University of Southern California Dissertations and Theses

Evaluating energy consuming behaviors and the sufficiency of urban systems in the context of extreme heat hazards

(USC Thesis Other)

Evaluating energy consuming behaviors and the sufficiency of urban systems in the context of extreme heat hazards

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Evaluating Energy Consuming Behaviors and the Sufficiency of Urban
Systems in the Context of Extreme Heat Hazards
by
Andrew Shida Jin
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(CIVIL ENGINEERING)
May 2024

ii
Acknowledgments
This work was sponsored in part by the University of Southern California’s Viterbi
Fellowship, and the United States Department of Defense’s Science, Mathematics, and Research
for Transformation. The views expressed in this paper are those of the authors and do not reflect
the official policy or position of the Department of the Army, Department of Defense, or the U.S.
Government, or any other sponsor affiliations.
I’d like to express my gratitude to my advisor, Dr. Kelly T. Sanders. You constantly
inspire me with your wit and grace. You have mentored and supported me throughout these
years. Words cannot express how much I have appreciated and grown from your guidance.
A special thanks to Dr. Igor Linkov, who has served as my mentor for over seven years
and as my DoD SMART mentor for the past four years. I would also like to thank my other
defense committee members, Dr. Burçin Becerik-Gerber and Dr. Adam Rose, for their helpful
feedback and guidance throughout my PhD. Furthermore, thank you to Dr. Lucio Soibelman for
his insights during my thesis prospectus.
I would like to thank my group members and colleagues throughout the years, especially
Stepp Mayes and McKenna Peplinski who have been with me every step of the way at USC, and
to Dr. Benjamin Trump who supported me for years at USACE.
Thank you to the researchers, mentors, and professors who have pushed me to apply my
skills to research I find meaningful. There are so many on this list who have helped me here --
Ms. Sophia Capelli, Ms. Patricia Duncan, Mr. Keith Hannigan, Dr. Shura Mankin, Dr. Nora
Vazquez-Laslop, Dr. Michael Federle, Mr. Ryan Fedewa, Dr. Khalid Kadir, Dr. Deborah Sunter,
and Dr. George Ban Weiss amongst so many others.

iii
To my friends and family, thank you for inspiring me. Mom, Dad, and David, thank you
for raising me to be the researcher I am.

iv
Table of Contents
Acknowledgments........................................................................................................................... ii
List of Tables................................................................................................................................viii
List of Figures................................................................................................................................ ix
Abstract........................................................................................................................................ xiv
Chapter 1 Introduction .................................................................................................................... 1
1.1 Motivation ..............................................................................................................................1
1.2 Summary of research gaps, research questions, and the contributions of this work..............3
1.3 Structure of this document and resulting publications to date ...............................................5
Chapter 2 - Changes to U.S Municipal Heat Response Plans for Compound Extreme Heat and
COVID-19 Events............................................................................................................... 7
2.1 Introduction ............................................................................................................................7
2.2 Methodology ........................................................................................................................11
2.3 Results..................................................................................................................................12
2.3.1 Municipal Responses to COVID-19-EHE Events...................................................................... 12
2.3.2 Cooling Centers........................................................................................................................... 21
2.3.3 Alternative Cooling Options ....................................................................................................... 22
2.3.4 Financial Assistance and Individual Cooling Resources ............................................................ 23
2.4 Discussion ............................................................................................................................25
2.5 Conclusion............................................................................................................................31

v
Chapter 3 - Characterizing Contributions to the Residential Sector Load Curves from Smart
Meter Datasets .................................................................................................................. 32
3.1 Introduction ..........................................................................................................................32
3.2 Literature Review.................................................................................................................36
3.2.1 Load Profiles............................................................................................................................... 36
3.2.2 Smart Meters Data Analysis in the Residential Sector ............................................................... 38
3.3 Methods................................................................................................................................41
3.3.1 Data Description ......................................................................................................................... 41
3.3.2 Individual Load Profile Definition.............................................................................................. 41
3.3.3 Filtering of Non-Residential and Outlier Residential Electricity Customer............................... 42
3.3.4 Customer Electricity Use Decile Bins ........................................................................................ 43
3.3.5 Average Daily Load Profiles....................................................................................................... 44
3.3.6 Gini Coefficient Calculation ....................................................................................................... 45
3.4 Results..................................................................................................................................46
3.4.1 Average Load Profiles by Electricity Use Decile Bin................................................................. 48
3.4.2 Dispersion of Electricity Customers throughout the Seasons..................................................... 49
3.4.3 Climate Zone Differences in Electricity Use .............................................................................. 52
3.5 Discussion ............................................................................................................................53
3.6 Conclusion............................................................................................................................57
Chapter 4 - Characterizing Individual and Collective Peaking Behaviors in the Residential
Electricity Sector............................................................................................................... 59
4.1 Introduction ..........................................................................................................................59
4.2 Methodology ........................................................................................................................63
4.2.1 Data Description ......................................................................................................................... 63

vi
4.2.2 Daily Peaking Metrics................................................................................................................. 65
4.2.3 Aggregate Statistics..................................................................................................................... 67
4.3 Results..................................................................................................................................68
4.3.1 Summary of Collective Peaking Behaviors................................................................................ 68
4.3.2 Peak Behaviors by Individual Customers................................................................................... 71
4.4 Discussion ............................................................................................................................75
4.4.1 Direct Implications for Time of Use: .......................................................................................... 76
Chapter 5 - Conclusion ................................................................................................................. 79
5.1 Limitations ...........................................................................................................................80
5.1.1 Limitations to Cooling Center Analysis...................................................................................... 80
5.1.2 Limitations to Energy Analysis................................................................................................... 81
5.2 Policy Recommendations and Future Work.........................................................................83
5.2.1 Contributions of this work .......................................................................................................... 86
References:.................................................................................................................................... 87
Appendix..................................................................................................................................... 113
Appendix A Appendix - Supplemental Information for Chapter 2 ..........................................113
A.1 Geographical distribution of Southern California Edison dataset.....................................113
A.2 Data Filtering.....................................................................................................................113
A.3 Validation of Residential Zoning................................................................................................. 115
A.4 Breakdown of User Deciles...............................................................................................115
A.5 Summary of average load profiles by season and day-type ..............................................116
A.6 Lorenz Curve.....................................................................................................................116
A.7 Summary of average load profiles by user decile and season...........................................118

vii
A.8 Summary of Climate Zone Load Profiles by Decile .........................................................119
A.9 Usage decile breakpoint in different climate zones .................................................................... 119

viii
List of Tables
Table 2-1 - The review of the COVID-19-EH response strategies for 25 largest cities the
United States from April 1st, 2020, to October 30th, 2020 are highlighted here.
Information in this table has been derived from publicly available press releases,
news articles, and social media posts by cities. Four key excessive heat reduction
strategies – heat warnings, cooling centers, alternative cooling resources, and cost
reduction strategies – are summarized in this table. ..........................................................13
Table 4-1: Summary of Peaking Behavior Metrics.....................................................................66
Table 4-2 Summary of Peaking Metrics for Customer-days .......................................................69
Table A-1 - Summary of Filtering Process. Data in each column represents the number
of users, user-years, and user-days of data that remained after each given .....................114
Table A-2 Distribution of borders of each electricity use decile bin. Data represent
annually averaged daily consumption..............................................................................115
Table A-3 Summary statistics for the residential load profile by season and
weekday/weekend. Peak hour represents the hour of the load profile with the
highest amount of electricity used, and minimum of electricity......................................116
Table A-4 Summary statistics of load profiles by usage decile and season on weekdays.
The contribution columns show the hour where that decile is contributing most or
least to the overall curve and when those periods occur. The usage shows the hour
with the highest and lowest electricity use. .....................................................................118

ix
List of Figures
Figure 1-1 A plot of preliminary hazardous weather statistics for 2021 from the
National Weather Service’s Natural Hazard Statistic on fatalities from weather
related hazards. The data was compiled from NWS forecast offices in the 50
states, Puerto Rico, Guam, and the US Virgin Island. Excessive Heat Events are
reported whenever heat index values meet or exceed a locally/regionally
established heat warning threshold.1 Because these values are only values that
are reported by the NWS forecast office tracking themselves, they are often
significantly lower than CDC Values for temperature related moralities. .........................1
Figure 2-1 Heatmap Representation of Extreme Heat and COVID-19 in from April 1st,
2020, to October 30th, 2020 for the 25 largest US Cities by Population. Cities
are ordered by the number of total 90F or warmer days in the study period from
top (highest) to bottom (lowest). (a.) Daily Maximum Temperatures are shown
from the U.S. National Oceanic and Atmospheric Administration’s Global
Historical Climate Network – Daily (GHCN-Daily) dataset. Maximum
Temperatures are calculated as the average daily maximum temperature of all
stations in the county containing each city. New York City weather is calculated
as Queens County, NY. (b.) COVID-19 7-day Average Daily Case Rates are
shown. Data is smoothed using a cubic spline to reduce reporting noise and to
match CDC criteria for high incidence. (C.) COVID-19 7-day Average Daily
Case Rates shown for any day that had a Daily Maximum Temperature from (b)
greater than 90F to highlight concurrent high temperatures and COVID-19 cases
..........................................................................................................................................26

x
Figure 3-1 The average daily load of a residential electricity customer in the SCE
service area is highest during the hottest months (i.e., in JASO; July, August,
September and October). Peak electricity usage occurs earlier and is much higher
in hot months than other months. Daytime weekend usage is typically higher
than weekday usage. Data represent the average hourly load profiles of
approximately 160,000 SCE customers spanning the time period January 2015
through December 2019...................................................................................................47
Figure 3-2 The average load profiles of SCE customers segmented by electricity use
decile bins show that the highest consuming customers represent a
disproportionate fraction of daily load. Note: Each unique customer is assigned
to the same decile bin in all seasons based on their average daily usage across
the entire time period of study. Darker colors represent higher usage customers. ..........48
Figure 3-3 Lorenz curve of electricity use in two different hours on weekdays in July,
August , September, and October. The dispersion is lower .............................................50
Figure 3-4 The GINI Coefficient of electricity users suggests that the largest disparities
in electricity usage across customers occur in the middle of the day, on
weekdays, and in mild months. (Higher values of GINI values represent higher
dispersions in electricity usage across SCE customers.) .................................................51
Figure 3-5 Electricity load profiles differentiated by climate zone and season indicate
that the hottest regions represent a disproportionate amount of electricity usage
in SCE, particularly during the hottest months. Note: The hottest climate zones
are denoted in red, moderate in yellow, and the coolest in blue. .....................................53

