, there was an average error of less than 20% when comparing with on-site data from 17
existing buildings, and it did not consider occupancy variation.
Elsawaf et. al. used regression analysis to evaluate the effectiveness of using heat pumps for space
heating. The evaluation was based on comparisons of different methods of heating: heat pump, gas
heating, electrical heater and some combination. Data was collected from surveys and actual
energy consumption in terms of both gas and electrical. The predictor variables, in this case, was
house size, the number of occupants, the number of stories, the age of construction, orientation,
heating set points. Results supported the hypothesis that residential homes using heat pumps are
more energy efficient compared to the other heating methods. This may well be the byproduct of
the climate, which was not considered. Heat pumps typically used in milder climates than gas
furnaces because heat pumps cease to function outside of a moderate temperature range.
For this study, the response would be energy use and the predictors are climate and occupancy
(Table 2.1). The rank of importance for factors influencing energy use can be determined. Linear
regression methods that will be included in this study are Correlation, Linear Regression, and
Stepwise Regression Models.
Table 2.1: Classification of building related data (Yu, Fung and Fariborz 2013)
No Building-related data Example Variable type
1 Climatic data Air temperature, relative humidity, etc. Predictor
2 System
operation
data
HVAC systems Air temperature, Airflow rates, etc. Predictor
Energy Data Electricity consumption, end-use
loads, etc.
Response
3 Building physical parameters Areas, Orientation, Window wall ratio Predictor
4 Occupancy Schedules Predictor
29
2.3.2 Other Statistical Methods for analyzing data for energy consumption
Cluster analysis (K-means) is another approach that was used to develop a methodology for
examining the effects of occupant behavior on building energy consumption. Yu recommended
cluster analysis, a basic data mining technique and collected information on many residential
buildings, using building characteristic as a variate and energy consumption on the macro level
(Yu, 2013). The process here merges data into different clusters so that instances in the same
cluster have a high similarity.
Figure 2.2: Boxplot of annual EUI of various end-use loads (Yu, Benjamin, et al. 2011)
(Yu, Benjamin, et al. 2011) Organized similar residential buildings in six different districts of
Japan and using cluster analysis technique studied the effects of occupant behavior on building
energy consumption. The study shows variability in annual Energy Use Intensity (EUI) of different
end-uses induced by occupant behavior. The mean value for the EUI is two kBtu/ft
2
/year, however
as in Figure 2.2; we can see a significant variability that ranges from close to zero and five. This
implies the high potential for energy savings by improving behavior based on the end uses.
However, this study was done using residential buildings and through various climate zones where
residents will sure to have different needs of conditioning interior conditions.
Decision tree method is another approach to predict building energy use in practice. This method
uses a flowchart-like tree structure to obtain useful predictors. Zhun et al. (Yu et al. 2010)
demonstrated that a decision tree method can predict building energy demand by 93% accuracy
for training data and 92% accuracy for test data. Decision tree can be utilized to predict energy,
provide significant predictors and their values.
30
Chapter Summary
There is a possibility that, in order to understand building energy performance, data mining can be
employed to extract hidden useful knowledge from huge amounts of building-related data. This
chapter discussed research that was done and their limitations. Some research focuses on modeling
occupancy while some focus on mining building-related data. This study will integrate real-time
occupant data to find the correlation with energy use using some proposed data mining methods.
This study of buildings after being occupied, can improve current conditions and guide the design
of future of buildings. In Chapter 3, the methodology to achieve the goals of this research will be
elaborated.
31
CHAPTER 3 : METHODOLOGY
Overview
The purpose of this research is to identify the impact of influencing factors of energy consumption
and investigate the relationship between occupancy and building end-use consumption by using
regression techniques. In this chapter, the methodology to achieve the purpose will be discussed.
In general, real-time data of all parameters that affect energy use which are climate, indoor
conditions, occupancy schedules, and building envelope; are collected. Each step is further
elaborated below with a Gantt chart at the end of the chapter summarizing steps done.
3.1 Methodology
Figure 3.1: Flow diagram showing methodology
A systematic process of using building-related data to find the relationship of occupants to building
energy consumption is proposed and illustrated in Figure 3.1.
a) Problem definition and objective setting. Understand and list influencing factors of energy.
b) Data source selection : Building selected to collect building-related data
c) Install sensors to collect building-related data
d) Data collection: Extracting and gathering all building-related data through weather station,
building automation systems, data loggers, occupancy sensors, and energy use sub-
metering, and then construct a database.
e) Data pre-processing / preparation: detect and remove outliers and noise, handle missing
values, deal with inconsistencies. Create a database where all collected data is synced.
f) Data mining using Clustering, Correlation, Simple Linear Regression and Stepwise
Regression Analysis.
g) Results analysis and evaluation: Identify the significant trends and patterns.
h) Knowledge discovery and document presentation.
32
3.1.1 Problem Definition and Objective Setting
Understand and list influencing factors of energy as described in Chapter 1.2 Factors affecting . The
main influences are a function of- a) weather parameters, b) building characteristics, c) energy
systems features and d) occupants’ behavior. (Fumo, 2013).
Figure 3.2: List of measurable parameters for energy influencing factors for space in study.
3.1.2 Data Source Selection
The space chosen for this study will be in downtown Los Angeles, California because of
permission to access was granted, and the building has a typical office setting in a city. The said
office is 12,336 square-footage in size, and is situated on the 16
th
floor at the corner of Wilshire
Boulevard and Flower Street in Los Angeles. The commercial building was built in 1971, last
major renovation in 2007 where a new chiller was installed. All floors are occupied as office space
with normal working hours. On each floor, a central Air Handling Unit distributes chilled air,
which is circulated through ducts using under floor air distributed system that spans the entire open
workspace, and variable-air-volume system for the gathering areas. Other building characteristics
are its Window Wall Ratio of 65%. There is single pane clear glazing, and roll-down opaque
shading on the east and south sides. Since the subject space used here has an open office layout,
single occupied rooms are not covered in the study.
33
Figure 3.3: Picture of Open Office Layout of space in study
3.1.3 Installing Data Collection Tools
Factors that affect energy consumption can be measured across in a time series format. Sensors
were placed for collecting outdoor weather parameters using a weather station; indoor parameters
are collected from the building management system, and data loggers, sub-metering measuring
energy consumption at the circuit level, and occupants are counted using a bi-directional
occupancy counter location illustrated in Figure 3.4. Each sensors’ function and capabilities are
explained below.
Figure 3.4: Location of sensors installed in study space
a) Weather Station
Tool: Wireless Vantage Pro2 Plus with UV& Solar Radiation Sensors (Davis Instruments n.d.)
34
Figure 3.5: Integrated Sensor Suite used to collect weather data (Davis Instruments n.d.)
This tool consists of six sensors which are rain collector, temperature and humidity sensors,
anemometer, solar radiation, and UV sensor. The parameters collected are explained in Table 3.1:
Table 3.1: Description of parameters measured for climate conditions (Davis Instruments n.d.)
No. Parameter Description
1 Wind
The anemometer measures wind speed / run (nautical miles) and
direction. The console calculates a 10-minute average wind speed and
10-minute dominant wind direction.
2 Temperature
The Integrated Sensor Suite houses the outside temperature sensor.
Vented and shielded to minimize the solar radiation induced
temperature error.
3 Humidity
Humidity refers to the amount of water vapor in the air. Relative
humidity reflects the amount of water vapor in the air as a percentage
of the amount the air is capable of holding
4 Dew Point
Dew Point is the temperature to which air must be cooled for saturation
(100% relative humidity) to occur, providing there is no change in
water vapor content.
5 Rain
A tipping-bucket rain collector in the Sensor Suite measures 0.01” for
each tip of the bucket.
6
Solar
Radiation
A measure of the intensity of the sun’s radiation reaching the
horizontal surface. This includes both the direct component from the
sun and the reflected component from the rest of the sky.
35
b) Indoor conditions logger
In commercial buildings, the common system used is a central conditioning system. As the name
suggests, a centralized cooling/ heating system has all the plant located in a single area, in this
case, the basement plant room. This plant supplies chilled water to an air-handling unit located on
each floor, which is then supplied by ductwork to the floors/spaces within the floor. The chiller
energy is not considered in this study; only the Air-Handling Unit fan power is taken into
consideration as it 's hard to extrapolate the floor’s chiller energy use from the central system.
Indoor conditions are measured using two different tools. The first is the buildings inbuilt Building
Management System that provides zonal information and control system through a software
“Andover Continuum.” (Schneider Electric n.d.) For this research purpose, the data is collected,
but not controlled. In addition, HOBO Data Loggers were placed for spot measurements. (Onset
Computer Corporation n.d.) (Figure 3.4).
Table 3.2: Indoor Parameters and Tools Used
No Parameter Tool Description
1 Zonal Temperature Building Management System, Andover Continuum (Schneider
Electric n.d.)(Figure 3.6)
2 Spot Temperature HOBO Data Logger (Figure 3.7)
3 Relative Humidity HOBO Data Logger (Figure 3.7)
4 Lighting Intensity HOBO Data Logger (Figure 3.7)
5 Carbon Dioxide “Telaire” Carbon Dioxide Sensor (Figure 3.7)
Figure 3.6: Screenshot of Building Management System displaying zones and its temperature.
36
Figure 3.7: HOBO Temperature and Humidity Data Logger and Carbon Dioxide Sensor
c) Sub-metering for energy consumption
Sub-metering is tracking electricity usage on a micro level. Sub-metering in itself does not save
energy; it provides information about how much energy is being used, and where it is being used.
With effective monitoring, sub-metering has the potential to optimize energy use for a better
performance space. Wireless sensors developed by Panoramic Power are used on a circuit level
deployed across two main power boards. Data is collected by transmitting information every 10
seconds to be stored in the cloud and monitored online by real-time dashboards or generate reports.
(Panoramic Power n.d.)
Figure 3.8: Wireless Circuit Level Clip on sensors for collecting sub-metered data
Hardware Used for Sub-metering
PAN-14 high-current wireless sensor; powered by magnetic fields (does not require
maintenance, service or battery replacement)
37
Size: 1.33 x 1.14 x 1.67 inch Manufacturer: Panoramic Power
A total of 73 clip-on were used to measure energy use at the circuit level.
List of energy-using items in study space and categorized by type:
Table 3.3: Location of Wireless Clip-Ons and what they represent
No. Parameter Category Description
1 Boiler (4)
Hot Water Heater 7, Hot Water Heater 9, Hot Water Heater 11,
Instahot at men’s shower 10
2 Servers (4) Server Rcpt 10, Server Rcpt 12, Server Rcpt 14, Server Rcpt 16.
3 Computer (8) Furniture Partition 1,2,3,4,5,6,7,8
4 Display Screen (2) TV Rcpt 18, TV Receptacle 33.
5 Fan (2) EF1 & EF2 6, Ceiling Fan 13
6 HVAC (6) AHU 19, 21, 23, RA-1 HVAC, VAV 4, HVAC Controls 25
7
Kitchen Equipment
(7)
Ice Machine, Dish Washer #1 38, Dish Washer #2 40, Microwave
23, Kitchen Rcpt 20, Kitchen Rcpt 22, Garbage Disposal 25
8 Lighting (10)
Electrical and machine lights 42, PE1, PE7, PE10, Exit signs 21,
Track Lighting 27, EM Egress Lights 15, EM Egress Lights 18,
EM Egress Lights 30, Shower Lights 14
9 Mains (12)
Work Room Rcpt 11, Work Room Receptacles 31, Conv. Outlets
27, Work Room Fixtures 19, Phone Boot Hallway 29, Conv and
phone room recept 30, Keyless Fixtures 15, Keyless Fixtures 17,
Conv. Outlet 29, Floor Receptacles Reception 10, Work Room
receptacles 9, Phone room receptacles 34.
10
Office Appliances
(7)
Copier Outlet 16, Copier Outlet 18, Copier Rcpt 21, Copier 32,
Plotter and IT Storage 15, Plotter and storage rcpt 13, Existing
Wi-Fi 29
11 Misc. (7)
CUI 35, 16L Panel 12, 16L Panel 17, CUI 37, GFI recept in
bathrooms 17, CSFD 23, Spare 13
12 Refrigeration (2) Refrigeration #1 24, Refrigeration #2 26
13
Sub Panel Mains
(2)
16A Panel Sub Main (W), 16B Sub-main (W)
38
For ease of comparison, the metering categories are compiled and categorized again to follow
DOEs Building Activity for office buildings, (1) HVAC (2) Water Heating (3) Lighting (4)
Kitchen Equipment (5) Refrigeration (6) Office Equipment (7) Computers (8) Others
d) Occupancy Sensors
For counting occupants, a bi-directional infra-red occupant counter is used. Two infrared beams
are transmitted from the transmitter to the receiver unit. As someone passes, it breaks the nearest
infrared beam, and the internal counter increases by one. Therefore, the total count for the space
occupants at any given time is the difference between the counters. The data is stored in the
counter, and the data can be downloaded to a PC via the USB drive provided.