xi
Figure 4-1 The top panel shows the percent of all customer weekdays with specific
peak time. The middle and bottom panels show the mean relative peak height
and mean absolute peak height for all customer weekdays by peak time. Results
are shown by season and for the full study period. Peak times are more likely to
occur earlier in the day in summer months, where we also see higher absolute,
but not relative, peaks. .....................................................................................................70
Figure 4-2. Comparison of Peak Time, Relative Peak Height, and Absolute Peak
Height between winter and summer months across climate zones (ordered from
coolest to hottest by CDDs or mean summer max temperature…). Summertime
peaks, especially in hotter climate zones, occur earlier in the day than winter
peaks and are higher in magnitude, showing that peaking behaviors can be driven
by temperature. ................................................................................................................71
Figure 4-3 Boxplot of Weekday Customer Mean (Left) and Customer Standard
Deviations (Middle) of Relative Peak Height (Bottom) and Absolute Peak
Height (Top). Outliers were excluded using the 1.5 IQR rule. The right most
panel shows the magnitude of change from DJF to JJA in relative peak heights
and absolute peak heights, in increasing order, for all customers in our dataset.
While the customer-day analysis showed higher magnitude summer peaks,
analysis of individual customers reveals that some customers have lower
magnitude peaks in the summer.......................................................................................72
Figure 4-4 Comparison of Mean Peak Time, Peak Time Variance, Mode Peak Time,
and Mode Frequency across seasons for all customers. Peaking behaviors are
less variable when examined within a customer than in the customer-day analysis

xii
that revealed large between-customer variation. However, internal variation in
peaking behaviors is still significant................................................................................74
Figure 4-5 Boxplots of Customer Mean Absolute Peak Height, Customer Mean
Relative Peak Height, and Customer Peak Variance for distinct census tract
income levels (Top) and quintiles of user electricity consumption (Bottom).
Customers in higher-income census tracts and those with higher levels of
average load have similar peak magnitudes in winter, but much larger magnitude
peaks in summer months..................................................................................................75
Figure 4-6 The proportion of customers by best rate structure (top) for customers
divided by electricity usage quintile (left) and neighborhood income (right)
assuming behaviors from 2015, 2016, and 2018, shows time of use rates are
better rates for the highest users of electricity. The pricing differences of these
rates (bottom), on average are within one cent of one another. .......................................77
Figure A-1 Plot of customers included in the SCE dataset. Users are counted by 1 km
by 1 km grid. Grid cells with less than 5 households are suppressed in our
analysis........................................................................................................................... 113
Figure A-2 Weekday load profiles by season and year. NB: Winter NDJF includes the
first two and last two months of the year....................................................................... 114
Figure A-3 Lorenz curve of electricity use for the daily load (left) and system peak
4pm to 9pm period (right). The solid dark-blue line represents the Lorenz-curve.
The dashed line represents a line of perfect equality, where each user uses the
same amount of electricity. The GINI index is the proportion of the area under

xiii
the line of perfect equality that is contained by the area between the line of
perfect equality and the Lorenz curve............................................................................ 117
Figure A-4 – Breakpoints of user decile bins by climate zones, where each red-dot
represents the breakpoint in electricity usage by the average daily electricity of
users. For example, an electricity customer with an annual-daily average
electricity consumption of 30 kWh per day in Climate Zone 15 would be
classified as CZ15-Decile 7 while a customer with the same annual-daily
average electricity consumption in Climate Zone 8 would be classified as CZ8-
Decile 10. ....................................................................................................................... 119

xiv
Abstract
As climate-induced extreme events, such as extreme heat, continue to grow in frequency,
duration, and intensity, managing these challenges is becoming increasingly difficult. First, this
dissertation analyzes the challenges of managing compound climate-extreme events, by
exploring municipal responses to compound global pandemic and extreme heat events during the
2020 COVID-19 pandemic. This body of work then explores electricity consumption behaviors
in Southern California by analyzing a dataset of smart meter electricity records for
approximately 200,000 homes in the greater Los Angeles area over a five-year span. This
dissertation develops frameworks to construct of data-driven hourly load profiles of the
residential electricity sector and analyzes individual peaking behaviors of residential customers.
The results of this body of work can be used by grid operators, utilities, and policymakers, to
better design electrification policies, rate structures, and heat action plans to protect communities
from the impacts of climate change while preparing for future energy needs.

1
Chapter 1 Introduction
1.1 Motivation
Figure 1-1 A plot of preliminary hazardous weather statistics for 2021 from the National Weather Service’s Natural
Hazard Statistic on fatalities from weather related hazards. The data was compiled from NWS forecast offices in the 50 states,
Puerto Rico, Guam, and the US Virgin Island. Excessive Heat Events are reported whenever heat index values meet or exceed a
locally/regionally established heat warning threshold.1 Because these values are only values that are reported by the NWS
forecast office tracking themselves, they are often significantly lower than CDC Values for temperature related moralities.
According to the most recent report from the Intergovernmental Panel on Climate Change
(IPCC), greenhouse gases from human activities are responsible for approximately 1.09°C of
warming above 1850-1900 temperatures, and global mean temperatures are more likely than not
over the next to reach or exceed 1.5°C of warming by mid-century.2 Climate change is
dramatically changing regional climates throughout the world. As current temperature trends
towards an increase of the global mean temperature of 1.5 °C or more above preindustrial levels,
more people will be exposed to high temperatures.
3 Alongside the temperature increases, climate
change is also increasing the frequency and intensity of extreme events, such as drought, tropical
cyclones, flooding.2

2
Climate Change-induced extreme events are impacting increasingly complex, optimized,
and interconnected urban infrastructure.4 This increased complexity, when paired with the
uncertainties of a changing climate is increasing the difficulty of traditional risk assessment, a
process in which failure probabilities of systems are assessed and associated consequences
evaluated.5 As climate change continues to increase the likelihood of extreme weather events
throughout the world, weather threats have the potential to be undervalued by traditional risk
assessment, especially in low-probability, high-impact events.
One such hazard is the threat of extreme heat events (EHEs), which are expected to
increase in frequency, magnitude, and duration.6 Extreme heat is a significant public health issue.
In the United States, extreme heat is the deadliest weather-related event. Exposure to heat
attributed by the US Center for Disease Control as killing 937 persons in 2021.7 Furthermore,
these values are limited by the reporting difficult of extreme heat. While the common morbidities
of other forms of extreme weather, such as tornados or hurricanes, are associated with trauma,
the impacts of extreme temperatures are more difficult to directly attribute and lead to underreporting of EHE-related mortalities.
8–10 The challenge of managing EHEs was even more
significant during the 2020 global COVID-19 pandemic. One of the main challenges was the
need to balance the implementation of measures to reduce the spread of the virus, such as social
distancing and masking, while trying to protect vulnerable populations from extreme heat. Thus,
this work investigates how municipal heat adaptation plans changed during the COVID-19
pandemic in response to two co-occurring public health crises.
Climate change-induced high temperatures can have several impacts on the power
system. Modern electric grids are expected to meet the needs of urban systems whenever
electricity is demanded. EHEs exacerbate the temporal mismatch between the intra-day patterns

3
of VRE generation and electricity demand. EHEs increase the demand for electricity, especially
through increased demand for air conditioning and other cooling resources. This increased
demand can put strain on the electricity grid and require utilities to increase generation to meet
demand. The largest sector of end-use utilities in the United States is the residential sector and
comprises 38.9% of all electricity used.11 At an individual household level, reliable electricity is
a key climate adaptation tool, as it enables the use of life-saving equipment and systems, such as
refrigeration for medicines and food, and lighting and air-conditioning in homes. Thus, this
work explores to build our understanding of the residential electricity sector using highresolution smart meter datasets to elucidate our understanding of the overall load profile and key
peaking behaviors.
1.2 Summary of research gaps, research questions, and the
contributions of this work
U.S. heat response plans tend to be created and implemented by local governments and
therefore vary regionally with each municipality creating heat response plans to meet local needs
and allocate available resources to protect public health during extreme heat events. This body of
work addresses a key understanding of extreme heat management strategies in the face of
compound threats. Heat adaptation plans made by cities have focused on only the threat of
extreme heat, and often do not have contingencies for co-occurring crises. For this reason, I
performed retrospective analysis of municipal policies and actions for the twenty-five largest
cities in the United States during the summer of 2020 to understand the differences among
strategies to protect the public from high temperatures while also maintaining COVID-19 related
health precautions. This work asks the following research questions and is presented in Chapter 2
of this dissertation.

4
1. Did cities experience heat advisories or excessive heat warnings during the COVID-19
pandemic between April 1st and October 30th, 2020?
2. What resources were available for communities to access during extreme heat events and
how did their availability change as a a result of the COVID-19 pandemic?
Chapter 3 and 4 develop insights into the impact of seasons on the residential electricity
usage of individual households. I utilize a large smart meter dataset consisting of more than
150,000 homes to analysis two main characteristics of the residential electricity load. First, I
investigate the residential intra-day load profile and how residential electricity consumption
changes throughout the day. While the overall residential load profile has been well investigated
through physics-based models or aggregate load analyses, our understanding of which customers
contribute most to the overall load profile, especially during hours of high electricity grid stress,
remains limited. Chapter 3 asks the following research questions.
3. What are the diurnal and seasonal characteristics in the aggregate residential load curve
in Southern California Edison’s service area?
4. How do customers within different electricity use decile bins contribute to the overall
load curve throughout seasons and do higher consumption deciles have an outsized
contribution on residential electricity load during specific hours of the day?
5. How do customers in different climate zones contribute to the overall load curve (i.e., in
terms of shape and magnitude)?
Chapter 4 characterizes peaking behaviors in the residential electricity sector. While there
has been extensive research into the behaviors of customers, most studies either rely on survey
data or on clustering customers together by similar behaviors. To address this gap, I develop
various metrics to calculate key features of electricity peaks within 200 million customer-day

5
records and characterize the timing and magnitude of peak behaviors (i.e., the time of day that
customers use the most electricity). Furthermore, the understanding of customers’ consistency in
those peak behaviors is key to our understanding of how to change behaviors to ensure stability
of the electricity grid during peak hours. Thus, I measure customers’ consistency individually,
and the consistency of the overall distribution of customers. In summation, chapter 4 asks the
following research questions.
6. What is the distribution of peaking behaviors in terms of magnitude and timing observed
in Southern California? How does seasonality and geography impact these differences?
7. What are the observable trends in timing and magnitude of peak behaviors of individuals
in the residential sector?
8. How much do the behaviors of individuals in the residential sector vary in time and
magnitude between days? What is the impact of seasonality on this difference?
1.3 Structure of this document and resulting publications to date
This document is organized into 5 chapters, with Chapters 2, 3, and 4 answering research
questions 1 and 2, 3 through 5, and 6 through 8 respectively. Chapter 5 then summarizes and
concludes this body of work.
Chapter 2 has been published in the peer-reviewed article below:
• Chapter 2: Jin, Andrew Shida, and Kelly T. Sanders. "Analyzing changes to US
municipal heat response plans during the COVID-19 pandemic." Environmental Science
& Policy 128 (2022): 347-358.
Chapter 3 has been submitted, peer-reviewed, and is currently under revision in the
article below:

6
• Chapter 3: Jin, Andrew Shida, and Kelly T. Sanders. “Characterizing Contributions to the
Residential Sector Load Curves from Smart Meter Datasets” Applied Energy (In Review)

7
Chapter 2 - Changes to U.S Municipal Heat Response
Plans for Compound Extreme Heat and COVID-19
Events
The content included in this Chapter is published in:
Jin, Andrew Shida, and Kelly T. Sanders. "Analyzing changes to US municipal heat
response plans during the COVID-19 pandemic." Environmental Science & Policy 128 (2022):
347-358.
2.1 Introduction
Extreme heat events (EHEs) are currently the deadliest weather related events in the
United States12, and they are expected to grow in intensity, frequency, and duration due to a
warming climate.13 In the period spanning 2004 to 2018, an average of 701 people died annually
in the U.S. due to heat related illness.14 The health impacts of EHE’s are being amplified by
factors such as the intensification of the urban heat island effect and increased vulnerability due
to aging populations.15
In response to the growing threat of EHEs, many communities have developed heat
response plans that identify risk factors, high-risk populations, and methods to protect their most
vulnerable residents that might not otherwise have protection from heat exposure.3 In the United
States, heat response plans are coordinated plans to organize activities to prevent heat-related
morbidity and mortality, often in response to National Weather Service advisories (e.g. heat
watches, heat warnings, or heat advisories).16 U.S. heat response plans tend to be created and
implemented by local governments and therefore vary regionally with each municipality creating
heat response plans to meet local needs and available resources.
Local government heat responses typically include measures to (a) define factors that
magnify the health-based consequences of EHE exposure, (b) identify and inform high-risk

8
populations of their options to mitigate EHE threats at a household level, and (c) provide
facilities and resources for underserved and at-risk communities (e.g., cooling centers).3 These
plans can range from cursory written measures to comprehensive heat response plans that
involve many city departments and partnerships with nongovernmental organizations in efforts to
implement strategies to target at-risk individuals and provide resources to cooling resources.17
They typically include measures to reduce heat-related health impacts by implementing public
outreach and heat adaptation strategies such as working with local media to broadcast emergency
warnings, opening cooling centers, distributing instructions on how to identify heat illness,
extending open hours at beaches and public pools, and implementing outreach programs to check
on the heat vulnerable.17–19 Local contexts, such as political will, infrastructural and financial
resources, demographics, and experiences with past heat waves, impact the types of interventions
considered and the ability to implement those interventions.20
Even when heat response plans are comprehensive, individual perceptions of extreme
heat can prevent heat interventions from being effective. While multiple survey studies
conducted throughout United States cities have found that a majority of respondents are aware of
heat events, many respondents in these studies claim that they chose not to modify their
behaviors as a result of financial insecurity (i.e., residents could not afford to use air conditioners
due to electricity costs) or because respondents did not consider heat to pose a significant danger
to their health.21–25
One major risk factor for heat-illness and heat-related mortality is access to air
conditioning.26 Cities have historically opened public cooling centers in an attempt to reduce
heat-related morbidity and mortality by providing cool public spaces in air-conditioned buildings
and are designed to provide respite and safety during extreme heat.27While cooling centers have

9
been widely used throughout U.S. cities, their efficacy to reduce heat-related mortality is
undetermined.28 Cooling centers are only effective if the heat-vulnerable can access them.
Multiple surveys of heat vulnerable households have found that respondents do not know cooling
resources exist or felt that resources were unavailable to them (e.g. believing they were only for
the elderly).22,25,28,29 Residents and community leaders in multiple survey studies also reported
that transportation access and lack of walkability to cooling centers are key barriers to utilization
due to the diverse and unique needs of the heat-vulnerable, such as elderly persons with limited
mobility.21
While the effectiveness of official cooling centers is not well understood, there is
significant evidence to suggest that accessibility to cooling resources outside of the residence is
effective in reducing the health impact of EHEs. Physical improvements that support safe,
walkable streets provide significant protection to the heat vulnerable by allowing residents to
access informal cooling resources like parks, beaches, pools, retail stores, and libraries.18 Fraser
et al. found that official cooling centers in Los Angeles County, California, and Maricopa
County, Arizona were accessible to less than 5% of the population, while free or low-cost
cooling resources like shopping malls, libraries and other de-facto cooling centers were within
walking distance to much higher numbers of households (80% in Los Angeles, 39% in
Maricopa).30
The COVID-19 pandemic made opening public spaces for congregation more difficult in
the summer of 2020 due to the nature of disease transmission. As a result, creating robust plans
to protect the public against the public health threats of EHEs was much more difficult because
of key epidemic safety protocols, namely social distancing, capacity restrictions, and/or the
closure of many public spaces altogether. City managers had to weigh the reopening of many

10
cool-spaces like pools, beaches, and businesses with COVID-19 concerns31, as emergency
response agencies, at both federal and local levels, were tasked with coordinating responses to
both COVID-19 and climate disasters.32
Furthermore, the economic impact of COVID-19 related policies (e.g. job loss, furloughs,
increased costs for utilities, reduced access to public transportation etc.) tended to exacerbate the
long-standing socioeconomic and racial disparities that increase risks of poor health outcomes in
specific populations. In addition to raising risks to severe disease, these economic hardships
reduced the ability of vulnerable populations to afford interventions that provide protection from
extreme heat (e.g., air conditioning, paying utility bills, transportation to cooling centers, etc.).
More broadly speaking, protecting the public from the threats of acute and recurring
weather-related events such as heat waves, hurricanes, tornados, and wildfires has become more
expensive and more challenging over time.33,34 While these seasonal phenomena have already
stretched limited governmental resources at the federal, state and local levels in recent years,
responding to these natural disasters while simultaneously managing a contagious disease during
the COVID-19 pandemic added a great deal of complexity to developing adequate governmental
responses. Given that most experts expect that contagious epidemics are likely to reoccur in the
future,35 analyzing the successes and failures of governments to respond to concurrent disasters
during the summer of 2020 is prudent.
In this manuscript we investigate how compound extreme heat-pandemic hazards were
managed in the 25 largest cities in the United States (by population) during the period spanning
from April through September 2020. First, we compiled a database detailing the municipal
response of these cities to compound EHE-pandemic hazards in Summer 2020, then we discuss

11
these municipal responses across a range of heat response classifications. Finally, we comment
on lessons learned and potential interventions to improve future responses to compound events.
2.2 Methodology
Our study analyzed the heat response of the 25 largest U.S. cities by population size
through the period spanning from April 1st to September 30th, 2020.36 After an initial search of
response policies, we identified four key categories of COVID-19-EHE adaptation strategies,
which include heat warnings, cooling centers, alternative cooling resources, and utility assistance
and shutoffs. For each category of cooling strategy, we tracked whether a strategy was initiated,
whether there were specific modifications to those strategies for COVID-19 related concerns,
and whether a given strategy was reduced in magnitude in comparison to pre-pandemic
strategies.
For each city, we performed keyword searches on Google News for each month included
in our study period. We also searched each city’s website, including public health and emergency
departments, as well as local news outlets for heat related notifications. For heat warning
systems, we looked at both news posts as well as the local National Weather Service’s Twitter
account.37 We did not track cooling centers or alternative cooling facilities provided by nongovernmental organizations unless they were working directly in conjunction with city or county
governments to open centers.
This methodology assumes that information on a heat plan would be released and made
publicly available via online documents, including government released sites and news-based
sites. One key limitation of our search is that non-public sources and plans developed and
conceived offline may not appear. Thus, the findings of our search more closely reflect the
experience of residents and their access to cooling resources than a comprehensive list of all

12
measures taken during COVID-19. As a result, the findings of this search are likely to present
more results for cities which experience more extreme events in our study period. Cities with few
or no EHEs during our study period may never publicize heat response plans, even if they were
developed.
2.3 Results
2.3.1 Municipal Responses to COVID-19-EHE Events
Here we discuss the results of our review of heat response plans, which are summarized
in Table 2-1.
Heat-health warning systems and heat-health action plans are key components of reduce
the morbidity and mortality of extreme heat on the populations by notifying the public of
inclement extreme heat and initiating emergency public health interventions.38,39 Table 2-1
demonstrates that of the cities in our study group, only Seattle did not have a National Weather
Service heat notice (i.e., met local qualifications to issue a heat advisory or excessive heat
warning).
Heat plans varied throughout the United States. Most cities within our study did not post
a formal comprehensive heat plan to the public. Only three cities, New York, Chicago, and
Houston posted formal heat plans to the public that were adapted to COVID-19 restrictions. Two
others, Los Angeles and Philadelphia, reported COVID-19 heat-plans were in progress at the end
of the summer of 2020. While not all cities developed publicly outlined heat plans, a majority of
cities initiated heat-related adaptation strategies.

Table 2-1 - The review of the COVID-19-EH response strategies for 25 largest cities the United States from April 1st, 2020, to October 30th, 2020, are highlighted here.
13
Information in this table has been derived from publicly available press releases, news articles, and social media posts by cities. Four key excessive heat reduction strategies – heat
warnings, cooling centers, alternative cooling resources, and cost reduction strategies – are summarized in this table.
DATAA AS OF APRIL 15,
2021
COOLING CENTERS ALTERNATIVE
COOLING
COST REDUCTION COVID-19
HEATLH
CITY
Heat Advisory or Excessive Heat
Warning (+)
Heat Plan
Opened Cooling Centers?
Capacity/
Time Limits
Cooling Center Sites
Sites in 2019 and 2020
Social Distancing and Facial
Coverings
Spraying/Misting Stations/Splash
Pads
Pools Open
Beaches Open
Free AC/Fans
COVID-19 Specific Utility
Assistance
Moratorium on Shutoffs
Closed Testing Cites
NEW YORK CITY
Jul. 8-940,
Jul. 19-2041,
Jul. 26-2742,
Aug. 10-1243
Adapted to COVID-1944
Yes45
Yes45
2020: 145-200+45
2019: 50046
Yes45
Yes45
Limited45
Limited47
74,000 Air Conditioners48
Up to $140 per Family ($70
Million Total) 48
Statewide-PUC Ban 49

14
LOS ANGELES
Apr
.24
-2550
,
May 6
-
751
,
June 852
,
Jul 1153
,
Sep
.
6
-
754+,
Oct
.
1
-
255
In Progress56
Yes57
Yes58
• Parks • Libraries • Community • Senior Centers59
2020: 1959
2019:100+60
Yes61
Limited62
Closed62
Limited62
Up to $500 per Family ($50 Million
Total) in November 63
Mayoral Ban64
Drive- Through Testing Closed65
CHICAGO
Jul
. 17
-1866
Adapted to COVID-1967
Yes67
Yes68
• Senior Centers • Police Stations67 • 50 CTA Cooling Buses • various Chicago Public
Schools 67
2019: 6+
2020: 6+
Yes68
Yes68
Closed69
Closed69
Up to $500 through LIHEAP
70
Up to $300 grant in
November70
Utility Decision
HOUSTON
Jun
. 9 71
,
Jul
. 1172
,
Aug
. 2873
Adapted to COVID-1974
Yes75
• Salvation Army
2019: 1176
2020: 475
Yes75
Yes 77
Yes 77
220 Air Conditioners78
Statewide-PUC Ban79
Testing Schedule Moved Earlier in
Day80
Test Sites Closed Early81

15
PHOENIX
Apr
. 26
-30,
May 6
-7,
May 27
-31,
Jun.
2
-4,
Jul
.10
-13,
Jul 19,
Jul 29
-Aug 4,
Aug
.
9
-10,
Aug
. 12
-20,
Aug. 24
-28,
Sep.
4
-7,
Sep
. 1782
Yes
Yes
• Convention Center, • Climate-controlled tents.83 • Salvation Army
2019: 12 84
2020: 12 + Conv. Ctr 85
Yes 83
Closed86
Statewide-PUC Ban87
PHILADELPHIA
Jul 19
-2088
In Progress89
Yes90
Yes90
• Libraries • Schools90 • SEPTA Cooling Buses90
2019: 3891
2020: 5 +, 5 Buses92
Yes90
Yes90
Closed93
Up to $800 from LIHEAP
Recovery Crisis Program.94
Statewide Ban95

16
SAN ANTONIO
May
.
1
-3 96
,
Jun
.
997
,
Jul
. 1398+,
Aug
. 13
-1499,
Aug. 29100,
Sep
.
1101
Activated102
Yes103
• Libraries • Senior centers • Community centers 102
2019: 48104
2020: 28102
Yes103
Closed105
Closed105
5000 Box Fans106
Statewide Ban79
SAN DIEGO
Apr. 24
-25107
,
May 6108
,
Jun 9109
,
Jul 30
- Aug
1110
,
Aug 15
-17111
+
,
Sep
.
4
-
6112
+
In Progress113
Yes114,115
(started May)114,115
Yes116
• Libraries, • Community Centers 117
2019: 100+ 118
2020: 9 117
Yes116
Limited119
Limited120,119
Free Electric Fans for Low
Income Persons116
Statewide Ban121
DALLAS
Jun
. 29
-Jul
. 2 122
,
Jul
. 10
-12123
,
Aug 14
-15124+,
Aug 28125
+
Limited126, Salvation
Army Only127
Yes127
• Salvation Army • Community Centers127
2020: 13127
2019: 6+16
Recreation
Centers+29
Libraries128 Yes127
Yes129
Closed130
Statewide Ban79
SAN JOSE
May
. 25
- 28131
,
Jun
.
2132
,
Aug
. 14
-16133
Yes134
Yes134
• Community Centers134
2020: 5 134
2019: 5 135
Yes 134
Closed 136
Up to $100 in bill
payment assistance137
Statewide Ban121
AUSTIN
May 30138
,
Jul 11
-12139
,
Jul 13140+,
Aug
. 30 141
Limited142
$46 million
in utility bill
relief
Statewide
Ban79

17
JACKSONVILLE
Sep
.
4143
Limited144
Yes145
up to 12 months of rent and utility
payments in March 2021
($29M Total) .146
Utility Decision - Limited (Ended
Jul 7)147
FORT WORTH
Jun 29
-Jul
2122
,
Jul
. 10
-12123
,
A
ug. 14
-15124+,
Aug 28125
+
Limited City148 &
Salvation Army Only
127
Yes 148
• Community Centers148
2020: 2148
Yes148
Closed 149
Statewide Ban79
COLUMBUS
Jul
8
-
9150
,
Jul 19151
No152
Closed152
Closed152
2000 Fans152
Statewide
Ban(EPA) 153
(September 1)
CHARLOTTE
Jul. 19154
Aug. 1155
City & Salvation
Army 156
Yes 156
• Community Centers127
2020: 1156
2019: 4 157
Yes 156
Fans Distributed 156

18
SAN FRANCISCO
Aug. 14
-16158, Sep.
5
-
7159,
Oct. 1160
Yes161
Yes162
• Community Center, • Libraries161
2020: 4161
2017: 11163
Yes162
Closed164
Limited165
Statewide Ban 121
INDIANAPOLIS
Jul 19166, Aug 10123
Limited 167
Limited168
Statewide Ban
Limited (Aug 14) 169
SEATTLE
Jul 27170
No171
Closed172
Closed 172
Yes173
Statewide
Ban 174
DENVER
Closed.175
Statewide Ban
Limited ( June 12)176
WASHINGTON
Jul 20177
Activated178
Yes 178
Yes178
• Shelters
,
• Schools, • Recreation centers 179
2020: 16179
Closed180
Closed180
$250-600
LIHEAP181
Yes182
Testing Closed
due to Extreme
heat180

19
BOSTON
Jul 19-20183,
Jul 26 – Jul 28184 ,
August 9-12185
Yes.186
Yes187
2020: 20186
2019: 35188
Yes187
Open186
Limited187
Yes189
Statewide190
EL PASO
Jun 4-6191,
Jun. 17192,
Jul. 10-12193 + ,
Jul. 30-31194,
Aug. 11-16195
Aug. 20-21 196
Yes197
• Recreation centers,
• Senior Centers,
• Pool Facility
2020: 9197
Yes197
Closed198
Closed198
Electric
Fans. 199
Statewide ban
79
NASHVILLE
Jul 19 200
Closed201
Limited202
Statewide Ban
(August 29).203
DETROIT
Jul 7191,
Jul 19192
Yes204
Yes204
2020: 5204
2019: 20205
Yes204
Closed206
Yes.207
Utility-led
(Jul 29)208
OKLAHOMA CITY
Jul 10,
Jul 18
Yes209
Utility Led 210
(September 6)
A This table is not meant to be comprehensive. Rather it aims to identify the diverse strategies and portfolio of policies different cities employed to adapt to COVID-19

20
20
Though best practices for joint COVID-19-EHE preparedness remain elusive, some
jurisdictions adapted existing strategies or resourced new initiatives to address EHE throughout
the summer of 2020. New York City (NYC) took an aggressive stance to mitigate heat risk
during COVID-19 with a prevention focused policy. The city created a $55 million initiative to
provide 74,000 air conditioning units to vulnerable populations and identified existing facilities,
cooling elements, and other resources that could be used as key cooling centers in high-risk
communities, planned appropriate social distancing and providing personal protective equipment
(PPE).44 In addition to this air conditioning program, the New York State Public Service
commission approved an additional $70.56 million emergency cooling bill relief program for
four months that made electric customers eligible to receive up to $40 a month in electricity bill
relief.211
Not all cities were able to take such an aggressive heat response measures as NYC during
the summer of 2020, typically due to limited financial resources. For example, in Philadelphia,
pandemic-related budget cuts led to a 20% reduction in its parks and recreation budgets forcing
74 public pools and 152 recreational centers to close for the summer 2020 season. To alleviate
the lost cooling facilities, the city department raised $600,000 in six weeks for its Playstreets
program, which acted as an extension of federal free lunch programs to provide 100 streets
throughout Philadelphia with free tents, patio umbrellas, misting fans, super soakers, water jugs,
water balloons, and neck cooling rags.212

21
21
2.3.2 Cooling Centers
Cooling center response strategies were largely characterized in this study according to
reductions in the number of locations available, capacity of centers, and social distancing
requirements. Every city that opened cooling centers in our study had some form of social
distancing or mask mandate required for entrance into cooling centers. Many cities traditionally
utilized libraries or recreation centers as cooling centers but closed them during summer 2020
due to COVID-19 restrictions. Furthermore, cities such as Columbus, Dallas, and San Jose cut
library and parks department budgets to meet COVID-19 budgetary shortfalls leading to reduced
operations and staff furloughs.213–215 The number of cooling center sites in cities such as New
York City, Los Angeles, and Houston were halved in comparison to years before. Additionally,
public spaces that might otherwise provide de-facto heat relief, such as malls and museums, were
also closed in many cities.
Some cities preemptively opened cooling centers before heat watches or warnings were
issued, including San Antonio and Chicago. In Phoenix and El Paso, cooling centers were
opened daily for those looking for heat relief throughout the summer, even when heat
emergencies were not declared. Non-traditional facilities were used as cooling centers over the
summer of 2020. For example, Philadelphia and Chicago both implemented “cooling bus”
programs to leverage their public transportation system’s air-conditioned buses as cooling centers
that could strategically be placed in locations that needed cooling access.67,90 Phoenix, on the
other hand, used their convention center as a larger cooling center with more space availability to
physically distance visitors.85
However, even when open, attendance in cooling centers was another area of concern for
many municipalities. A number of cities reported cooling center underutilization, including

22
22
Washington DC, Detroit, and El Paso.216–218 During a four-day heatwave spanning Labor Day
(Sept 3-6, 2020), slots in the Los Angeles’s six cooling centers averaged under one-third
utilization, with many community outreach volunteers citing poor walkability or lack of
transportation options to cooling centers as key reasons for underutilization by vulnerable
populations.219 Previous work found that even in non-pandemic conditions, less than 10% of
houses in Los Angeles and Phoenix were within walking distance of an official cooling center.30
Reduced openings because of COVID-19 restrictions significantly reduced geographic
accessibility to cooling centers, exacerbating these accessibility challenges.
Cooling center attendance could also have decreased because of reduced programming at
cooling centers that would have traditionally helped to attract visitors. For example, although
cooling centers were often still opened at libraries, recreation centers, senior centers, or park
buildings during summer 2020, many ceased programming and standard services (e.g., book
browsing at libraries). Significant restrictions of allowed usage for designated cooling centers,
such as one-hour per person per day limits in Fort Worth148, or opening hours limited to before
afternoon heat such as in Dallas 126, could have also dissuaded people from using cooling
facilities offered.
2.3.3 Alternative Cooling Options
Pools and beaches have traditionally been a key aide for citizens to beat the heat in the
past and reported heavy usage on hot days, even despite concerns about COVID-19.220,221
However, COVID-19 related public health restrictions drove many cities to close these
alternative cooling facilities. The same budgetary cuts that reduced cooling center availability
also impacted the ability for cities to open alternative cooling facilities. For example, the Seattle
Parks and Recreation department prioritized utilizing staff resources to provide outdoor