Hardware – USB Based Infra-red People Counter with PC Data Viewing (PC-002U)
Dimensions - 116.4 x 68.6 x 22.3 mm
Material - ABS Plastic, Power supply - 2 x 1.5v AA approximately 120uA
Battery life - Approximately 4 years, Maximum counting distance: 33 feet (10 m)
Maximum Data Memory: >1month at 1min. interval
Figure 3.9: USB Based Infra-red People Counter
Figure 3.10: Entryway location of infrared bi-directional occupant sensor
39
3.1.4 Data Monitoring and Collection
In total, 25 parameters were monitored at a 1 to 15-minute interval (Table 3.4). While most data
is collected on a minute resolution, the database will consist of data compiled at 15 minutes
resolution. In addition, daily data is compiled as well.
Table 3.4: Time Resolution collected for each parameter
No. Category Parameter Time Resolution
(Collected)
Time Resolution
(Compiled)
1 Climate Outdoor Temperature 1 min 15 mins
2 Relative Humidity 1 min 15 mins
3 Wind Speed 10 min 15 mins
4 Solar Radiation 1 min 15 mins
5 BMS Chilled Water Temp 1 min 15 mins
6 Chilled Water Temp 1 min 15 mins
7 Outside Air Temp 1 min 15 mins
8 AHU Supply Air 1 min 15 mins
9 AHU Return Air 1 min 15 mins
10 AHU Static Pressure 1 min 15 mins
11 OSA Cfm 1 min 15 mins
12 Data Logger Indoor Temperature 10 min 15 mins
13 Relative Humidity 10 min 15 mins
14 Intensity 10 min 15 mins
15 Carbon Dioxide 10 min 15 mins
16 Energy HVAC 1 min 15 mins
17 Water Heating 1 min 15 mins
18 Lighting 1 min 15 mins
19 Kitchen Equip 1 min 15 mins
20 Refrigeration 1 min 15 mins
21 Office Equipment 1 min 15 mins
22 Computer 1 min 15 mins
23 Others 1 min 15 mins
24 Total 1 min 15 mins
25 Occupant Number 15 mins 15 mins
40
All sensors mentioned in Section 3.1.3 were allowed to run for a period of Six Weeks (Table 3.5).
Extracting and gathering all building-related data through weather station, building automation
systems, data loggers, occupancy sensors, energy use sub-metering, and then compiled in a
database. Since each tool has a specific software and does not share the same extension for data
files, all data are converted and combined in spreadsheet file type (Microsoft Excel).
Table 3.5: Breakdown of study period in weeks
Week Period
0 02 Aug to 08 Aug
1 11 Oct to 17 Oct
2 18 Oct to 24 Oct
3 25 Oct to 31 Oct
4 1 Nov to 7 Nov
5 8 Nov to 14 Nov
6 15 Nov to 21 Nov
3.1.5 Data Pre-Processing
The Air Handling Unit (AHU) unit was found to be a major contributor to the energy consumption
of this space and it runs on a fixed schedule from 6am to 6pm on a workday. Since, the objective
of this study is to find the relationship of occupants and energy use, data within the shut-off time
will not be taken into consideration. As explained later in Chapter 5.1, taking the whole data set
would saturate energy consumption values with a large variation. Therefore, of the 3,537 data sets
of 15 min interval is reduced to 30 workdays of 1470 data points of 15 min interval (6am to 6pm).
Other pre-processing steps were removing outliers, converting to metric units and syncing all data
collected from various sensors to represent the same time.
3.1.6 Data Mining
A sensitivity analysis on the collected data using simple linear regression will provide an overview
on how parameters could influence building energy consumption based on real collected data, by
category as a whole or narrowed down to a particular parameter.
To develop the regression model, multiple methods and specialized software are available, such
as SPSS Statistics (IBM 2014), SAS, STATSTICA, R and etc. In this research, a statistical
41
analysis tool, Minitab
®
17 (Minitab 2014b) is adopted for data analysis and regression model
development. By using Minitab, a large amount of data can be processed for basic statistical
analysis, regression and correlation analysis, hypothesis tests, model validation, prediction, and
graphs making, etc. The correlation between each parameter and energy use could be analyzed by
calculating Pearson’s correlation coefficient. Then stepwise regression models is used to determine
the largest impact from occupants. The effect of occupancy, climate, or indoor conditions on
energy consumption throughout the periods are analyzed. Types of statistics methods used are:
a) Clustering
Data is merged into different clusters so that instances in the same cluster have a high similarity.
b) Correlation
Correlation value provides the Direction and Strength of the points; it’s always between -1 and
+1, with -1 being a negative slope and +1 a positive slope. The strength does not necessarily mean
the correlation is statistically significant and does not explain causation.
c) Simple Linear Regression Analysis
Simple Linear Regression is always a comparison to what we would have if we only two variables,
in this case between each of the independent variables versus the dependent variables.
Figure 3.11: Illustrated relationship between independent and dependent variables
From Figure 3.11, the value of one dependent (y) variable, is a function of the other independent
(x) variable, [y = f(x)]. X1, X2, and X3 are the independent (predictors) while Y is the dependent
42
(response) variable. From the analysis, effects of an independent variable can be seen by a)
identifying a trend or b) forecasting an impact of change or c) Predicting a trend or future value
(Solutions n.d.). Simple linear regression is designed to find the best fitting line through the data
that minimizes the Sum of Squares Error (SSE) generating a correlation and a regression equation
explaining the nature of the best-fit line.
d) Multiple Linear Regression / Stepwise
Figure 3.12: Illustrated difference between simple linear regression and multiple regression
Multiple linear regression is a many-to-one relationship used “to describe data and to explain the
relationship between one dependent variable and two or more independent variables” (Solutions
n.d.). MLR is used to find out the change of each influencing factor (i.e. climate, indoor conditions,
and user presence) on the change of energy consumption if other factors remain constant. Some
Important Questions multiple linear regression can answer is:
a) Is at least one of the predictors (climate, user presence, indoor conditions) useful in
predicting the response (energy use)?
b) Do all / one/ combination of the predictors (climate, user presence, indoor conditions) help
to explain a trend X in energy consumption,
c) The significance level of how well the model fits the data/ how accurate is the prediction.
Upon analyzing data, the indoor conditions of the said space were very constant. This is because
the space is a controlled environment; therefore, the indoor condition variables becomes dependent
43
on the energy use instead. The analysis structure or variables left to be predictors for this analysis
is climate parameters and occupants, depicted in Figure 3.13.
Figure 3.13: Final structure for multiple regression analysis
Figure 3.14: Illustration of relationship of variables for Climate and Occupants with Energy.
Sub-predictors (parameters) that are Outdoor Temperature, Relative Humidity, Solar Radiation,
Wind, and Precipitation are measured. However, because there were no operable windows in this
facility, Wind Speed and Wind Direction are removed from the predictor set. In addition, from the
44
collected period, there were no precipitation data. This is also withdrawn from the predictor set.
Leaving the final structure for Multiple Linear Regression Analysis as illustrated in Figure 3.15.
Figure 3.15: Illustration of final structure of variables for Multiple Linear Regression Analysis.
An ideal multi-regression analysis is having dependent variables with some positive correlation
with the dependent variables. If there is no relationship to begin with, it will not provide additional
information to the model, if it were included in the predictor set. Another important aspect is
having the independent variables not to be potentially related to each other (potential
multicollinearity). It will then be hard to tell which is affecting the dependent variable. This
condition can produce ambiguous results that are associated with unstable regression coefficients.
(Fumo and Biswas, 2015). The ideal is for all of the independent variables to be correlated with
the dependent variable but NOT with each other.
Multi Regression Prep:
1) Generate a list of potential variables; independent(s) and dependent
2) Collect data on the variables
3) Check the relationships between each independent variable and the dependent variable
using scatterplots and correlations
4) Check the relationships among the independent variables (Solutions n.d.)
Stepwise is then used as a technique where variables are entered into the regression model in the
order of explanatory power, to identify the strength of the effect that the independent variable have
on a dependent variable.
45
Quality of the model:
The results from regression analysis or more accurately the quality of line-of-fit from the observed
data can be judged by using the coefficient of determination (R-sq or R
2
). The value of R
2
varies
between 0 and 1; for example, a value of R
2
= 0.9 indicates that 90% of the total variability in the
response variable is accounted for by the predictor variables. R-sq values from the stepwise
regression are compared and drawn conclusions from. Other methods used to determine the quality
of the model is the p-value, where values below 0.05 are considered ideal and renders the results
significant.
3.1.7. Results and Document Presentation
Results are documented using graphical images and tabulated data in a word document.
3.2 Gantt Chart
Table 3.6: Gantt chart
Task AUG SEPT OCT NOV DEC JAN FEB
Literature Review
Problem definition
Data source selection
Install sensors
Data collection
Data pre-processing
Data mining
Results
Document presentation
Chapter 3 discussed the methodology carried out. The following chapter presents graphical
representations and discussions of data collected and some key findings.
46
CHAPTER 4 : DATA COLLECTED
Overview
In Chapter 3: Methodology, description of the sensors and tools that were installed to collect data
were presented. Data collected is presented here for the total of six weeks for each parameter.
Although data was collected as minute resolutions for climate, indoor temperature, and energy use;
Occupant sensor was only able to gather data at 15 min intervals. With that the database is
compiled at 15 minutes intervals example (4:00, 4:15, 4:30 and 4:45). Data collected starts from
11th October 2015 and ends 21st November 2015.
4.1 Climate Data
a) Outdoor Temperature in degree Celsius
Figure 4.1: Collected data for Outdoor Temperature
Table 4.1: Peak temperature values for by week
Week 1 & 2 Week 3 & 4 Week 5 & 6
Date Range Oct 11 to Oct 24th Oct 25
th
to Nov 7th Nov 8
th
to Nov 21st
Max Temperature (⁰C) 35.44 32.00 32.28
Min Temperature (⁰C) 16.89 11.44 9.39
Delta Temperature 18.55 20.22 22.89
Due to significant variance in the temperature range, data is broken down and analyzed in three
parts, i.e. Week 1 and 2, Week 3 and 4; and Week 5 and 6. Table 4.1 summarize significant
temperature data for three of two weeks set. Of the six weeks study period, Week 1 and 2 had
higher temperatures compared to the later weeks and Week 5 and 6 had lower temperatures
0
5
10
15
20
25
30
35
40
Temperature (⁰C)
Outdoor Temperature (⁰C)
47
compared to the first four weeks. The maximum temperature of the study period 35.44 ⁰C fell in
Week 1 and 2 and the minimum temperature of the study period, 9.39 ⁰C fell in Week 5 and 6. This
temperature difference seems to be due to lower solar radiation levels as seen in Figure 4.2. Solar
Radiation levels appear to be lower in Week 5 and 6 compared to levels in Week 1 and 2.
b) Outdoor Solar Radiation (Watt/m
2
)
Figure 4.2 (a) & (b): Collected data for Outdoor Solar Radiation
0
100
200
300
400
500
600
700
800
900
1000
0
100
200
300
400
500
600
700
800
900
1000
Solar Radiation (W/m
2
)
Solar Radiation (W/m2)
Outdoor Solar Radiaton (W/m
2
)
Rad. (W/m2) Daily Peak Average Peak Value
12000
13000
14000
15000
16000
17000
18000
0
5000
10000
15000
20000
25000
Week 1 Week 2 Week 3 Week 4 Week 5 Week 6
Average Solar Radiation (W/m
2
)
Daily Solar Radiation (W/m2)
Cummulative Daily Solar Radiation (W/m2)
Monday Tuesday Wednesday Thursday
Friday Saturday Sunday Average
48
Figure 4.3: Outdoor Condition for (a) Week 1 and 2, (b) Week 3 and 4, and (c) Week 5 and 6.
A weekly breakdown of outdoor temperatures is shown in Figure 4.3 (a), (b) and (c) and includes
relative humidity values. It was found the hottest recorded temperature in six weeks fell in Week
1 at 3:00 PM (Figure 4.4) while the lowest temperature fell on Nov 17th (Figure 4.5).
0
20
40
60
80
100
0
5
10
15
20
25
30
35
40
Relative Humidity (%)
Temperature (⁰C)
Outdoor Condition for Week 1 and 2 (⁰C)
0
20
40
60
80
100
0
5
10
15
20
25
30
35
40
Relative Humidity (%)
Temperature (⁰C)
Outdoor Condition for Week 3 and 4 (⁰C)
0
20
40
60
80
100
0
5
10
15
20
25
30
35
40
Relative Humidity (%)
Temperature (⁰C)
Outdoor Condition for Week 5 and 6 (⁰C)
Outdoor Temperature Hum (%RH)
49
Figure 4.4: Comparison of recorded for hottest day and average TMY2 values for October.
Figure 4.5: Comparison of recorded for coldest day and average TMY2 values for November
Comparing with historical averages (TMY2 weather file) the recorded data shows a typical pattern
for both temperatures and relative humidity. However, it shows a significant difference in values,
about 8 to 15 degrees Celsius in Figure 4.4 depending on the time of day. This difference is not as
apparent in the colder week (Figure 4.5). This may seem due to Downtown Los Angeles being
warmer than most of Los Angeles due to urban heat island effect and also the location of TMY2
recorded files are based off the Los Angeles Airport, which is located closer to coast.
0
10
20
30
40
50
60
70
80
90
100
0
5
10
15
20
25
30
35
40
3:30 AM
4:15 AM
5:00 AM
5:45 AM
6:30 AM
7:15 AM
8:00 AM
8:45 AM
9:30 AM
10:15 AM
11:00 AM
11:45 AM
12:30 PM
1:15 PM
2:00 PM
2:45 PM
3:30 PM
4:15 PM
5:00 PM
5:45 PM
6:30 PM
7:15 PM
8:00 PM
8:45 PM
9:31 PM
10:15 PM
11:00 PM
11:45 PM
Relative Humidity (%)
Temperature ⁰C
Hourly comparison of recorded climate and historical average
Temperature Temperature TMY2 Relative Humidity (%) RH TMY2
0
10
20
30
40
50
60
70
80
90
100
0
5
10
15
20
25
30
35
40
12:00 AM
12:45 AM
1:30 AM
2:15 AM
3:00 AM
3:45 AM
4:30 AM
5:15 AM
6:00 AM
6:45 AM
7:30 AM
8:15 AM
9:00 AM
9:45 AM
10:30 AM
11:15 AM
12:00 PM
12:45 PM
1:30 PM
2:15 PM
4:00 PM
4:45 PM
5:30 PM
6:15 PM
7:00 PM
7:45 PM
8:30 PM
9:15 PM
10:00 PM
10:45 PM
11:30 PM
Relative HUmidity (%)
Temperature ⁰C
Hourly comparison of recorded climate and historical average
Low Temperature Temperature TMY2 Relative Humidity RH TMY2
50
c) Outdoor Wind Direction
Figure 4.6 (a) and below (b): (a) Snapshot of Wind Rose from Los Angeles TMY2 Weather file
and (b) Collected Data for Cumulative Wind Direction
The TMY2 historical average of wind rose taken from Climate Consultant, October and November
showed winds coming from mostly the West direction. The collected data show similar trends.
North
NNE
NE
ENE
East
ESE
SE
SSE
S
SSW
SW
WSW
W
WNW
NW
NNW
Wind Direction
51
4.2 Indoor Conditions – Building Management System Data & Loggers
a) Zonal Temperature from 17 zones
Figure 4.7: Zonal Temperature from Building Management System
Figure 4.8: Screenshot of Zones in Office Layout
The cooling set point for the office is at 23⁰C (74.0⁰F). The shaded band (ranging from 20 to 24C
(68 to 76F) in Figure 4.8 shows temperatures in the office satisfies the ASHRAE Standards for
indoor comfort. The values outside the comfort band is explained by weekends where the space is
unoccupied and the peak is due to the server room.
16
18
20
22
24
26
28
30
32
34
36
38
40
Temperature ⁰C
Zonal Temperatures
Zone01 Zone03 Zone05 Zone06 Zone08 Zone10
Zone11 Zone12 Zone13 Zone14 Zone15 Zone16
Zone17 Zone18 Zone19 Zone20 Zone21
52
b) Indoor Spot Measurement of Temperature and Relative Humidity
Figure 4.9: Indoor Spot Measurement of Indoor Temperature and Relative Humidity
c) Indoor Spot Measurement of Lighting Intensity (lux)
Figure 4.10: Indoor Spot Measurement – Lighting Intensity (lux)
The Illuminating Engineering Society (IES) recommends illuminance levels for offices facilities
to be 200 to 500 lux depending on tasks as highlighted in Figure 4.10. This data show the office
could consider daylighting as a light source, as given hours of operation for an office facility
daylighting could be used at an advantage.
0
10
20
30
40
50
60
70
0
5
10
15
20
25
30
35
Relative Humidity (%)
Temperature (⁰C)
Indoor Spot Temperature and Retive Humidity (%)
Temp (C) RH, %
0
100
200
300
400
500
600
Lighting (lux)
Indoor Spot Lighting (lux)
Intensity (lux)
53
Figure 4.11: Average Lighting Levels for a typical workday, 6am to 6pm
d) Indoor Spot Measurement of Carbon Dioxide Levels (ppm)
Figure 4.12: Indoor Spot measurement – Indoor Carbon Dioxide Levels (ppm)
According to ASHRAE, a typical office, CO2 concentrations of about 1,000 to 1,200 ppm indicate
a sufficient ventilation rate of 7.5 L/s.person (15 cfm/person). Levels above these do not impose
direct health risk, however, is used as an indicator of occupant acceptance of odors. The collected
data show this office have a healthy ventilation rate, but not overall building air quality as it
requires other data such as carbon monoxide, volatile organic compounds, etc.
0
100
200
300
400
500
600
Lighting (lux)
Average Lighting level (lux) for workday (6am to 6 pm)
0
200
400
600
800
1000
1200
1400
parts per million (ppm)
Indoor Spot Carbon Dioxide Levels (ppm)
54
4.3 Sub-metering Data
a) Pie Chart for distribution of total energy use by end-use category
Figure 4.13: Pie Chart summarizing energy consumption by end uses
Figure 4.13 displays the summary of energy consumption from the six weeks of study. The top
three categories are in descending order Computers, Lighting, and AHU + Fan.
b) Cumulative Energy Consumption
Table 4.2: Cumulative Energy Consumption and Occupants for Random Weeks
Week Period / Total
Office Energy
Consumption
02 August to 08
August 2016
18 October to 24
October 2016
1 November to 07
November 2016
No. Of Occupants 53 45 36
Monday 17,068 15,228 13,274
Tuesday 17,696 15,919 12,591
Wednesday 17,321 15,693 12,723
Thursday 15,039 15,803 13,266
Friday 9,617 9,660 7,585
Average (Wh) 15,348 14,461 11,888
AHU + Fan
23%
Water Heating
1%
Lighting
24%
Kitchen
Equipment
6%
Refrigerator
1%
Office Equipment
5%
Computers
33%
Others
7%
Energy Consumption by End-Use Category
55
c) Sub-metering Data – AHU Fan
Figure 4.14: Sub-metering data for AHU Fan Category
Figure 4.15: Day profile for AHU Fan Energy Use
Figure 4.15 shows energy consumption levels for the AHU Fan. The AHU operates on a fixed
schedule, which is from 6am to 6pm on a workday (Monday through Friday). As the indoor
condition has a controlled set point, the AHU can be seen to work harder at the beginning of the
day and then slows down when the indoor conditions are cooler. It picks up again through to the
middle of the day with additional gains.
0
2000
4000
6000
8000
10000
12000
Energy (Wh)
AHU Fan Energy Use (Wh)
6,000
7,000
8,000
9,000
10,000
11,000
12,000
Energy (Wh)
AHU Fan Energy Use (Wh)
Peak Day (Wh) Average (Wh)
56
d) Sub-metering Data – Water Heating
Figure 4.16: Sub-metering data for Water Heating Category
Figure 4.17: Day profiles for day of highest Water Heating Energy Use
Water Heating measured here refers to instant electric hot water in the male and female shower.
As some occupants cycle to work and freshen up in the morning, it explains the peak at 8:30am.
0
1000
2000
3000
4000
5000
6000
7000
Energy (Wh)
Water Heating Energy Use (Wh)
Water Heating (Wh)
0
200
400
600
800
1,000
1,200
1,400
Energy (Wh)
Average Water Heating Energy Use
57
e) Sub-metering Data – Lighting
Figure 4.18: Collected sub-metering data for lighting category
Figure 4.19: Day profile for day of highest Lighting Energy Use
It is noted lighting use peaks about 7:30 am and remains relatively constant throughout the work
day. When compared against the average occupant schedule, it does not co-vary. This is
explainable as the fixtures are motion-sensitive and does not take into account number of
occupants. If the first person enters the office, lights will be turned on as seen in the start of the
graph. This is also seen at the end of the day, where occupants start to leave, lighting use remains
the same. It is also noted that lighting starts at 2.8 kWh. This could mean some lights are switched
on all hours such as egress or corridor lights.
0
500
1000
1500
2000
2500
3000
3500
4000
Energy (Wh)
Lighting Energy Use (Wh)
0
5
10
15
20
25
30
35
40
45
2,600
2,700
2,800
2,900
3,000
3,100
3,200
3,300
3,400
Occupants
Energy (Wh)
Average Daily Lighting Energy Use (Wh)
Average Lighting Occupants
58
f) Sub-metering Data – Kitchen Equipment
Figure 4.20: Sub-metering data for Kitchen Equipment Category
Figure 4.21: Average day profile for Kitchen Equipment Energy Use
Kitchen equipment here refers to a coffee maker, coffee grinder, microwave, toaster, dishwashers,
and sink incinerator. From Figure 4.21, the equipments are most used from periods of 8am to 10am
and 12 to 2pm. Again, from the full data in Error! Reference source not found., when space is
unoccupied, or kitchen equipment not used, the energy use does not drop to zero.
0
500
1000
1500
2000
2500
3000
3500
4000
Energy (Wh)
Kitchen Equipment Energy Use (Wh)
0
200
400
600
800
1000
1200
1400
1600
1800
6:00 AM 8:00 AM 10:00 AM 12:00 PM 2:00 PM 4:00 PM 6:00 PM
100,000
150,000
200,000
250,000
300,000
350,000
Energy (Wh)
Average daily Kitchen Equipment Energy Use
Cummulative Average
59
g) Sub-metering Data – Refrigerator
Figure 4.22: Sub-metering data for Refrigerator Category
Figure 4.23: Average day profile for refrigerator energy use
The kitchen is also equipped with two commercial size refrigerators. It was not included in the
Kitchen Equipment category because it operates all day 24 hours. The daily profile pattern is
typical and can be related to how the kitchen space is used, peaks in the morning, constant all day.
0
50
100
150
200
250
300
350
400
450
500
Energy (Wh)
Refrigerator Energy Use (Wh)
0
20
40
60
80
100
120
140
160
180
200
Energy (Wh)
Average Daily Refrigerator Energy Use (Wh)
60
h) Sub-metering Data – Office Equipment
Figure 4.24: Sub-metering data for Office Equipment Category
Figure 4.25: Average daily profile of Energy Use Consumption for Office Equipment
0
500
1000
1500
2000
2500
3000
Energy (Wh)
Office Equipment Energy Use (Wh)
0
100
200
300
400
500
600
700
800
900
1,000
Energy (Wh)
Average Daily Office Equipment Energy Use Profile (Wh)
61
i) Sub-metering Data – Computer
Figure 4.26: Sub-metering data for Computer Category
Figure 4.27: Average daily profile of Energy Use Consumption for Computer Category
From the average use of energy consumption in the end-use category of Computers, the pattern
seems to follow the average occupancy levels of the space.
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
Energy (Wh)
Computer Energy Use (Wh)
0
5
10
15
20
25
30
35
40
45
3,000
3,500
4,000
4,500
5,000
5,500
6,000
Occupants
Energy (Wh)
Average Daily Computer Energy Use Profile (Wh)
Average Computer Average Occupancy
62
j) Sub-metering Data – Total Energy Consumption
Figure 4.28: Sub-metering data for Total Office Energy Consumption
Figure 4.29: Energy Consumption for 30 Working Days, 6am to 6pm
4.4 Occupancy Data
a) Occupant Data from Office Administration
The office was asked how many people were expected to be in the office at the start of each week.
Data from three weeks stood out (Table 4.3). In Week A, when the study started there were 53
people in the office. From August to mid-October, eight people left the company. Hence, there are
45 people left in the office. In November, the management level (9 people) in the company were
away, leaving 36 people in the office (Week C).
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
Energy (Wh)
Total Energy Consumption (Wh)
0
5
10
15
20
25
30
35
Energy Use (kWh)
Office Energy Consumption (kWh) Workday 6am to 6pm
63
Table 4.3: Cumulative Occupant Count for Random weeks
Week Week Period People
A 02 August to 08 August 2016 53
B 18 October to 24 October 2016 45
C 1 November to 07 November 2016 36
b) Occupant Data from Occupant Sensor
Figure 4.30: Occupancy Schedule collected from Occupant Sensors
Figure 4.31: Graphical summary for Occupants
Chapter Summary
This chapter presented the data collected using the tools and sensors placed. Although each tool
had a different extraction software, all data presented here were combined and synced with time
resolutions at 15 minutes in an Excel spreadsheet database. The next chapter will discuss
relationship results from statistical analysis of the data.