23
23
lifeguarded beaches rather than their pools.173 Financial and human resource limits also slowed
reopening, as in Jacksonville, Florida, where the gradual reopening of city pools was largely
driven by staffing shortages for pools.144 While some city pools were open in the summer of
2020, many cities used reservation and lap pool systems. For example, pools in Austin, Texas
normally charge a fee, but during the summer of 2020 they were available for free, yet required
reservations to enter.142 These types of policies, enforced to reduce virus transmission, may have
restricted vulnerable populations from accessing alternative cooling facilities, particularly those
that did not have internet access.
Spraying features, such as showers and sprays at parks, interactive fountains, or spray
grounds, were another major component of cooling plans in many cities. The relatively low
public health risks of opening splashpads, despite reduced municipal staffing and budgetary
constraints, likely drove the decision for 5 cities (New York City, Chicago, Philadelphia, Dallas,
Boston) in Table 2-1 to reopen spraygrounds despite some form of limitation on public pools.
Spraying resources were a more dispatchable service in comparison to pools. Chicago reopened
splashpads specifically during the heat wave to protect health.67 New York City’s Department of
Environmental Protection and Fire Department New York worked to proactively install 320 fire
hydrant spray caps in zones found to have high heat vulnerability with the goal of maintaining
less than a quarter mile distance between spray features or cooling resources.44
2.3.4 Financial Assistance and Individual Cooling Resources
The economic constraints and restrictions caused by COVID-19 intensified household
energy insecurity, and caused people that were already household energy insecure to become
more susceptible to both heat and COVID-19 (e.g. poor ventilation leading to both warmer
buildings and increased COVID-19 exposure).222 Energy insecurity intensified because of the

24
24
COVID-19 pandemic due to the direct impact of household lost wages due to COVID-19
positive members, unemployment following COVID-19 related public health closures and/or
fears of contracting the virus, and other factors placing financial hardships on families (e.g. lack
of childcare).223 These conditions made it even harder for the energy insecure to install new air
conditioning or other cooling systems, and/or afford the costs of operating cooling systems.
Many cities in 2020 had services that provided free air conditioners and box fans to the
elderly and vulnerable. New York City’s $55 million initiative to provide 74,000 air conditioning
units to vulnerable populations was the largest.44 Most other programs were significantly smaller
or relied on community donations (e.g. Houston, San Diego) . It is important to note that even
when cooling infrastructure is available in households, it is not a guarantee that people will use it
due to the high electricity costs.28 Thus, utility assistance and utility shut-off bans are also key
policy levers that can be used to make cooling resources affordable and accessible. We found that
all 25 of the cities studied here had some form of moratorium on utility disconnections
throughout the summer of 2020. However, some moratoriums ended before the summer heat
ended. Furthermore, these moratoriums on shutoffs did not reduce the accumulation of financial
stress on COVID-19 impacted families. Many families were still likely to avoid the use of air
conditioning and/or allow indoor temperatures to reach unsafe levels due to concerns over future
financial burdens, as has been found in pre-pandemic studies of household energy
insecurity.224,225
The federally funded Low Income Home Energy Assistance Program (LIHEAP)
program226 is an important component for reducing energy insecurity, with households who
receive LIHEAP money reporting significantly improved health and economic outcomes.227 In
years past, only 20% of eligible households were able to be funded through the program, but the

25
25
Coronavirus Aid, Relief, and Economic Security (CARES) Act of 2020228 released an increased
$900 million in supplemental funds for LIHEAP.226 However, many of these funds became
available well after the summer of 2020, and thus, may not have been effective in offering
resources to help vulnerable residents.
Despite efforts to improve LIHEAP program accessibility, persistent barriers continue to
exist. For example, vulnerable populations require knowledge of these programs, the resources to
apply (i.e. internet for online applications), and in some cases, help from experienced
professionals to navigate the application process.
2.4 Discussion
The year 2020 tied with 2016 as the warmest year on record, according to the National
Weather Service, based on global average surface temperature. However, the record heat in 2020
came during a La Niña year, when a cooling effect would be expected, as opposed to the 2016 El
Niño year, when a heating effect would be expected.229 In the United States, this translated to
record high average temperatures throughout many cities such as Chicago, Illinois; Phoenix,
Arizona; and New York City, New York. State health officials in Arizona found that 467 heatrelated deaths occurred in 2020, far exceeding the previous record of 283 heat-related deaths
reported in 2019.230 Record temperatures were concurrent with increased infections of COVID19 throughout the summer of 2020. From June 1st to July 31st, 24% of United States counties and
63% of the U.S. population were designated as COVID-19 high-incidence counties (. i.e., >100
weekly cases and increases in cases in the preceding 3–7 days).231 Figure 1 illustrates (a.) Daily
Maximum Temperatures and (b.) week-averaged daily COVID-19 cases per 100,000 persons
reported throughout the period of April 2020 to September 2020 for the counties containing the
25 largest municipalities in the United States investigated in this study. Highlighting case rates
during days when temperatures are over 90 degrees Fahrenheit highlights the high concurrency

26
26
of COVID-19 with high temperatures (Figure 1.c). In ten cities, forty or more days of
temperatures over 90 degrees coincided with high-incidence of COVID-19 (i.e., El Paso: 84
coincident days, Houston: 81, Dallas:69, Phoenix: 66, San Antonio: 65, Austin: 60, Fort Worth:
60, Jacksonville: 56, Oklahoma City: 49, Charlotte: 40). While these relationships are not causal,
there is evidence to suggest that increased temperatures may have increased the spread of
COVID-19 as a result increased congregation indoors. 232 Regardless, outbreaks during these
periods made executing heat response plans more difficult due to social distancing protocols.
Figure 2-1 Heatmap Representation of Extreme Heat and COVID-19 in from April 1st, 2020, to October 30th, 2020 for the 25
largest US Cities by Population. Cities are ordered by the number of total 90F or warmer days in the study period from top
(highest) to bottom (lowest). (a.) Daily Maximum Temperatures are shown from the U.S. National Oceanic and Atmospheric
Administration’s Global Historical Climate Network – Daily (GHCN-Daily) dataset. Maximum Temperatures are calculated as
the average daily maximum temperature of all stations in the county containing each city. New York City weather is calculated as
Queens County, NY. (b.) COVID-19 7-day Average Daily Case Rates are shown. Data is smoothed using a cubic spline to reduce
reporting noise and to match CDC criteria for high incidence. (C.) COVID-19 7-day Average Daily Case Rates shown for any
day that had a Daily Maximum Temperature from (b) greater than 90F to highlight concurrent high temperatures and COVID-19
cases

27
27
These health impacts were not unique to the year 2020. In 2021, a heat wave in the
Pacific Northwest of the United States brought record breaking temperatures over five days that
resulted in an estimated death toll of 800 across the region.233 The unique challenges of
managing both COVID-19 and extreme heat continued to place stress upon urban systems.
Hospital system was overwhelmed, with waiting times of five to seven hours, especially during
the early afternoons.234 Limited access to cooling as well as COVID-19 related social isolation
played a major role in adverse heat health effects, especially in the South and the West of the
US.235 As the US and other countries global deal with COVID-19 as a potentially endemic
challenge, the complex challenges of managing infectious disease during extreme weather events
will remain a challenge as both extreme temperatures and disease burden increase in frequency
and intensity as a result of Climate Change.
Extreme heat events and COVID-19 pandemic concurrent events were not isolated from
other climate-related risks. In California and Arizona, a prolonged heat event in August 2020
induced record-breaking temperatures and increased fire danger.236 Combined with lightning
strikes and high wind gusts, hundreds of wildfires sparked across the state of California and
combined to form the August Complex Fire, the largest wildfire in California history.237 The fire
resulted in one million acres burned with several hundred structures destroyed, tens of thousands
of residents forced to evacuate, and multiple injuries and deaths.238 Additionally, at one point, the
wildfires led to the worst measured air quality worldwide.239 These compounding risks created
strained conditions for medical care. Fire, heat, electricity shutoffs, and poor air quality forced
the evacuation of some hospitals and prevented some COVID-19 testing sites from opening.
240,241 Similarly, moisture leftover from the remnants of Hurricane Laura combined with late
summer sunshine to produce heat index values of 105 degrees or greater across much of the Mid-

28
28
South. Over 300,000 customers in Louisiana and Texas lost power after Hurricane Laura
inflicted major damage to infrastructure.242 The loss of electricity and heat led some hospitals to
evacuate. 243
These compound crises underscore a key challenge in risk-based approaches; riskmanagement of compound events requires understanding the probability, severity, and
consequence of numerous threats and the failures those threats could cause on complex and
interconnected infrastructure systems.
Many of the same strategies for heat-adaptation, such as cooling shelters, are transferable
to other climate disasters, such as wildfire and post-hurricane shelter requirements. These
congregant sheltering solutions provide efficient ways to protect public extreme weather but pose
major public health risks during a contagious pandemic. The impact of COVID-19 on heatadaptation strategies provides insight into how we can prepare for future pandemic-climate
events. Many of the same racial inequities to cooling access have been mirrored by access to
health care, such as vaccinations.244 Developing methods to spatially allocate congregant
resources, such as cooling centers, could help cities to quickly adapt to new compound
challenges.
There are also major lessons learned for public health departments in developing best
practices to opening facilities to COVID-19. The on-the-ground challenges of opening a cooling
center while maintaining social distancing was a key challenge for the emergency response
community. While Interim CDC guidance has suggested new social physical distancing, air
filtration, cleaning, and personal protective equipment to operate cooling centers, 245 operators at
a state and local level also developed their own guidelines and best practices ranging from
individual screening questions to rules about food delivery, masking, and re-entry.58,85 While

29
29
congregant solutions like cooling centers were limited by social distancing requirements, noncongregant solutions allowed better access to cooling resources by using a high number of
smaller, more distributed resources. City buses, for example, have played multi-purpose roles
throughout the pandemic by being a mobile resource that promotes social distancing by
increasing the number of available resources and bringing services directly to those who need it.
Beyond extreme heat, busses played an important role as mobile wifi-hotspots to provide internet
access for school children.246 As urban systems start to integrate new distributed technologies,
such as rooftop solar electricity, electric vehicles with vehicle-to-grid-functions, and microgrid
technologies, emergency management departments have an opportunity to develop agile
strategies that can move life-saving resources to vulnerable populations during adverse events.
We found that the cities that were most successful in providing access to cooling
resources utilized resources beyond traditional public health and emergency management
departments. For example, prioritizing often overlooked components of urban infrastructure such
as libraries and pools was integral in continuing to provide heat resources to the public during
EHEs in 2020. Informing vulnerable populations of and helping them reach key services were
also critical aspects of successful heat-resilience strategies. The elderly and other heat-vulnerable
populations were significantly more likely to seek out these heat-resources during extreme heat
events because of the role libraries, pools, and parks play in civic life.18,247 Engaging
stakeholders across different city departments enabled new strategies that leveraged diversified,
distributed resources that were already integrated into communities, such as public transportation
and schools. City departments can also work with grassroots actions undertaken by business,
community leaders, and organizations to identify options that reflect their options and
concerns.248