-10
0
10
20
30
40
50
60
70
Number
Occupants
1st Quartile 20.000
Median 32.000
3rd Quartile 37.000
Maximum 66.000
27.134 28.526
31.000 33.000
12.308 13.293
A-Squared 52.16
P-Value <0.005
Mean 27.830
StDev 12.781
Variance 163.360
Skewness -0.816874
Kurtosis -0.336648
N 1299
Minimum 0.000
Anderson-Darling Normality Test
95% Confidence Interval for Mean
95% Confidence Interval for Median
95% Confidence Interval for StDev
60 50 40 30 20 10 0
Median
Mean
33 32 31 30 29 28 27
95% Confidence Intervals
Summary Report for Occupants (No.)
64
CHAPTER 5 : DATA ANALYSIS
Chapter 4 presented data collected from the weather station, indoor data loggers, sub-metering
sensors and occupant sensor. Chapter 5 will discuss results gathered from statistical analysis of the
parameters carried out using the Minitab statistical software. Data is first analyzed as a single
variable using box plots, interval plots, clustering and then with two variables with simple linear
regression. Lastly, a stepwise regression analysis was employed to determine the impacts of
predictors on categorized end-use energy consumption.
5.1 Box Plot
A boxplot presents a graphical summary of the sample distribution. It is a straightforward and
standardized way to show data distribution based on five items, minimum, first quartile, median,
third quartile and maximum. Boxplots are also useful to see which way the data is skewed (if any).
Figure 5.1: Box Plot of Energy Consumption of whole study period
Energy consumption data was first used as a whole, i.e.; all data sets including after hours and
weekends. The boxplot display in Figure 5.1 shows its distribution which consist of:
1. Blue box represents the middle 50% (interquartile range) of the data, values between 25%
and 75% of the data. (7.726 kWh to 19.187 kWh )
2. The median value indicated by the horizontal line inside the box (9.514 kWh)
3. The upper 50% of the data are represented by everything above the median line. The data
is skewed towards the upper percentile.
35
30
25
20
15
10
5
Total Energy Consumption (kWh)
Boxplot of Total Energy Consumption (kWh)
65
4. Lines (called “whiskers”) extending from the box representing the upper and lower 25%
of the distribution (excluding outliers)
The focus of this research is how occupants affect building performance, therefore from the six
weeks of data collected, 3922 data sets of 15 min intervals are reduced to 1470 data sets of 15 min
intervals, which focus on 30 workdays (6am to 6pm). The twelve-hour range was chosen because
the AHU on the office floor operates on a fixed schedule which is 6am to 6pm. As this category
takes up a large portion of energy use, times outside the schedule is not considered. Including the
data will affect the overall energy consumption impact study.
Figure 5.2: Box Plot of Energy Consumption of 30 working days (6am to 6pm)
Outliers from the data (an observation that is beyond the normal) are removed. The boxplot display
in Figure 5.2 is a graphical summary of the distribution showing its shape, central tendency and
variability, which consist of:
5. Rectangular box (blue) represents the middle 50% (interquartile range) of the data. This
are values between 25% and 75% of the data. (18.307 kWh to 22.540 kWh )
6. The median value indicated by the horizontal line inside the box (20.546 kWh)
7. The upper 50% of the data (735 data points) are represented by everything above the
median line.
8. Lines extending from the box representing the upper and lower 25% of the distribution
With only one variable which is the total energy use, and using no other information, the best
prediction for an average value is the mean value of the whole sample (1470 points). From this
35
30
25
20
15
10
5
Total Energy Consumption (kWh)
Boxplot of Total Energy Consumption (kWh)
66
box plot, we can answer the question “How much does is the average energy use of this office
space?” or if we were required to predict the next value of energy consumption, it would be the
mean which is 20.217 kWh.
Comparing both box plots from Figure 5.1 and Figure 5.2, data from the reduced data set ( 6am to
6pm) have a smaller variance or is less saturated. The remaining analysis uses the 6am to 6pm
workday time range.
Figure 5.3: Average Energy Use by End-Use Category
The average use is calculated for different end-use category. On average, HVAC uses the most
energy followed by computer and lighting.
5.2 Aggregated Data
The process here merges data into various groups so that instances in the same group have a high
similarity, in this case, the number of occupants. Data from three weeks with a significantly
different number of people for the same working space were used. In Week A, there were 53 people
on the office payroll. In Week B, there were 45 people on the office payroll (18% less than Week
A). While in Week C, the management team left for a meeting leaving 36 people using the space
(25% less than Week B). Interval plot is used to compare visually energy use among the three
group on one graph; it is interesting to see if change occurs in energy use if occupants change.
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
Energy Use (kWh)
Average Energy Use by End-Use Category
67
Figure 5.4: Interval Plot for Energy Use
Table 5.1: Tabulated Occupancy Usage for Weeks A, B, and C
Week Ref # A B C
Week Period 02 August to 08
August 2016
18 October to 24
October 2016
1 November to 07
November 2016
No. Of Occupants 53 45 36
Mean Energy Use (Wh) 15,348 14,461 11,888
Median Energy Use (Wh) 17,068 15,693 12,723
Minimum Energy Use (Wh) 9,617 9,660 7,585
Maximum Energy Use (Wh) 17,696 15,919 13,274
Peak Temperature (⁰C) 31.2 32.1 26.7
The interval plots are useful for comparing groups for its central tendency. From Figure 5.4,
interval plots display a graphical summary of the distribution, its central tendency, and variability.
A 95% confidence interval means we are 95% certain the range of values contains the true mean
of the population. From the interval plot, the central tendency shows Week A with the higher
number of occupants consume more energy than Week C with the lowest number of occupants
and Week B falls between the two. The same applies for Median, Minimum, and Maximum values
which are tabulated in Table 5.1.
With this aggregated data, we can tell occupants do affect energy use. However this method is not
accurate enough as another variable (outdoor temperature) could also play a role in the impact. It
Week C Week B Week A
18000
16000
14000
12000
10000
Data
Interval Plot of Energy Use
95% CI for the Mean
Individual standard deviations are used to calculate the intervals.
68
was discussed in Chapter 4.1 that temperatures early in the study were higher. Such aggregated
method is insufficient to identify causation of energy use. Hence, such method is not appropriate
to conclude the relationship between occupants and energy consumption.
5.3 Correlation Analysis
The correlation analysis evaluates the covariance between two variables. The correlation
coefficient explains the extent to which two variables tend to change together, describing both the
strength and the direction of the change if any (positive, negative, near zero). Two different
correlation analyses are the Pearson and Spearman. The Pearson correlation evaluates between
two continuous variables while the Spearman correlation evaluates based on ranked values
involving ordinal variables (Minitab 17 n.d.). As our data is not in ordered, (The temperature does
not increase from Day 1 rather it fluctuates throughout the day), the Spearman correlation is not
suitable for this analysis and the Pearson is used.
The Pearson’s correlation coefficient (r) is always a value between -1 and +1 showing a measure
which scales is independent of the scale of the variables themselves. This is very handy as it allows
comparison of variables that are measured in different ways for example; temperature outside and
energy usage which are in different metrics. Correlation is only applicable to linear relationships.
By squaring the correlation coefficient (r) to calculate R-sq., it will represent the proportion of
variability in one variable that is accounted for by another variable. R-sq is also known as the
coefficient of determination.
The p-value is for determining whether the results are statistically significant. This is used to
determine whether to accept or reject a hypothesis. A p-value ranges from 0 to 1. The p-value is
traditionally compared to α (alpha) values of less than 0.05. From here on, all p-values which are
lower than 0.05 is statically significant. Before computing correlations, a visual interpretation
between Independent Variable (IV) and Dependent Variable (DV) is done using scatterplots.
a) Correlation between Occupant and Total Energy Consumption
In general, the lower the number of occupants will result in a lower energy use and vice versa.
Using correlation, the strength of that direction can be found. Since the energy use is dependent
on the number of occupants, the energy use is the dependent (y) variable and the occupant is the
independent (x) variable.
69
Figure 5.5: Scatterplot of Total Energy Consumption (kWh) vs. Occupants (No.)
In the scatterplot above Figure 5.5, plainly descriptive it is not randomly all over, clearly upwards,
however, more centered blob on the right. Using the Minitab tool, the Pearson correlation of Total
Energy Consumption (kW) and Occupants (No.) is 0.514 with a p-value of 0.000, which shows
significance.
Equation 5.1: Relationship Rule of Thumb
𝐼𝑓 |𝑟 | ≥
2
√𝑛 , then a relationship exists
Solving Equation 5.1, for 1470 data points (n), the correlation (r) 0.514 is strong. At 0.514, there
is indeed a positive linear relationship between these two variables.
b) Correlation between Occupants and Energy Use by End-Use Category
Table 5.2: Tabulated Correlation between Occupants and Energy Use by Category
Refrigerator
(kWh)
Kitchen
Eq.(kWh)
AHU
(kWh)
Lighting
(kWh)
Office
Eq.(kWh)
Computer
(kWh)
Total
(kWh)
Occupant 0.065 0.155 0.288 0.296 0.328 0.396 0.514
0.014 0.000 0.000 0.000 0.000 0.000 0.000
Cell Contents: Pearson correlation
p-value
The Pearson’s correlation are all positive values which indicate that as occupants increase, energy
use in each category increases. There is a positive relationship between this predictor (occupants)
60 50 40 30 20 10 0
35
30
25
20
15
10
Occupants (No.)
Total Energy Consumption (kWh)
Scatterplot of Total Energy Consumption (kWh) vs Occupants (No.)
70
and the response (energy use). In accordance, the largest positive relationship between occupants
is Computer, followed by Office Equipment, Lighting, AHU, Kitchen Equipment and Refrigerator.
In these results, all the p-values are less than the significance level of 0.05, which indicates that
the correlation coefficients are significant.
c) Correlation between Climate and Energy Consumption
Table 5.2: Tabulated Correlation between Total Energy Consumption and Predictors
Outdoor
Temperature
Relative
Humidity
Solar Radiation Occupants
Total Energy
Consumption (kWh)
0.723 -0.0432 0.578 0.514
0.000 0.112 0.000 0.000
Cell Contents: Pearson correlation
p-value
The Correlation between Outdoor Temperature and Total Energy Consumption is the highest
followed by Solar Radiation and Occupants. The p-value of Relative Humidity is higher than 0.05
making it non-significant.
5.4 Simple Linear Regression
Figure 5.6: Simple Linear Regression between Independent Variables and Dependent Variables
71
a) Occupants vs. Total Energy Use
From Correlation, it was found that occupants affect energy use in a positive relationship, the
coefficient of impact can be obtained using simple linear regression. Minitab gave us an equation
for this graph Figure 5.7:
Equation 5.2
𝐸𝑛𝑒𝑟𝑔𝑦 𝐶𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 (𝑘𝑊 ℎ) = 16.43 + 0.1275 𝑂𝑐𝑐𝑢𝑝𝑎𝑛𝑡𝑠
For every increase of 1 occupant, we would expect or predict an increase of 0.1275 kWh of energy
usage. The intercept 16.43 show that if the number of occupants is zero, then the
expected/predicted energy use is 16.43kWh. R-sq of 23.8% from Figure 5.7 is the percentage of
response variation that is explained by its relationship with one or more predictor variables.
Figure 5.7: Fitted Line Plot for Total Energy Use against Occupants
The fitted line plot shows the line is upwards. However, it is noted visually that the data varies
larger between energy consumption at the front compared to the back of occupancy data. Data is
further analyzed by splitting the occupants into two groups, namely 0 to 22 number of occupants
and 23 to 44 number of occupants. This provides more information if the line equation holds true
for the entire occupant group. If both graphs have the same slope with Figure 5.7 , the lesser
occupants give no additional information to the line.
60 50 40 30 20 10 0
35
30
25
20
15
10
S 3.00709
R-Sq 23.8%
R-Sq(adj) 23.7%
Occupants (No.)
Total Energy Consumption (kWh)
Fitted Line Plot
Total Energy Consumption (kWh) = 16.43 + 0.1275 Occupants (No.)
72
Pearson correlation = 0.121; P-Value = 0.018
Pearson correlation = 0.253; P-Value = 0.000
Figure 5.8 (a) & (b): Fitted line plot for (a) 0 to 22 and (b) 23 to 44 number of occupants
Both graphs have a p-value of lower than 0.00, which denotes results are statistically significant.
The correlation for occupant group 0 to 22 is 0.121 and for occupant group 23 to 44 is 0.253. This
shows that the change in the larger group is stronger or more positive than the former group. From
both the Pearson correlation values and equations, we can say a number of occupants do affect
energy use, with the 23 to 44 range affecting more than 0 to 22 range of occupants. The first
regression model of 0 to 22 occupants has an R-sq. value of 1.5%, which explains only 1.5% of
25 20 15 10 5 0
27.5
25.0
22.5
20.0
17.5
15.0
12.5
10.0
S 3.94615
R-Sq 1.5%
R-Sq(adj) 1.2%
Occupants (No.)
Total Energy Consumption (kWh)
Fitted Line Plot
Total Energy Consumption (kWh) = 16.83 + 0.06433 Occupants (No.)
45 40 35 30 25 20
35
30
25
20
15
10
S 2.56219
R-Sq 6.4%
R-Sq(adj) 6.3%
Occupants (No.)
Total Energy Consumption (kWh)
Fitted Line Plot
Total Energy Consumption (kWh) = 16.52 + 0.1287 Occupants (No.)
73
the variance while the second one explains 6.4%. The second model with the higher variance
explains that the data points being closer to the fitted regression line.
b) Outdoor Temperature and Energy Use
Pearson correlation = 0.723 (P-Value = 0.000)
Figure 5.9: Fitted Line Plot for Energy Consumption vs. Outdoor Temperature
c) Solar Radiation and Energy Use
Pearson correlation = 0.578 (P-Value = 0.000)
Figure 5.10: Fitted Line Plot for Energy Consumption vs Solar Radiation
35 30 25 20 15 10
35
30
25
20
15
10
S 2.42495
R-Sq 52.1%
R-Sq(adj) 52.0%
Outdoor Temperature ( ⁰C)
Total Energy Consumption (kW)
Fitted Line Plot
Total Energy Consumption (kW) = 8.700 + 0.5242 Outdoor Temperature ( ⁰C)
900 800 700 600 500 400 300 200 100 0
35
30
25
20
15
10
S 2.88687
R-Sq 32.2%
R-Sq(adj) 32.1%
Solar Radiation (W/m2)
Total Energy Consumption (kW)
Fitted Line Plot
Total Energy Consumption (kW) = 17.73 + 0.008158 Solar Radiation (W/m2)
74
d) Outdoor Relative Humidity and Energy Use
Pearson correlation = 0.0432 (P-Value = 0.112)
Figure 5.11: Fitted Line Plot for Energy Consumption vs. Outdoor Relative Humidity
From the climate predictor set, outdoor temperature seems to be the most correlated with total
energy consumption with highest R-sq value of 52.1%. From simple linear regression analysis,
outdoor temperature impacts energy consumption greater than solar radiation and relative humidity.
However, this analysis limits to only two variables. To include more variables and check if this
hold true, if outdoor temperature affects energy use regardless of the change in solar radiation,
multiple linear regression (MLR) is used. In order to be included as multiple predictors in MLR,
the relationship between IV and DV must have some linear relationship summarized below:-
Scatterplot Summary
1. Occupancy appears highly correlated with Energy Use
2. Temperature appears highly correlated with Energy Use
3. Solar Radiation appears highly correlated with Energy Use
4. Relative Humidity appears slightly correlated with Energy Use
It is noted that correlation does not imply causation. Unless data was collected from properly
controlled experiments such as testing one variable being gradually increased at a constant rate in
a fixed environment, can we determine if the relationship is accurate.
90 80 70 60 50 40 30 20 10 0
35
30
25
20
15
10
S 3.49532
R-Sq 0.2%
R-Sq(adj) 0.1%
Hum (%RH)
Total Energy Consumption (kW)
Fitted Line Plot
Total Energy Consumption (kW) = 20.62 - 0.007641 Hum (%RH)
75
5.5 Multiple Linear Regression
Figure 5.12: Illustration of final structure of variables for Multiple Linear Regression Analysis.
Multiple regression is an extension of simple linear regression. Multiple Regression is used to
model the linear relationship between continuous responses and more than two continuous
predictors. Stepwise regression in multiple regression analysis is a step-by-step analysis when
there are many variables to identify a useful subset of the predictors. In the Minitab tool, stepwise
regression procedure both adds and removes predictors one at a time. “Minitab stops when all
variables not included in the model have p-values that are greater than a specified Alpha-to-Enter
value and when all variables that are in the model have p-values that are less than or equal to a
specified Alpha-to-Remove value.” (Minitab)
Before Multiple Linear Regression Analysis (MLR) there has to be some linear relationship for it
to become an Independent Variable (Predictor) in MLR. If there is no connection between the
dependent variable, adding it as a predictor will not contribute any value to the data. From the
scatter plots, since all have some linear relationship we can include them into MLR. The goal here
is to narrow the predictor variables into a list of the top predictors of energy usage.
5.5.1 Stepwise Regression Example
Stepwise regression was conducted to produce a more predictive model when more than two
variables are involved. This method which consists of a progressive introduction of predictor
76
variables automatically took out variables that did not contribute to analysis and retained variables
with a higher significance level at the top. To identify a subset of useful predictors, and ranked in
order, stepwise regression is an appropriate analysis. Although many predictors are measured
earlier in this study, not all possible predictors are included because of availability of data;
1. Uses large amounts of trustworthy data (real-time collection) and a small number of predictors
that have well established causal relationship (done in Section 5.4)
2. Uses logic for including variables (Although wind speed is tempted to be used as a predictor,
the office space is enclosed and have no operable windows)
Thus, the final list of responses and predictors for this study are as below:
Responses include:
1. AHU (Wh)
2. Lighting (Wh)
3. Kitchen Equipment (Wh)
4. Refrigerator (Wh)
5. Office Equipment (Wh)
6. Computer (Wh)
7. Total (kWh)
Continuous Predictors:
1. Outdoor Temperature (⁰C)
2. Outdoor Relative Humidity (%)
3. Solar Radiation (W/m
2
)
4. Occupancy (No.)
Stepwise regression analysis generates results to describe the statistical relationship between one
or more predictor variables and the response variable. Figure 5.13 is a sample screenshot of step-
wise results from how the AHU energy end-use category is being responsive to its predictors; Solar
Radiation, Outdoor Temperature, and Occupants. The results are interpreted below Figure 5.13:
R-squared denoted in the figure as R-sq is a statistical measure of how close the data are to the
fitted regression line. In general, the higher the R-square, the better the model fits your data.
Theoretically, if a model could explain 100% of the variance, all the data points would fall on the
fitted regression line. However, a low R-squared does not mean values are bad or not applicable,
fields that attempt to predict human behavior has low R-squares simply because humans are
77
stochastic in nature. Since we have statistically significant predictors, one can still use this value
to draw important conclusions about how changes in the predictors are associated with the
response. Here, Solar Radiation can explain 25.81% of energy consumption in the AHU end-use
category.
Stepwise Regression Analysis: AHU (Wh)
Figure 5.13: Stepwise Regression Analysis of AHU End-Use Category
78
A low p-value (<0.05) indicates that the predictor is likely to be a meaningful addition to the model
because changes in the predictor’s value are related to change in the response variable. Conversely,
a larger p-value (>0.05) shows insignificance or can be said suggests that changes in the predictor
are not associated with changes in the response. In the output above, it is demonstrated that the
predictor variables of Outdoor Temperature, Solar Radiation and Occupants are significant
because their p-values are 0.000.
Regression coefficients denoted by “Coef” in the figure above represent the mean change in the
response variable (HVAC) for one unit of change in the predictor variable while holding other
predictors (Outdoor Temperature, Solar Radiation, and Occupants) in the model constant. This
control is important because it isolates the impact of one variable from all the others in the model.
In algebra terms, these coefficients act as slopes and compared as a slope ratio. From these slope
coefficients, regression analysis generates an equation to describe the statistical relationship
between one or more predictor variables and the response variable. From the results in Figure 5.13
that shows how HVAC (AHU) energy is affected by the predictors, generates the equation;
Equation 5.3
𝐻𝑉𝐴𝐶 (𝑊 ℎ) = 3677 + 169.4 𝑂𝑢𝑡𝑑𝑜𝑜𝑟 𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 (
0
𝐶 ) + 3.611 𝑆𝑜𝑙𝑎𝑟 𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛 (
𝑊 𝑚 2
)
− 25.07 𝑂𝑐𝑐𝑢𝑝𝑎𝑛𝑡𝑠 (𝑁𝑜 . )
The equation shows that the coefficient for Outdoor Temperature is 169.4 Watt-hour. This
coefficient indicates that for every additional outdoor temperature in degree Celsius, one can
expect energy use in the HVAC Category to increase by an average of 169.4 Watt-hour. This
concept holds true for the other predictors in the equation, Solar Radiation and Occupants.
The constant term in the front, 3677 also known as the y-intercept is simply just that, the value at
which the fitted line crosses the y-axis. That is, if the outdoor temperature, solar radiation, and
occupants are zero the HVAC will still use 3677 Watts-hour. While solar radiation and occupants
could be set to zero, it is very unlikely the Outdoor Temperature is 0 degree Celsius is our office
setting in Los Angeles, California. This renders the constant meaningless or only applicable within
a limit, as in statistics, the analyst is crucial in deciding when variables should be associated. Even
79
if it is possible for Outdoor Temperature to equal zero, the point might be outside the range of data
observed. It should also be noted that this equation is only valid in the range of the observed data
because the variables might change. For example, when temperatures are cold, heating systems
might be needed, and the AHU may not work as hard as before. Since data for this all-zero range
is not collected, the value of the constant can’t be trusted and not draw solid conclusions. However,
the constant is important here for the line to float else the fitted is forced to go through the origin,
rendering the predicted line biased.
If the slope coefficient is zero, it would not be displayed here i.e. the Outdoor Relative Humidity.
Hence, the expected value of HVAC Energy Consumption would not change no matter how high,
or low Outdoor Relative Humidity gets. So a low p-value also suggests that the changes in the
predictor variables are associated with changes in the response variable. (Minitab 17 n.d.) For
order of impact, Solar Radiation has the largest impact according to the T-statistics. Outdoor
temperature is second, followed by occupants.
The same stepwise analysis is repeated for all the predictors in different end-use categories. The
following table (Table 5.3) summarizes all the stepwise results collected from Minitab. R-sq values
for p-values that are greater than the alpha level of 0.05, which indicates that it is not statistically
significant, is not represented. The same table with P-values is available in Appendix A. Six end-
use categories are used as responses in four sets of data, a) whole six weeks, b) Week 1 and 2, c)
Week 3 and 4, and d) Week 5 and 6. The columns in the table display the main predictors, which
are Outdoor temperature, Outdoor Relative Humidity, Solar Radiation and Occupancy. Their roles
change in terms of end-use category.
5.5.2 Stepwise Regression Results Summary
In the AHU Category, climate impacts energy consumption more than occupants. Solar Radiation
could explain 25.81% variation in AHU energy consumption when all days are accounted for. It
was assumed that the outdoor temperature would play a more important role as at higher outdoor
temperatures, AHU would have to work harder to condition the air. From the regression results,
AHU works in variation to Solar Radiation instead of Outdoor Temperature. This seems possible
as Solar Radiation causes heat gains through the envelope. Similar impacts can be seen when data
is broken down by weeks, compared to outdoor temperature, solar radiation seems to impact this
80
Table 5.3: Summary of Results from Stepwise Regression Analysis
Outdoor
Temperature
Outdoor
Relative
Humidity
Solar
Radiation
Occupancy
R-sq Ranki
ng
R-sq Ranki
ng
R-sq Ranki
ng
R-sq Ranki
ng
AHU
Full
8.02 2 - - 25.81 1 1.40 3
Week 1-2
28.38 1 1.26 4 10.19 3 10.72 2
Week 3 -4
3.01 2 1.33 3 19.41 1 1.14 4
Week 5 -6
8.93 2 - - 56.04 1 1.88 3
Lighting
Full
- - 8.56 1
Week 1-2
- - 24.78 1
Week 3 -4
- - 5.07 1
Week 5 -6
- - 12.18 1
Kitchen
Equipme
nt
Full
0.29 4 0.43 3 - 2 2.48 1
Week 1-2
- - 1.93 2 - - 2.17 1
Week 3 -4
- - 0.82 2 2.1 1 - -
Week 5 -6
3 1.54 2 6.22 1 1.12 4
Refrigera
tor
Full
- - 0.87 2 - - 0.51 1
Week 1-2
- - - - - - - -
Week 3 -4
- - - - - - - -
Week 5 -6
- - - - - - - -
Office
Equipme
nt
Full
10.77 1
Week 1-2
12.20 1
Week 3 -4
3.24 1
Week 5 -6
19.45 1
Compute
r
Full
15.69 1
Week 1-2
48.42 1
Week 3 -4
25.40 1
Week 5 -6
66.7 1
Total
Energy
Use
Full
49.58 1 2.81 3 7.12 2 0.55 4
Week 1-2
47.68 1 0.76 3 2.57 2 4
Week 3 -4
44.42 1 2.57 3 5.99 2 - -
Week 5 -6
2.76 3 0.27 - 57.63 1 13.12 2
81
category usage more; 19.4% and 56.04% in Week 3-4 and Week 5-6 respectively. However as
mentioned before the temperatures through Week 1 to 6 gradually decreases, for the week with
higher temperatures (Week 1 – 2) outdoor temperature impacts AHU energy use at 28.4% in the
highest order followed by occupancy at 10.72% accountability. This can be due to the introduction
of outdoor air flow and HVAC playing a significant percentage of the office energy consumption.