30
30
A large breadth of new data sources and analytical methods can be leveraged to identify
vulnerable populations beyond traditional demographic factors, in order to target and optimize
heat response strategies. Recent research illustrates how residential smart meter data can be used
to detect whether a household uses an air conditioning unit.249 Given that Advanced Metering
Infrastructure is now abundant, particularly in densely packed cities, utilities could use these data
to identify heat vulnerable populations.250 Mobility data from cell phones has been a key source
of data for the influence of travel on COVID-19 spread.251 These data could be used to help to
better quantify the de-facto usage of cooling resources and identify sites for cooling centers. As
physical access to cooling resources improves, the heat vulnerable will need better access to
information about what de-facto or de-jure cooling facilities are available and accessible. One
strategy to improve access to cooling access may be to add whether buildings have air
conditioning to the results of search engine (e.g. Yelp, Google Maps, etc.) or on public websites
to help disseminate information about how and where low-cost air conditioning access is
available.
In the long term, cities can also work to reduce the likelihood of extreme temperatures.
For example, city planning to require cool pavement, energy efficient buildings, and green
infrastructure to combat the urban heat island may help to reduce the magnitude of climate
impacts faced by residents.252,253 Large scale policy initiatives to mitigate climate change can
help may help to reduce the more frequent, more severe, and longer lasting heat waves
experienced by cities.254 However, cities must also build in the institutional and financial support
systems to improve the resilience of cities by increasing access to cooling resources as the
climate warms.

31
31
2.5 Conclusion
The COVID-19 pandemic disrupted the day-to-day operations, health infrastructure, and
economies, of cities throughout the world, exacerbating the already difficult challenge of
adapting to a climate-change exacerbated climate hazards. Given the growing frequency,
intensity, and duration of EHEs, it is likely that future management of EHEs will require
navigating two simultaneous challenges in the future. Our analysis explored how cities around
the United States changed their EH-response strategies to meet the unique conditions created by
the COVID-19 pandemic.
Our analysis found that EHE-response strategies were impacted by the COVID-19
pandemic not only by policies to enacted limit the spread of COVID-19, but also the economic
and staffing limitations induced by those pandemic-related public health restrictions. We found
that the cities most able to provide resources to residents leveraged resources beyond traditional
public health and emergency management departments. Our work underscores that evaluating
multipurpose uses of physical infrastructure and developing improved analytical tools to identify
the vulnerable will be a critical step to enabling improved responses to complex compound
challenges.

32
32
Chapter 3 - Characterizing Contributions to the
Residential Sector Load Curves from Smart Meter
Datasets
The content included in this chapter is currently under review in:
Jin, Andrew Shida, and Kelly T. Sanders. "Characterizing Contributions to the
Residential Sector Load Curves from Smart Meter Datasets." Applied Energy 2024 (Under
Review)
3.1 Introduction
In recent years, the global energy landscape has begun a transformative shift towards less
carbon-intensive forms of electricity generation in efforts to mitigate the negative consequences
of climate change. Renewable electricity is projected to be the largest source of global electricity
generation by early 2025 and is the only electricity generation source whose share is expected to
grow.255 Renewables are expected to displace fossil fuels in the United States electric power
sector due to declining renewable technology costs for renewable power through 2050.256 Unlike
conventional thermal generators, wind and solar PV power have variable power outputs that
create operational challenges in balancing supply and demand.
257 To deal with these challenges,
utilities around the United States and internationally have implemented new strategies to better
forecast and manage energy supplies and demands.258
In California, this challenge of balancing supply and demand is already growing more
difficult because of a changing climate and changes in demand patterns due to increased
electrification. Solar generation within the service territory overseen by California Independent
System Operator (CAISO, which covers 80% of California's bulk power transmission259),
represented 17% of total system electricity generated in 2022.260 In CAISO, high penetrations of
solar power have created a so-called “duck curve” that is characterized by a deep daytime net-

33
33
load (where net load is the total load less variable renewable electricity) dip followed by a steep
increase in net load that occurs as solar resources go offline in the early evening and CAISO
simultaneously enters its evening peak.261 The duck curve can lead to overgeneration and
curtailment of renewables during the midday “belly” of the curve, while the “neck” of the curve
requires conventional generators to rapidly ramp operations to address peak demand in the early
evening, leading to higher costs and more emissions.262 As California continues to grow its solar
power portfolio, CAISO's midday net load continues to dip lower year over year, exacerbating
issues with overgeneration and ramping challenges.263 While CAISO's battery storage capacity,
which increased tenfold from 2020 to 2023, provided key net peak capacity and energy during
the summer of 2022, its deployment is still relatively limited and not yet sufficient to fully
ameliorate the challenges of managing CAISO’s duck curve.
Because grid operators must match supply constraints with electricity demand, systems
like CAISO have moved from a paradigm of “matching available supply with dynamic demand”
to “matching dynamic supply with dynamic demand” through demand side management
strategies, like demand response. 264 Demand response strategies are designed to influence
customer use of electricity in order to encourage customers to use less power during peak times,
or to shift energy use to off-peak hours. 265 Demand response strategies, as defined by the United
States Federal Energy Regulatory Commission, are “changes in electric usage by demand-side
resources from their normal consumption patterns in response to changes in the price of
electricity over time, or to incentive payments designed to induce lower electricity use at times of
high wholesale market prices or when system reliability is jeopardized.”266 Demand response
programs can include incentive-based demand response programs, such as direct load control,
emergency demand response programs, or interruptible/curtailable rates.267

34
34
While most demand side management strategies have historically been targeted at larger
electricity customers in non-residential sectors, state policies in California have accelerated the
implementation of new time-of-use (TOU) rates in recent years.268 In 2003, a residential TOUpilot program was designed and implemented with the help of the three major investor-owned
utilities in California: Pacific Gas and Electric Company, Southern California Edison, and San
Diego Gas and Electric Company.269 Since then, a 2016 rule making effort has been put into
place an effort to analyze the electricity load of CAISO and develop analyses to develop new
TOU designs,270 with a pilot study commissioned in 2018 proposing the implementation of
default TOU pricing for all residential electricity customers.271 Time-of-use rates have been
implemented as the default pricing scheme for customers in PG&E 272 and SDG&E273 since 2019
and SCE since 2021.274 However, while there has been a concerted effort to develop such rates
and understand the impact of such pricing rates on decreasing the load at peak hours,
understanding the load shape of the residential sector remains a key gap in the literature.
Currently, the end-use load profile models developed to understand the California’s
demand in end-use sectors give a poor understanding of how residential electricity demand
fluctuates throughout the day. Until 2017, the residential load profile used by the California
Energy Commission was informed by the CEC’s Hourly Electric Load Model (HELM)275 ,
which was based on a small number of metered homes in the late 1980’s. The most recent
California end-use load shape report uses aggregated smart meter data from various houses to
develop average daily residential load profiles.276 However, the report only details an annual load
profile and does not offer detailed insights into differences between customers within the
residential sector (e.g., by housing type, by climate zone, by season, etc... ). Hence, there are key
gaps in our understanding of the differences in how customers within the same or adjacent

35
35
geographic regions contribute to the aggregate residential load curve. While demand response
strategies incentivize behavioral changes in individuals, the primary goal of demand response
strategies is to reduce net demand during peak times.277 Thus, understanding how different
customers contribute to the overall load curve can help utilities design more effective and
equitable residential demand response programs.
This study develops a longitudinal data-based approach that constructs aggregate
residential load curves and evaluates the contributions of utility customers based on their annual
electricity usage across different day types (i.e., weekday vs weekend), seasons, and climate
zones. It uses the real-world hourly smart meter data of a statistically representative sample of
electricity users within Southern California Edison’s service area to answer three major research
questions regarding residential load curves:
a. What are the diurnal and seasonal characteristics in the aggregate residential load
curve in Southern California Edison’s service area?
b. How do customers within different electricity use decile bins contribute to the
overall load curve throughout seasons and do higher consumption deciles have an
outsized contribution on residential electricity load during specific hours of the
day?
c. How do customers in different climate zones contribute to the overall load curve
(i.e., in terms of shape and magnitude)?
In this study, we evaluate a large, multi-year smart meter dataset to create average
electricity consumption profiles for households in Southern California. We classify utility
customers into one of ten electricity use decile bins, which represent the percentile ranking of
each household based on its typical daily electricity consumption. These deciles are determined

36
36
by computing the average daily electricity consumption for each customer based on all their
records in our multi-year dataset, which could span up to four years depending on the household.
This study develops average load profile shapes for customers in Southern California that can be
used to assess differences in the shape of the electricity load by season, day of the week, climate
zone, and consumption-based decile bins. We then use a GINI metric to characterize disparities
in the contribution of the overall load curve by customers in different decile bins.278 Thus, this
study not only characterizes the load shape of the residential sector electricity load profile in
Southern California, but also gives insight to the disparate contribution that customers who use
different amounts of electricity have on the diurnal patterns of the residential load profile. These
insights can help enable (a) the tracking of how the residential electricity consumption shifts over
time and (b) more equitable demand response targeted policies.
In the following sections, we review the state of the literature in the load profiling
community and identify existing knowledge gaps. Then, we present our methods to construct
load profiles and characterize disparity and describe the results of our study by analyzing the
load profiles in Southern California by season, decile bin, and climate zone. Finally, we discuss
the implications of the load profile shapes and how they can be used to inform improved demand
response strategies.
3.2 Literature Review
3.2.1 Load Profiles
Load Profiles (also referred to as load shapes, demand shape, load curve, demand curve, etc.)
are a way of characterizing the variation of demand and electrical load over a specific period. In
the academic literature, these have been largely referred to the diurnal hourly or sub-hourly graph
of electricity use of individual customers (i.e. individual meters), groups of individuals, or