Instead of a fixed schedule, an ideal conditioning method would be taking occupants into account
conditioning for occupants instead of space.
For Lighting Category, both outdoor temperature and outdoor relative humidity have been
removed from the predictor set, as they are irrelevant to the change in lighting usage. This leaves
Solar Radiation and Occupancy as predictors to change in lighting end-use. Solar Radiation was
included because it would have been interesting to note if a change in solar radiation would affect
energy use in the lighting category. From the results of all week’s dataset, the R-sq value is 8.56,
meaning that overall only 8.56% of the variability in lighting is responsible by occupants. The
percentages for Week 1-2, 3-4 and 5-6 are 24.78, 5.07 and 12.18 respectively. It was first assumed
that lighting would play a more important role by occupants. This can explain how lighting is used
in an open office space. The fixtures are equipped with sensors that switch off if no motion is
detected. However, it does not detect the number of occupants. Having one occupant in the space
is the same as having 10 occupants in the space in terms of lighting usage. The usage is not
consumed by an occupant rather a minimum lighting level is required for occupants regardless the
population. Solar Radiation was not included as an important predictor by stepwise, which can
also be explained by how the fixtures operate. Since there are motion detectors rather than light
detectors, lighting usage is apparent regardless of how much daylight is available. There can be
savings potential here. Another explanation for this data is that the sub-metering includes
emergency lighting or essential corridor lighting that is required by law to be switched on at all
times, making the data less responsive towards occupants.
The R-sq values for both Kitchen Equipment are very low (<2.5%) hence the relationship by
predictors is weak (p-values for refrigeration were mostly not significant). Kitchen equipment here
refers to the coffee makers, coffee grinder, dishwasher, microwave, and toaster. Since the occupant
counter refers to the office space as a whole, and not in the kitchen specifically, the whole occupant
82
count does not accurately reflect how this space is used. There can be arguments that it is related
to the proportion of occupants in the office. Not everyone in the office would have a coffee or use
the toaster daily. A stepwise regression analysis might not be appropriate or is not the best way to
interpret how occupants affect use in this category. A better way is to understand the energy use
profile in the kitchen is using an hourly day profile. In Chapter 4, a cumulative use of kitchen
equipment category showed higher usage in the morning (8 to 10 am) and lunch (12 to 2 pm).
However it is increasingly common to have an open kitchen / pantry space in the open office
layout, it is important to include in total energy use of an office space.
For the Refrigerator category, this refers to the two refrigerators in the kitchen space. It is not
included in the kitchen equipment category as refrigerators need to be switched on 24 hours
regardless of the change in predictors. As shown, values here are rendered invalid by regression.
What would constitute change is the number of times the refrigerator is opened, as an exchange of
warmer air outside would have the refrigerator to work harder to condition the new air. Hence, this
change can be related to the cumulative usage in kitchen equipment (8 to 10 am and 12 to 2 pm on
workdays). On another note, keeping the refrigerator fully stocked allows less air volume to be
exchanged.
In the office equipment and computer end-use category, the climate is taken out of the predictor
category as it is irrelevant to the usage. The aim of these runs then becomes how many occupants
are accounted for the usage of these categories. For office equipment, the largest R-sq values for
occupants 19.45, meaning that occupants could explain 19.45 % variation in office equipment
energy use. for office equipment’s the first person in the office turns on the office equipment, and
it is used by many occupants throughout the day.
In the computer category, every occupant in this office space has similar desk size and working
equipment, i.e. two monitors and one processing unit. It was assumed that values would be high
as computer power would be accounted for when the user is on their workstation. The highest
variation was in Week 5 -6 with 66.7% accountability in computer use. However, the lowest was
15.69 from the whole period data set. This can be further explained by the method of use of
computers when occupants leave the workspace, they shut off computers or allow it to be on
83
standby. The numbers for variation should potentially be higher for a healthy energy use in this
category.
The last regression analysis was done for the total energy use of the space. This includes the six
categories above which are AHU, Lighting, Refrigerator, Office Equipment, Computer and in
addition Water Heating and Others. Overall, of the predictors, outdoor temperature causes the
highest impact in total energy use (about 49.5% accountability) followed by Solar Radiation. As
HVAC systems is a significant percentage of overall usage, the effects in the category seem to be
similar in this category. As previously mentioned, the temperatures were higher in Week 1-2
compared to Week 5-6. In terms of effect in temperature variation, higher outdoor temperature
affects higher energy use (47.68% in Week 1-2) however, when the temperature goes lower (Week
5-6) Solar Radiation becomes accountable for energy consumption (57.63%). Occupants play a
very small role as seen in depicted by the R-sq values or ordered ranking number.
It should be noted correlation explained here is not causation. For example, although it is said that
solar radiation affects energy use more than outdoor temperature when the outdoor temperature is
lower by 3 degrees Celsius, the envelope type allowing heat transfer is an important too. This
conclusion is only valid for office space or similar parameters and time period.
5.5.3 Stepwise Regression Results with Occupant Time Lag
Energy consumption can be related to the time lag of heat gain by occupants. The results in Table
5.4 of the analysis show that occupant time lag is not crucial in a change of total energy
consumption in the space. This also applies to the end-use categories. At 30 mins of time lag,
results seem to be affected negatively or no change at all. It is, therefore, sufficient to collect data
at 15 minutes intervals.
On the other hand, time lag analysis could be more appropriate on the building envelope and how
thermal lag can affect the building energy consumption. However, this is beyond the scope of
work.
84
Table 5.4: Summary of Results from Stepwise Regression Analysis with Time Lag
Outdoor
Temperature
Outdoor Relative
Humidity
Solar Radiation Occupancy
R-sq Rank R-sq Rank R-sq Rank R-sq Rank
AHU Occupancy 8.02 2 - - 25.81 1 1.4 3
15 min lag 8.02 2 - - 25.81 1 1.04 3
30 min lag 8.02 2 - - 25.81 1 0.64 3
45 min lag 8.02 2 - - 25.81 1 0.34 3
60 min lag 8.02 2 - - 25.81 1 3-
Lightin
g
Occupancy - - 8.56 1
15 min lag - - 7.41 1
30 min lag - - 6.05 1
45 min lag - - 4.66 1
60 min lag - 2- 3.38 1
Kitchen
Equipm
ent
Occupancy 0.29 4 0.43 3 0.25 2- 2.48 1
15 min lag 0.22 3- 0.30 2 0.26 4- 4.05 1
30 min lag 0.33 3 0.39 2 -- -- 6.13 1
45 min lag 0.36 3 0.45 2 - - 8.63 1
60 min lag - - 0.15 3- 0.72 2 10.96 1
Refrige
rator
Occupancy - - 0.87 2 - - 0.51 1
15 min lag - - 0.88 2 - - 0.65 1
30 min lag - - 0.87 2 - - 0.75 1
45 min lag - - 0.84 2 - - 0.78 1
60 min lag - - 0.81 2 - - 0.85 1
Office
Equipm
ent
Occupancy
10.77 1
15 min lag 9.02 1
30 min lag 5.99 1
45 min lag 3.13 1
60 min lag 1.22 1
Comput
er
Occupancy 15.69 1
15 min lag 17.22 1
30 min lag 15.97 1
45 min lag 14.12 1
60 min lag 11.73 1
Total
Energy
Use
Occupancy 49.58 1 2.81 3 7.1 2 0.55 4
15 min lag 49.58 1 2.81 3 7.12 2 0.79 4
30 min lag 49.58 1 2.81 3 7.12 2 1.04 4
45 min lag 49.58 1 2.81 3 7.12 2 1.21 4
60 min lag 49.58 1 2.81 3 7.12 2 1.35 4
85
CHAPTER 6 : CONCLUSIONS
Many approaches have been used to establishing a relationship between energy consumption and
data such as physical building characteristics and/or social, economic data. Two main types include
engineering-based and statistics. An advantage of statistical over engineering-based is the
capability of taking into account the behavior of occupants. Many past studies use information
from power suppliers and mostly referred to the residential sector or as a whole building context.
With the advancement of sub-metering and cloud storage technology, energy consumption data
can now be broken down into end-use categories in a commercial setting and occupant’s impact
can be more refined. This study aimed at using statistics, regression analysis, in particular, to
understand the impact of occupants on end-use categories and also the order of influencing factors
on energy consumption. As office spaces take up the lot of commercial spaces and energy
consumption, the findings from this study hopes to give a deeper insight on how energy is used
and the energy savings potential.
A single test bed was used in this study having the predictor variable of building characteristic and
indoor conditions controlled and therefore excluded as an influencing factor on energy
consumption. A Large amount of data that are considered to influence energy consumption were
measured for six weeks and compiled i.e. climate, indoor conditions, sub-metering, and
occupancy. Minitab tool was used for finding a correlation, provision of graphs and illustrations,
linear and stepwise regression techniques. Key takeaway points are listed below:
a) Using box plots, the average use of this office space is found to be 20.217 kW/hr.
b) From clustering occupants and aggregated use of energy, the higher number of occupants
consume more energy. However, this method is not accurate enough as it did not take into
account the outdoor temperature which was also considerably higher earlier in the study.
c) Using correlation analysis, the Pearson correlation of Total Energy Consumption and
Occupants was 0.514, which show a strong positive linear relationship.
d) Upon breaking down to a number of occupants, it was found that occupant group of 23 to 44
had a larger impact (0.253) to energy use compared to occupancy group of 0 to 22 (0.121).
e) Using correlation analysis, the Pearson correlation for Occupants and Energy by End-Use,
shown that Occupants affect Computers and Office Equipment’s more than Lighting, AHU
and least affects Kitchen Equipment and Refrigerators.
86
f) In stepwise regression, for AHU end-use category, climate impacts energy use more than
occupants. Solar Radiation could explain 25.81% variation in AHU energy consumption.
Similar impacts can be seen when data is broken down by weeks, regardless of outdoor
temperature, solar radiation seems to impact this category usage more, followed by outdoor
temperature.
g) For Lighting Category, the R-sq value for all data is 8.45, meaning that 8.45% of the variability
in energy use is accounted for by Occupants. Solar Radiation has little effect. This is obvious.
Occupant’s impact lighting within weekly data Week 1-2, Week 3-4 and Week 5-6; at 24.78%,
5.07% and 12.18% respectively. There is potential to reduce lighting energy use in this office
as lux levels measured was at the maximum required (500 lux) and remain constant throughout
the work day. The facility could take advantage of daylighting.
h) For office equipment and computer category, the largest R-sq values for occupants were
respectively, 0.1945 and 0.667, meaning that occupants could explain 19.45 % variation in
office equipment energy use and 66.7% variation in computer use.
i) Results show at higher temperatures; outdoor temperatures account for 49.5% in energy use.
Overall, of the predictors, outdoor temperature causes the highest impact in total energy use
(about 49.5% accountability) followed by Solar Radiation. As previously mentioned, the
temperatures were higher in Week 1-2 compared to Week 5-6. In terms of effect in temperature
variation, higher outdoor temperature affects higher energy use (47.68% in Week 1-2)
however, when temperature goes lower (Week 5-6) Solar Radiation becomes accountable for
energy use (57.63%). This can be due to introduction of outdoor air flow and HVAC playing
a large percentage of the office energy consumption.
j) A lag time study for occupants show weaker results or no change in relationship at 30, 45 and
60 mins lag. For a similar study collecting current data every 15 minutes is suitable.
In conclusion, while occupants do affect energy use, it is more impacted by the change in outdoor
temperatures. By installing sensors for real-time circuit-level energy measurement that gives
visibility to energy usage by real-time dashboards, raw data that contain the full effects of
occupancy can now provide insights how occupants use energy in an office space. This work
presented relevant analysis for application on the commercial sector with focus on energy use in
six categories in an office setting. There is a positive relationship between the predictor (occupants)
and the response (energy end use). In accordance, the largest positive relationship between
87
occupants is Computer, followed by Office Equipment, Lighting, AHU, Kitchen Equipment and
Refrigerator. Human factor seem to be promising component and can offer significant
improvement of energy usage. This can be used as a basis to energy simulation to reduce energy
gaps between simulated and real building performance or achieve existing high-performance
buildings.
6.1 Study Limitations
In spite of the significant data points collected, there are still some limitations outlined below that
should be further investigated.
1. There is only one test bed, to generalize this idea, more than one office space in Downtown
Los Angeles should be added and studied using the same methodology.
2. The study could be extended to collect for at minimum one-year data to verify the findings
of climate impacts
3. Incentives or education for occupants on energy consumption for further reduction. The
testbed here was used on an as is basis, if occupants were given incentives to play a role
in lowering plug loads, for example, it could provide an additional intervention variable.
4. The same methodology could be used for other commercial building types
5. Continuous collection of such data in a working office space can help to detect
malfunctions (by identifying abnormality).
6. To quantify the correlation between occupancy and energy consumption and estimate
energy savings potential.
7. Extension to a city scale by quantifying the results found using Geographic Information
Systems (GIS) that offers the opportunity to characterize building stocks.
The priority of occupants in a design stage is relatively low and the current use of static schedules
is static and does not represent well the impact of occupants due to its complexity. This thesis
provides an insight on how occupants affect energy use by end-use categories which hopes to be
implemented to achieve high performance buildings or in energy simulation models of office
spaces in commercial building instead of using one static schedule.
88
APPENDIX A: STEPWISE RESULTS WITH P-VALUES
Outdoor
Temperature
Outdoor Relative
Humidity
Solar Radiation
Occupancy
R-sq Rank R-sq Rank R-sq Rank R-sq Rank
HVAC
Full 8.0 2 - - 25.8 1 1.4 3
0.000 0.000 0.000 0.000
Week 1-2 28.38 1 1.26 4 10.19 3 10.72 2
0.000 0.002 0.000 0.000
Week 3 -4 3.01 2 1.33 3 19.4 1 1.14 4
0.000 0.004 0.000 0.007
Week 5 -6 8.93 2 - - 56.0 1 1.88 3
0.000 0.000 0.000
Lighting
Full - - 8.56 1
0.000
Week 1-2 - - 24.78 1
0.000
Week 3 -4 - - 5.07 1
0.000
Week 5 -6 - - 12.18 1
0.003 0.000
Kitchen
Equipme
nt
Full - - 0.43 3 0.25 2 2.48 1
0.014 0.054 0.000
Week 1-2 - - 1.93 2 - - 2.17 1
0.003 0.002
Week 3 -4 - - 0.82 2 2.1 1 - -
0.044 0.001
Week 5 -6 0.60 3 1.54 2 6.22 1 1.12 -
0.079 0.005 0.000 0.015
Refriger
ation
Full - - 0.87 2 - - 0.51 1
0.000 0.007
Week 1-2 - - - - - - - -
Week 3 -4 - - - - 1.24 2 0.72 1
0.113 0.063
Week 5 -6 - - - - 0.55 1 - -
0.104
Office
Equipme
nt
Full 10.77 1
0.000
Week 1-2 12.20 1
89
0.000
Week 3 -4 3.24 1
0.000
Week 5 -6 19.45 1
0.000
Comput
er
Full 15.69 -
0.000
Week 1-2 48.42 1
0.000
Week 3 -4 25.40 1
0.000
Week 5 -6 66.7 1
0.000
Others
Full 5.16 2 - - - - 17.42 1
0.000 0.000
Week 1-2 0.41 4 0.37 3 0.62 2 29.91 1
0.111 0.125 0.051 0.000
Week 3 -4 13.73 1 1.52 2 1.36 3
0.000 0.003 0.005
Week 5 -6 29.51 1 1.67 3 5.75 2
0.000 0.000 0.000
Total
Energy
Use
Full 49.58 1 2.81 3 7.12 2 0.6 4
0.000 0.000 0.000 0.000
Week 1-2 47.68 1 0.76 3 2.57 2 0.25 4
0.000 0.010 0.000 0.144
Week 3 -4 44.42 1 2.57 3 5.99 2 - -
0.000 0.000 0.000
Week 5 -6 2.76 3 0.27 - 57.63 1 13.12 2
0.000 0.028 0.000 0.000
90
APPENDIX B: STEPWISE RESULTS FROM TIME LAG WITH P-VALUES
Outdoor
Temperature
Outdoor Relative
Humidity
Solar Radiation Occupancy
R-sq Rank R-sq Rank R-sq Rank R-sq Rank
AHU Occupancy 8.02 2 - - 25.81 1 1.4 3
0.000 0.000 0.000
15 min lag 8.02 2 - - 25.81 1 1.04 3
0.000 0.000 0.000
30 min lag 8.02 2 - - 25.81 1 0.64 3
0.000 0.000 0.000
45 min lag 8.02 2 - - 25.81 1 0.34 3
0.000 0.000 0.008
60 min lag 8.02 2 - - 25.81 1 0.12 3
0.000 0.000 0.113
Lighting Occupancy - - 8.56 1
0.000
15 min lag - - 7.41 1
0.000
30 min lag - - 6.05 1
0.000
45 min lag - - 4.66 1
0.000
60 min lag - 2 3.38 1
0.000
Kitchen
Equipm
ent
Occupancy 0.29 4 0.43 3 0.25 2 2.48 1
0.041 0.014 0.054 0.000
15 min lag 0.22 3 0.30 2 0.26 4 4.05 1
0.070 0.036 0.054 0.000
30 min lag 0.33 3 0.39 2 -- -- 6.13 1
0.026 0.016 0.000
45 min lag 0.36 3 0.45 2 - - 8.63 1
0.019 0.009 0.000
60 min lag - - 0.15 3 0.72 2 10.96 1
0.121 0.001 0.000
Refriger
ation
Occupancy - - 0.87 2 - - 0.51 1
0.000 0.007
15 min lag - - 0.88 2 - - 0.65 1
0.000 0.002
30 min lag - - 0.87 2 - - 0.75 1
0.000 0.001
91
45 min lag - - 0.84 2 - - 0.78 1
0.001 0.001
60 min lag - - 0.81 2 - - 0.85 1
0.001 0.001
Office
Equipm
ent
Occupancy
10.77 1
0.000
15 min lag 9.02 1
0.000
30 min lag 5.99 1
0.000
45 min lag 3.13 1
0.000
60 min lag 1.22 1
0.000
Comput
er
Occupancy 15.69 1
0.000
15 min lag 17.22 1
0.000
30 min lag 15.97 1
0.000
45 min lag 14.12 1
0.000
60 min lag 11.73 1
0.000
Total
Energy
Use
Occupancy 49.58 1 2.81 3 7.1 2 0.55 4
0.000 0.000 0.000 0.000
15 min lag 49.58 1 2.81 3 7.12 2 0.79 4
0.000 0.000 0.000 0.000
30 min lag 49.58 1 2.81 3 7.12 2 1.04 4
0.000 0.000 0.000 0.000
45 min lag 49.58 1 2.81 3 7.12 2 1.21 4
0.000 0.000 0.000 0.000
60 min lag 49.58 1 2.81 3 7.12 2 1.35 4
0.000 0.000 0.000 0.000
92
BIBLIOGRAPHY
A, Synnefa, Santamouris M, and H. Akbari. 2007. "Estimating the effect of using cool coatings on energy
loads and thermal comfort in residential buildings in various climatic conditions." Energy and
Buildings (Energy and Buildings 39) 1167-1174.
Ahn, Ki-Uhn, and Cheol-Soo Park. 2016. "Correlation between occupants and energy consumption."
Energy and Buildings 420-433.
Aksamija, Ajla. 2013. Sustainable Facades Design Methods for High-Performance Building Envelopes.
Hoboken, N.J.: John Wiley & Sons.
Aranda, Alfonso, German Ferreira, Mainar-Toledo, Sabina Scarpellini, and Eva Sastresa. 2012. "Multiple
regression models to predict the annual energy consumption in the Spanish banking sector."
Energy and Buildings 380-387.
ASHRAE Standard 55. 2013. ANSI/ASHRAE Standard 55-2013: Thermal Environmental Conditions for
Human Occupancy. Atlanta: American Society of Heating, Refrigerating and Air Conditioning
Engineers.
ASHRAE Standard 62.1. 2013. ANSI/ASHRAE Standard 62.1-2013: Ventilation for Acceptable Indoor
Air Quality. Atlanta: ASHRAE, Heating, Refrigerating and Air-Conditioning Engineers.
ASHRAE Standard 90.1. 2013. ANSI/ASHRAE Standard 90.1-2013: Energy Standard for Buildings
except low rise residential buildings. Atlanta: ASHRAE, Heating, Refrigerating and Air-
Conditioning Engineers.
ASHRAE, Heating, Refrigerating and Air-Conditioning Engineers. 2013. ANSI/ASHRAE Standard 62.1-
2013: Ventilation for Acceptable Indoor Air Quality. Atlanta: ASHRAE.
—. 2013. ANSI/ASHRAE Standard 90.1-2013: Energy Standard for Buildings except low rise residential
buildings. Atlanta: ASHRAE.
Azar, Elie, and Carol Menassa. 2012. "Agent-Based Modeling of Occupants and Their Impact on Energy
Use in Commercial Buildings." Journal Computing in Civil Engineering 26 506-518.
Balaji, Bharathan. 2011. Using Occupancy to Reduce Energy Consumption of Buildings. Thesis, San
Diego: University of California, San Diego.
Bianco, Vincenzo, Oronzio Manca, and Sergio Nardini. 2009. "Electricity consumption forecasting in
Italy using linear regression models." Energy 1413-1421.
Bonte, Mathieu, Thellier Francoise, and Lartigue Berangere. 2014. "Impact of occupant's actions on
energy buildings performance and thermal sensation." Energy and Buildings 76 219-227.
Brandemuehl, MJ, and JE Braun. 1999. "The impact of demand-controlled and economizer ventilation
strategies on energy use in buildings." ASHRAE. ASHRAE.
93
California Energy Commision. 2013. 2012 Building Energy Efficiency Standards for Residential and Non
Residential Buildings.
Catalina, Tiberiu, Joseph Virgone, and Vlad Iordache. 2011. "Study on the Impact of the building form on
the energy consumption." International Building Performance Simulation Assoc. Sydney. 1726-
1729.
Catalina, Tiberiu, Vlad Iordache, and Bogdan Caracaleanu. 2013. "Multiple regression model for fast
prediction of the heating energy demand." Energy and Buildings 57 302-312.
Catalina, Tiberiu, Vlad Iordache, and Bogdan Caracaleanu. 2013. "Multiple regression model for fast
prediction of the heating energy demand." Energy and Buildings 57 302-312.
Chang, Wen-Kuei, and Hong Tianzhen. 2013. "Statistical Analysis and Modeling of Occupancy Patterns
in Open-plan Offices Using Measured Lighting-switch Data." Build. Simul. Building Simulation
23-32.
Chao, Yang. 2015. Energy Use Intensity Estimation Method Based on Building Facade Features by using
Regression Models . Los Angeles.
Chen, Jun, Xiaohong Wang, and Koen Steemers. 2013. "A statistical analysis of a residential energy
consumption survey study in Hangzhou, China." Energy and Buildings 193-202.
Chen, Zhenghua, and Chai Soh Yeng. 2014. "Modeling building occupancy using a novel inhomogeneous
Markov Chain approach." IEEE International Conference on Automation Science and
Engineering . Taipei: CASE.
Chung, William, and Y.V Hui. 2009. "A study of energy efficiency of private office buildings in Hong
Kong." Energy and Buildings 41 696-701.
Cleveland, Cutler J. 2006. Dictionary of Energy. Amsterdam: Elsevier.
Clevenger, C.M, and J Haymaker. 2006. "The impact of the building occupant on energy modelling
simulations." Computing and Decision Making in Civil and Building Engineering. Montreal: Int.
Society for Computing in Civil and Building Engineering.
Clevenger, Caroline M., John R. Haymaker, and Maral Jalili. 2014. "Demonstrating the Impact of the
Occupant on Building Performance." Journal of Computing in Civil Engineering 99-102.
Davis Instruments. n.d. Wireless Vantage Pro2™ Plus including UV & Solar Radiation Sensors.
Accessed Jan 21, 2016.
http://www.davisnet.com/weather/products/weather_product.asp?pnum=06162.
Energy, United States Deparment of. 2012. 2011 Buildings Energy Data. Pacific Northwest National
Laboratory.
94
EnergyStar. n.d. Energy Use Intensity. Accessed Jan 21, 2016.
http://www.energystar.gov/buildings/facility-owners-and-managers/existing-buildings/use-
portfolio-manager/understand-metrics/what-energy.
Erickson, Varick L., Lin Yiqing, Kamthe Ankur, Brahme Rohini, Surana Amit, E. Cerpa Alberto, D.
Sohn Michael, and Narayanan Satish. 2009. "Energy Efficient Building Environment Control
Strategies Using Real-time Occupancy Measurements." First ACM Workshop on Embedded
Sensing Systems for Energy Efficienct in Buildings. BuildSys.
Feng, Xiaohang, Yan Da, and Tianzhen Hong. 2014. "Simulaiton of Occupancy in Buildings." Energy
and Buildings 348-359.