37
37
different customers sectors (i.e. residential, commercial, industrial). The construction of load
profiles have been used for a number of different tasks ranging from supply management
requirements, such as electricity forecasting279–282 and system design283–285, to demand-side
analyses, like tariff design.286–288 Load profiles can generally be divided into two general models,
one that characterizes a large group of potential customers to establish standard load profiles, an
aggregate electricity profile that represents the overall behavior of the residential sector, and one
that characterizes the behaviors of subgroups of customers that have similar behaviors. 289
The residential sector presents several unique challenges that make it harder to
characterize load profiles. The decentralized nature of residential customers makes it difficult to
characterize their consumption. While residential customers represent 39% of end-use electricity
consumption in the United States, they represent approximately 87% of utility customers. 290
Electricity customers have a high diversity in building characteristics and occupant
demographics that lead to dramatic differences in the load profiles of different households.291
Even with nearly identical houses with the same appliances and similar demographic occupants,
the load profiles of two different homes can be very different. 292 By contrast, other sectors, such
as commercial, industrial, agriculture, and transportation, generally attract more significant
financial motivation to analyze and understand their electricity usage patterns and have more
centralized platforms from their respective industries to compare patterns of behavior.293
Historically, load profiles of the residential sector can either be surveyed by utilities themselves
(typically limited to a small number of voluntary survey participants), or the load profile of a
customer group can be measured at the transformer or branch feeder level.294
Several reviews have reviewed methods for developing load profiles in the residential
sector.289,295–297 Because of data limitations, two main archetypes to have been used to interpolate

38
38
the load profile, bottom-up and top-down strategies. Top-down approaches rely on aggregate
data from utility operators, such as the all-sector load curve and apply statistical methods and
models to make inferences about what the residential sector may look like.298,299 One approach to
top-down modeling is Conditional Demand Analysis, which uses actual smart meter and
appliance data to train models that disaggregate residential electricity use into functions of
specific appliances, without the need for direct appliance measurement, and scales it up to the
overall electricity load profile. This approach is exemplified in studies by Parti and Parti300 and
Aigener et. al301 . Bottom-up approaches rely on detailed data from individual customers or
appliances and use simulation models or other engineering methods to aggregate a residential
load profile. 302 For example, Capasso et. al implemented a Monte-Carlo simulation based on
various factors such as weather, time of day, and household appliance curves to create an
aggregate total load.303 While top-down approaches have high uncertainty into the actual
behavior and preferences of individual customers, bottom-up approaches may not capture the
true nature of customers because of their limited sample sizes. Thus, generating data-driven load
profiles, which has been enabled by the proliferation smart meters, can provide a real-world
based analysis of electricity customers.
3.2.2 Smart Meters Data Analysis in the Residential Sector
Smart meters, or Automated Metering Infrastructure (AMI), that provide hourly or subhourly information about electricity usage, are being installed at households in many regions.
The growth of AMI has largely been driven by the operational benefits to utilities as automated
metering can reduce the number of site visits to read meters, identify disruptions to service
quickly, and more accurate and timely billing.304 Because these operational benefits have been

39
39
the driver of the implementation of smart meters, the use of smart meters as informational tools
to understand how customers consume electricity has been underexplored.305
Past reviews of smart meter analyses have found the large majority use relatively small
numbers of smart-meter readings for their analysis, with many using small open-source
residential datasets on the order of thousands of unique meters.306,307 The use of massive datasets
with more than 100,000 users that have been applied to academic literature has been limited only
a small number of distinct datasets.308–315 Of these aforementioned papers, two papers were
published directly by the utilities regarding internal operational research to improve the
forecasting of electricity usage 309 or to improve automated customer feedback in billing308. By
contrast, most of the papers that analyze these large datasets seek to understand how customers
use electricity through the construction of load profiling and load analysis techniques.310–314
One key topic of research is to understand how individual electricity customer’s
electricity load is affected by the use of different installed appliances. For example, Chen et. al
uses a dataset of over 180,000 customers in Southern California to analyze the existence of air
conditioning within households by analyzing the temperature sensitivity of electricity usage.315
Other analyses use smart meter electricity datasets tagged with the presence of different
appliance technologies to analyze the impact of those technologies on the electricity profiles of
houses. Gunkel et. al. studied a dataset of 667,373 houses in Denmark, with smart meter data
matched to data about each household regarding the income and age of occupants, heating
technology, and the presence of electric vehicles to understand differences in the level and timing
of consumption throughout the day.
310
Load analyses have largely focused on characterizing load profiles in the residential
sector by first segmenting customers by their similar electricity load profiles through clustering

40
40
techniques. Generally, these methods characterize load profiles for groups of customers. Kwac,
Flora, and Rajagopal (2014) first proposed and applied a method to segment households based
on their hourly electricity consumption patterns using a dictionary of load shapes to characterize
similar load profiles in 218,090 homes in Northern California. 316 A similar clustering method
simplifies the load shapes by developing average load profiles for each user by month and for 2.5
million Illinois customers.317 More sophisticated versions of this clustering methodology have
been applied as well. For example, Ushakova and Mikhaylov propose a Guassian-Mixture model
to encode individuals’ energy consumption over time based on 400,000 homes in the United
Kingdom.314
While the aforementioned group of papers characterize homes by similar behaviors,
fewer papers have aggregated multiple customers based on temporal factors, such as seasonal
and day of week differences, or subsets of the population before describing their load. A study
analyzing the impacts of COVID-19 restrictions on residential electricity usage used 230,000
smart meters to analyze week-long hourly resolution smart meter load profiles in Santiago,
Chile, finding that median household-loads were formed for individual neighborhoods to
compare residential demand.312 Gunkel et. al (2023) used the aforementioned Danish smart meter
dataset in 310 to analyze the contributions of individual customers within a customer class to the
aggregate profiles of that class.318
Current load profiling methods that utilize massive volume smart meter datasets have key
limitations that we address in this study. No high-volume smart meter dataset analysis has
tracked customers across a span of five years. Our study addresses a key gap in understanding
how a household’s load shape might vary across climate zones. Furthermore, no paper has
analyzed the contribution of utility customers, binned-according to their typical household

41
41
electricity use, to the aggregate load profile. We then analyze the diurnal patterns of disparity in
hourly electricity use of these binned customers. In the following section, we describe our
methodology of creating data-driven load profiles and analyzing the contributions of and
disparities between different users.
3.3 Methods
3.3.1 Data Description
Our data was sourced from Southern California Edison, a large investor-owned utility
serving the greater Los Angeles area. We obtained hourly electricity data from 200,000
residential smart meters, randomly selected so that the dataset is statistically representative of 5%
of the SCE’s 4.5 million residential households in their service area which spans the Greater Los
Angeles. [See Supplemental 1] Data spanned from 2015-2016 and 2018-2019.
Each residential smart meter was associated with a unique service-account ID,
representing data for one utility customer account. All electricity data were stored on USC’s
center for High-Performance Computing with a highly secure HPC Secure Data Account, to
remain in line with the security and confidentiality requirements of SCE. In our analysis, we only
consider the electricity load profiles of sub-metered residential units (e.g. apartment buildings
where each apartment has a meter) and filter out master-metered properties (e.g. apartments
buildings or trailer parks with one master meter).
3.3.2 Individual Load Profile Definition
In this paper, we begin by analyzing the individual load profiles of all utility customers
(each representing a household with a smart meter) within our dataset. The dataset was
organized into customer-day records, where each day is represented by a unique hourly

42
42
electricity load profile. Each individual load profile record, �, is represented by a set where �!
represents the electricity used in a one-hour period, measured in kWh, at the hour, starting at ℎ,
of day, �, for an individual customer �. [Equation 1]. Unlike the traditional electric industry-wide
term “hour ending”, we will refer to hours by the starting hour of the electricity use. For
example, hour ℎ=16 will refer to the electricity used during the hour starting at 4:00 pm and
ending at 5:00pm. We then calculate the total daily electricity, �, measured in kWh, for an
individual load profile.
�(�, �) = [�!, �", … , �#, … , �$%] 1
�(�, �) = ,�#
$%
#
2
For each individual utility customer, �, the set �" is the set of all complete sets of full-day
records in a customer’s dataset. Throughout this analysis, we will use the prime notation (�"
#
) to
describe the subset after filtering.
3.3.3 Filtering of Non-Residential and Outlier Residential Electricity
Customer
In the dataset provided by SCE, each customer is identified by a service-account ID
which is unique to each customer and is affiliated with a postal address. This address was parsed
and matched with building assessor data in Los Angeles, San Bernadino, Orange, and Riverside
Counties, and found that less than 1.5% of households were matched to non-residentially zoned
parcels. [Supplemental Methods – Section 2.D] The addresses were then attached to a geographic
coordinate and tagged with a California Energy Commission Building Climate Zone.
The set �", is the set of all complete sets of full-day records, �, available for an
individual customer, �, across �", the set of all years, �, in which that customer has data. We

43
43
include customer-years of data that meet the following conditions: First, the set of customer-day
records for a customer in a given year must contain more than 200 customer-day load profiles.
Second, the total electricity usage in that given customer-year must be greater than 20 kilowatthours (Equation 3). If a customer does not have at least one qualifying year of data (i.e. Y$
# is an
empty set), that customer is excluded from our analysis.
�&
' = {� ∈ �: 2�&,)2 > 200 ∧ ∑ �(�, �) *∈,!,#
> 20 kWh} 3
We performed a manual assessment of the largest electricity customers in our
dataset, whose average annual electricity usage was greater than 100,000 kWh, to check if these
customers represented one household. Records that were associated with addresses where more
than one household was collectively measured by one meter (e.g., trailer parks or housing
communities with only one SCE meter) were excluded. [See Supplemental 2.D] After all
filtering steps were completed, our final dataset constituted 184 million customer-days of hourly
smart meter readings data from 163,403 customers over 514,370 customer-years.
3.3.4 Customer Electricity Use Decile Bins
For each customer, we calculate the average daily electricity use value, �(�, �), for each
year � which satisfies our inclusion criteria �"
# (Equation 4). Averaging to the daily level reduces
the influence of potentially missing data in each year and allows comparison between different
customers. We then calculate the average all annual-daily average electricity use, �(�), by
averaging the set of eligible customer-years (Equation 5). Thus, �(�), measured in kWh per day
is the mean-annually-averaged daily electricity use across the time period of study for each
customer.

44
44
To characterize the relative usage of electricity customers to one another, we rank all
electricity customers based on their average annual electricity use, �(�). Annual electricity
usage is then used to order all individual customers with eligible data through a percentile
function, �(�) (Equation 6). Once percentile is calculated, each individual customer is assigned
an electricity use tier, such that each decile bin, �%, represents a decile of electricity usage, where
decile 1 (0-10%) represents the lowest consumption customers consuming a daily average of
0.05 and 6.6 kWh and decile 10 (90-100%) represent the highest consumption customers
consuming a daily average of more than 33.4 kWh (Equation 7). [See Supplemental 3]
�(�, �) = ∑ �(�, �) *∈,!,#
2�&,)2 4
�(�) =
∑ �(�, �) )∈-!
$
2�&,)2 5
�(�) = ��E�(�)F
|�'|
6
�(�) = ⌈�(�) ∗ 10⌉ 7
3.3.5 Average Daily Load Profiles
We define three season classifications primarily based on weather patterns in Southern
California’s Mediterranean climate: “Winter” defined as November-December-January-February
(NDJF), “Spring” defined as March-April-May-June (MAMJ), and “Hottest Months” defined as
July-August-September-October (JASO). JASO, combining summer and early fall, is the hottest
period of the year, characterized by extended heat and dryness typical of Southern California
summers, which often extend into the traditional autumn months. Because of diversity of
climate zones contained within the SCE service area, these seasonal categories do not necessarily
capture all of the region’s micro-climates but serve to capture its broad climate patterns.