Filippin, C, F. Ricard, and S. Larsen. 2013. "Evaluation of heating energy consumption patterns in the
residential building sector using stepwise selection and multivariate analysis." Energy and
Buildings 571-581.
Fumo, Nelson. 2013. "A review on the basics of building energy estimation." Renewable and Sustainable
Energy Reviews 31 53-60.
Fumo, Nelson, and Rafe Biswas. 2015. "Regression Analysis for prediction of residential energy
consumption." Renewable and Sustainable Energy Reviews (47): 332-343.
Gareth, James, Daniela Witten, Trevor Hastie, and Robert Tibshirani. 2014. An Introduction to Statistical
Learning with Applications in R. Springer.
Gronadzik, Walter, Kwok Alison, Stein Benjamin, and Reynolds John. 2010. Mechanical and Electrical
Equipment for Buildings 11th Edition. Wiley.
Haldi, Frederic, and Darren Robinson. 2011. "The impact of occupants' behaviour on building energy
demand." Building Performance Simulation 323-338.
Han, Jiawei, and Micheline Kamber. 2001. Data Mining: Concepts and Techniques. San Franciso:
Morgan Kaufmann.
Hand, David J. 2001. Principles of data mining. Mass: MIT Press.
Hong, Tianzhen. 2014. "Occupant Behavior: Impact on Energy Use of Private Offices." Lawrence
Berkeley National Laboratory LBNL-6128E.
JoonHo, Choi, Aziz Azizan, and Loftness Vivian. 2010. "Investigation on the impacts of different genders
and ages on satisfaction with thermal environments in office buildings." Building and
Environment 1529-1535.
Korolija, Ivan, Yi Zhang, Marjanovic-Halburd, and Victor Hanby. 2012. "Regression models for
predicting UK office building energy consumption from heating and cooling demands." Energy
and Buildings 214-227.
95
Lam, Joseph C, Kevin Wan, and K.I. Cheung. 2009. "An Analysis of Climatic Influences on Chiller Plant
Electrical Consumption." Applied Energy 933-940.
Lam, Joseph, Kevin Wan, K.L. Cheung, and Yang Liu. 2007. "Principal component analysis of electricity
use in office buildings." Energy and Buildings 828-836.
Li, Cheng, and Hong Tianzhen. 2014. "Revisit of Energy use and Technologies of High Performance
Buildings." ASHRAE Transactions ,
http://search.proquest.com.libproxy2.usc.edu/docview/1555408404?accountid=14749.
Linacre, Edward. 1992. Climate Data and Resources: A Reference and Guide. London: Routledge.
Loftness, Vivian, Azizan Aziz, JoonHo Choi, Kevin Kampschroer, Kevin Powell, Mike Atkinson, and
Judith Heerwagen. 2009. "The Value of Post-Occupancy Evaluation for Building Occupants and
Facility Managers." Intelligent Buildings International 249-268.
Mastrucci, Alessio, Olivier Baume, Francesca Stazi, and Ulrich Leopold. 2014. "Estimating energy
savings for the residential building stock of an entire city: A GIS-based statistical downscaling
approach applied to Rotterdam." Energy and Buildings 75 358-367.
Meir, Isaac A., Yaakov Garb, Dixin Jiao, and Alex Cicelsky. 2009. "Post-Occupancy Evaluation: An
Inevitable Step Toward Sustainability." Advances in Building Energy Research 3 189-219.
Minitab 17. n.d. Minitab User Guide on Stepwise Regression. Accessed March 10, 2016.
http://support.minitab.com/en-us/minitab/17/topic-library/modeling-statistics/regression-and-
correlation/basics/basics-of-stepwise-regression/.
Newsham, Guy, Sandra Mancini, and Benjamin Birt. 2009. "Do LEED-certified buildings save energy?
Yes, but..." Energy and Buildings 41 897-905.
Norford, L.K, R.H Socolow, E.S Hsieh, and G.V Spadaro. 1994. "Two-to-one discrepancy between
measured and predicted performance of a "low-energy' office building: insights from a
reconciliation based on the DOE-2 model." Energy and Buildings 21 121-131.
Oldewurtel, Frauke, David Sturzenegger, and Manfred Morari. 2013. "Importance of occupancy
information for building climate control." Applied Energy 521-532.
Onset Computer Corporation. n.d. Onset Products and Sensor. Accessed Jan 21, 2016.
http://www.onsetcomp.com/products/sensors/tel-7001.
—. n.d. Temperature Data Loggers & Sensors. Accessed Jan 21, 2016.
http://www.onsetcomp.com/products/data-loggers-sensors/temperature.
Ouyang, Jinlong, and Hokao Kazunori. 2009. "Energy-saving potential by Improving Occupants'
Behavior in Urban Residential Sector in Hangzhou City, China." Energy and Buildings 711-720.
Panoramic Power. n.d. Technology. Accessed Jan 21, 2016. http://www.panpwr.com/technology.
96
Parys, W, D Saelens, and H Hens. 2009. "Impact of occupant behavior on lighting energy use." Building
Simulation. Glasgow, UK. 1143-1150.
Perez-Lombard, Luis, Jose Ortiz, and Christine Pout. 2007. "A review on buildings energy consumption
information." Energy and Buildings 394-398.
Perez-Lombard, Luis, Jose Ortiz, and Christine Pout. 2008. "A review on buildings energy consumption
information." Energy and Buildings 40 394-398.
Reena, Mary K, T. Mathew Abraham, and Jacob Lilltkutty. 2015. "Energy Efficient Wireless Networked
Building Automation System Controlled by Real Occupancy." TENCON 2015- IEEE.
Ross, Barbara. 2009. Design with Energy in Mind. Thesis, Ontorio: University of Waterloo.
Ryan, Emily, and Thomas Sanquist. 2012. "Validation of building energy modeling tools under idealized
and realistic conditions." Energy and Buildings 47 375-382.
Samprit, Chatterjee, and Simonoff Jeffrey. 2013. Handbooks in Applied Statistics: Handbook of
Regression Analysis. John Wiley & Sons.
Sanchez, Marla, Carrie Webber, Richard Brown, John Busch, Margaret Pinckard, and Judy Roberson.
2007. Space Heaters, Computers, Cell Phone Chargers: How Plugged in Are Commercial
Buildings? Feb 28. http://escholarship.org/uc/item/8r101268.
Schneider Electric. n.d. Andover Continuum. Accessed Jan 21, 2016. http://www.schneider-
electric.com/en/product-range/6823-andover-continuum/.
Simona, D'Oca, and Hong Tianzhen. 2014. "Occupancy schedules learning process through a data mining
framework." Energy and Buildings 395-408.
Solutions, Statistics. n.d. Statistics Solutions. Accessed November 02, 2015.
http://www.statisticssolutions.com/assumptions-of-linear-regression.
Sonderegger, Robert. 1977. "Movers and Stayers: The Resident's Contribution to Variation across Houses
in Energy Consumption for Space Heating." Eneryg and Buildings 313-324.
Statistics Solutions. n.d. Conduct and Interpret a Multiple Linear Regression. Accessed November 02,
2015. http://www.statisticssolutions.com/multiple-linear-regression/.
Swan, Lukas, and Ismet Ugursal. 2008. "Modeling of end-use energy consumption in the residential
sector: A review of modeling techniques." Renewable and Sustainable Energy Reviews 1819-
1835.
U.S. Energy Information Administration. 2012. Commercial Buildings Energy Consumption Survey
(CBECS) Maps. Accessed Feb 26, 2016.
https://www.eia.gov/consumption/commercial/maps.cfm.
97
U.S. Energy Information Administration. 2006. Electricity Consumption and Conditional Energy
Intensity by Census Region for Non-Mall Buildings, 2003. Table, Washington DC: U.S.
Department of Energy.
Wang, Chuang, Yan Da, and Yi Jiang. 2011. "A Novel Approach for Buidling Occupancy Simulation."
Building Simulation 149-167.
Westergren, Karl-Erik, Hans Hogberg, and Urban Norlen. 1998. "Monitoring energy consumption in
single-family houses." Energy and Buildings 29 247-257.
Wu, Shaomin, and Derek Clements-Croome. 2006. "Understanding the indoor environment through
mining sensory data - A Case Study." Energy and Buildings, July 21: 1183-1191.
Yagi-Kim, Mika. 2013. "Watts per person" Paradigm to Design Net Zero Energy Buildings. Los Angeles:
University of Southern California.
Yan, Haiyan, Liu Yang, Wuxing Zheng, and Daoyi Li. 2016. "Influence of Outdoor Temperature on the
Indoor Environment and Thermal Adaptation in Chinese Residential Buildings." Energy and
Buildings 133-140.
Yang, Zhao, Wim Zeiler, Gert Boxem, and Timi Labeodan. 2015. "Virtual Occupancy Sensors for Real-
Time Occupancy Information in Buildings." Building and Environment 93 9-20.
Yang, Zheng, and Burcin Becerik-Gerber. 2014. "Coupling Occupancy Information within HVAC Energy
SImulation: A systematic review of simulation programs." Simulation Conference. Los Angeles.
Yu, Zhun, Benjamin Fung, and Haghighat Fariborz. 2013. "Extracting Knowledge from Building-Related
Data;A Data Mining Framework." Building Simulation 207-222.
Yu, Zhun, Fariborz Haghighat, Benjamin Fung, and Hiroshi Yoshino. 2010. "A Decision Tree Method for
Building Energy Demand Modeling." Energy and Buildings 1637-1646.
Yu, Zhun, Fariborz Haghighat, Benjamin Fung, and Liang Zhou. 2012. "A Novel Methodology for
Knowledge Discovery through Mining Associations between Building Operational Data." Energy
and Buildings 47 430-440.
Yu, Zhun, Fung Benjamin, Haghighat Fariboz, Yoshino Hiroshi, and Morofsky Edward. 2011. "A
Systematic Procedure to Study the Influence of Occupant Behavior on Building Energy
Consumption." Energy and Buildings 1409-1417.
Yun, Geun Young, Hyoin Kim, and Jeong Tai Kim. 2012. "Effects of Occupancy and LIghting Use
Patterns on Lighting Energy Consumption." Energy and Buildings 46 152-158.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Impacts of building performance on occupants' work productivity: a post occupancy evaluation study
PDF
Bridging performance gaps by occupancy and weather data-driven energy prediction modeling using neural networks
PDF
Developing environmental controls using a data-driven approach for enhancing environmental comfort and energy performance
PDF
Real-time simulation-based feedback on carbon impacts for user-engaged temperature management
PDF
"Watts per person" paradigm to design net zero energy buildings: examining technology interventions and integrating occupant feedback to reduce plug loads in a commercial building
PDF
Occupant-aware energy management: energy saving and comfort outcomes achievable through application of cooling setpoint adjustments
PDF
Building energy performance estimation approach: facade visual information-driven benchmark performance model
PDF
Energy use intensity estimation method based on building façade features by using regression models
PDF
Natural ventilation in tall buildings: development of design guidelines based on climate and building height
PDF
Exploration for the prediction of thermal comfort & sensation with application of building HVAC automation
PDF
Economizer performance and verification: effect of human behavior on economizer efficacy and thermal comfort in southern California
PDF
Mitigating thermal bridging in ventilated rainscreen envelope construction: Methods to reduce thermal transfer in net-zero envelope optimization
PDF
Impacts of indoor environmental quality on occupants environmental comfort: a post occupancy evaluation study
PDF
Building bridges: filling gaps between BIM and other tools in mechanical design
PDF
Facade retrofit: enhancing energy performance in existing buildings
PDF
An analysis of building component energy usage: a data driven approach to formulate a guideline
PDF
Considering occupants: comprehensive POE research on office environment of Southern California
PDF
A parametric study of the thermal performance of green roofs in different climates through energy modeling
PDF
Environmental adaptive design: building performance analysis considering change
PDF
Double skin façade in hot arid climates: computer simulations to find optimized energy and thermal performance of double skin façades
Asset Metadata
Creator
Cheong, Yun Kim
(author)
Core Title
Impact of occupants in building performance: extracting information from building data
School
School of Architecture
Degree
Master of Building Science
Degree Program
Building Science
Publication Date
04/25/2016
Defense Date
03/21/2016
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
building energy,building information,building occupants,OAI-PMH Harvest,office
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Choi, Joon-Ho (
committee chair
), Konis, Kyle (
committee member
), Schiler, Marc (
committee member
)
Creator Email
cheongyunkim@gmail.com,kcheong@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-243761
Unique identifier
UC11277533
Identifier
etd-CheongYunK-4378.pdf (filename),usctheses-c40-243761 (legacy record id)
Legacy Identifier
etd-CheongYunK-4378-1.pdf
Dmrecord
243761
Document Type
Thesis
Format
application/pdf (imt)
Rights
Cheong, Yun Kim
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
Tags
building energy
building information
building occupants