45
45
To characterize residential electricity use in SCE’s service area, we create characteristic
average daily load profiles that show the typical electricity load of a given set of customers on a
given type of day. Throughout our analysis, we subset customers according to California Energy
Commission Climate Zone (�) and electricity use decile bins (b). We subset customer-days by
day of the week (i.e. weekend vs weekday) and season (i.e. JASO, MAMJ, JASO). Thus, average
load profiles, �(�, �, �, �), represent the mean electricity in each hour, �!, from all customerdays for customers who lived in climate zone (�) with consumption in decile (b), on
weekday/weekend (w) days in season (s) (Equation 8 & 9).
�(�, �, �, �) = [�!, �", … , �#, … , �$%] 8
�# =
∑ ∑ �#(�, �) *∈,%,&
$ &∈.',(
$
2�/,0,1,2 ' 2 9
3.3.6 Gini Coefficient Calculation
To examine the distribution of electricity consumption among customers and quantify
their relative contributions to the SCE aggregate residential electricity load, we derived average
load profiles for each customer, i, denoted as �(�, �, �), based on season (�) and day type (�). At
each hour of the day, we use the typical customer profiles to construct a Lorenz curve 319 which
shows the cumulative proportion of electricity consumed at hour, h, up to the �th -ranked
customer by �(�) in ascending order C!(i):
C!(i) = ∑ l!(�, �, �&) $
'()
∑ l!(�, �, �&) *
'()

46
46
The Gini coefficient (GINI) is calculated using the differences between the Lorenz curve
and a line of perfect equality.278 [See Supplemental 5] We calculate the Gini coefficient as the:
GINI(h) = 1 − C(P+($) − P+($.)) )(C/(i) + C/(i − 1))
*
$()
In this case, the Gini coefficient is a value between 0 and 1, with 0 representing perfect
equality (everyone uses the same amount of electricity) and 1 representing perfect inequality
(one customer uses all electricity and no other customers use any electricity).
3.4 Results
In this section, we describe the average residential load profiles by season and year. Because
our data is statistically representative of the SCE service area, the shape and magnitude of the
load curves are representative of the residential sector in SCE. (Error! Reference source not
found.)
Weekend curves are similar in shape to weekday curves within the same season. The average
daily load of weekends is slightly higher than that of weekdays within the same season. The
majority of the differences between electricity use between weekdays and weekends occurs
during the 9 am through 5pm workday, with 75% of the variability between the two curves in
the hottest months and 62% in spring and winter months. During this 9am to 5pm period, there
is higher weekend usage within the same season.

47
47
Figure 3-1 The average daily load of a residential electricity customer in the SCE service area is highest during the hottest
months (i.e., in JASO; July, August, September and October). Peak electricity usage occurs earlier and is much higher in hot
months than other months. Daytime weekend usage is typically higher than weekday usage. Data represent the average hourly
load profiles of approximately 160,000 SCE customers spanning the time period January 2015 through December 2019.
Like CAISO’s all-sector electricity daily load profiles, average daily load in the hottest
months is higher than average daily load in spring and winter months. In JASO, weekday
average daily load is about 23.8 kWh across the period of study. By contrast, the MAMJ and
NDJF weekday average daily load is 17.0 kWh, 28% lower than the JASO daily load. The
seasonal load profiles are relatively similar in both shape and magnitude throughout the late
evening early morning periods (6pm to 8 am) in all seasons. The biggest differences that drive
the higher electricity in JASO occur during the mid-day, presumably due to cooling loads. In
JASO, the residential load profile peaks at 17:00, two hours earlier than the 19:00 peak in both
MAMJ and NDJF. This JASO peak hour electricity consumption is also significantly higher

48
48
(1.47 kWh), with non-JASO peak hour consumption less than 1 kWh about 35% lower in
electricity usage.
3.4.1 Average Load Profiles by Electricity Use Decile Bin
We analyzed diurnal electricity consumption profiles on weekdays in NDJF, JASO, and
MAMJ by electricity use decile bins (Figure 2). Each unique customer is assigned to the same
decile bin in all seasons based on their average daily usage across the entire time period of study.
Thus, some users may not be represented in all years. This figure shows that the differences in
electricity usage across utility customers at each hour of the day can be large. For example, in
JASO, customers in the highest electricity use decile bin consume 59.5 kWh of electricity per
day, more than twice the average daily load of the average daily load of all users and more than
12 times the average daily load of the lowest electricity use decile bin.
Figure 3-2 The average load profiles of SCE customers segmented by electricity use decile bins show that the highest consuming
customers represent a disproportionate fraction of daily load. Note: Each unique customer is assigned to the same decile bin in all
seasons based on their average daily usage across the entire time period of study. Darker colors represent higher usage customers.
The total daily load was highest in JASO for all decile bins of customers. Thus, while the
impact to the diurnal shape of electricity consumption changes throughout seasons, the increase
in magnitude of total daily load remains a consistent trend for all decile bins. Most bins have
about 40-50% lower weekday electricity consumption in non-JASO months compared to the

49
49
JASO months. The exception is the smallest two deciles of customers where electricity
differences hover around 15-20% lower weekday electricity use. This suggests that while there
are seasonal differences in the magnitude of electricity consumption, the smallest customers (by
electricity usage) exhibit less seasonal variability.
Generally, the average load profile of the middle 60% of customers is similar in shape to the
average residential weekday load profile of all SCE customers.” By contrast, the customers in the
two highest consumption decile bins vary from lower decile bins. These large consumers have an
average weekday load profile with a smooth singular peak in JASO and a double peak (i.e., one
in mid-day and one in the early evening) in the Spring and Winter. For lower deciles, Spring and
Winter load profiles generally consist of a smaller morning peak and a more pronounced evening
peak. The timing of the peak hour shifts earlier in the day (16:00 and 17:00) for most decile bins
in JASO, but the peak hour timing of the bottom three decile bins of customers stays consistent
around hour 19:00 and 20:00 across the whole year. Customers in the bottom three decile bins
have relatively flat profiles across seasons, when compared with higher usage customers. Like
other customers, the magnitude of their average electricity usage is highest in JASO, but the
difference in magnitude is smaller across seasons compared to higher usage customers.
3.4.2 Dispersion of Electricity Customers throughout the Seasons
Next, we measure the diurnal patterns in disparities in electricity between the average
electricity of users. Cumulatively, the lower half of electricity users ranked by annual electricity

50
50
use consumed less than 25% of total JASO weekday usage. By contrast, the top twenty percent
of electricity users comprised more than 40% of all JASO weekday electricity usage.
There are differences with the distribution of electricity consumption based on the timing
of the day. In the early morning, electricity disparity is at its lowest but increases through the
midday, as shown by a larger concavity in the Lorenz curves in the early morning compared to
mid-day. At 6 am, when electricity disparity is relatively low, the lower 50% of customers
consume about 30% of all electricity. By contrast, noon-time electricity the lower 50% of
customers only constitute 22% of electricity use.
The disparity of electricity usage is measured using the Gini coefficient, which represents
dispersion across a population on a scale from 0 to 1, changes by the time of day (Figure 4).
Higher values of GINI values represent higher dispersions across SCE customers (i.e. a smaller
number of high consumption customers consuming more than others). All three seasons
generally show a similar pattern of low disparity between households in the late evening and
early morning and reach the highest dispersion levels at around midday (i.e., prior to afternoon
Figure 3-3 Lorenz curve of electricity use in two different hours on weekdays in July, August , September,
and October. The dispersion is lower

51
51
peaking hours). This result suggests that a relatively small proportion of high consuming
households account for a disproportionate amount of midday electricity usage across the year in
SCE.
While the general shape of the diurnal disparity curve looks relatively similar across seasons,
the magnitude of disparity shows large seasonal differences. The Gini coefficient is highest in
magnitude during the mild spring months (green in Figure 4), particularly on weekdays, most
likely due to large heterogeneities of HVAC usage and time spent at home compared to hotter
and colder seasons. Cooling dominates midday loads across much of the customer base during
the hottest months (red in Figure 4), reducing midday dispersion across customers using AC
compared to in mild weather months. The dispersion in the hottest months is likely bigger than
the dispersion in the coldest months because space heating in California is dominated by natural
gas units, which cannot be seen in these electricity data. Hence, the larger disparity in electricity
Figure 3-4 The GINI Coefficient of electricity users suggests that the largest disparities in electricity usage across
customers occur in the middle of the day, on weekdays, and in mild months. (Higher values of GINI values represent
higher dispersions in electricity usage across SCE customers.)

52
52
consumption during hot months is being driven by large differences in residential cooling loads,
which can be impacted by home size, home insulation and weatherization characteristics, AC
unit technology, energy affordability concerns, etc.320 More generally, users across SCE
demonstrate the widest range in loads in the midday during typical working days in every season.
We conjecture that the disparity is higher on weekdays during typical working hours because
there are bigger differences in time spent at home across populations of customers compared to
weekends.
3.4.3 Climate Zone Differences in Electricity Use
SCE encompasses many different climate zones. Each climate zone represents a distinct
geographic region, which encompasses not only differences in the climate that customers
experience, but also differences in housing stock, built infrastructure, and demographics. Cool
climate zones include CZ 6 (coastal) and CZ 8 (inland coast). By contrast, the two climate zones
in the valley, CZ 9 (Inland Valley) and CZ 10 (Interior Valley), are classified as moderate climate
zones, with CZ 10 having summers that are classified as hot. CZ 14 (Low Desert) and CZ 15
(High Desert) are the hottest and driest climate zones in the region. The cool coastal climate
zones maintain relatively similar load shapes year-round, with only minor increases in peak
height and total daily electricity usage in the summer. By contrast, total daily electricity use
during the hottest months significantly increases in moderate and hot climate zones, and peakhour electricity shifts earlier in the afternoon during these hot periods. Decile breakdowns of
residents within each climate zone can be found in Supplemental Information Figure 6 and 7. In
climate zones 10 and 14, the ratio between decile 10 (i.e., the highest use customers) and decile 5
electricity loads is lower in the hottest months than compared to cooler months, suggesting that

53
53
there is more diversity in electricity consuming behavior when loads are not being driven by
cooling, which is consistent with the insights derived in the Gini coefficient analysis. In cooler
climate zones, by contrast, the ratio between the decile 10 and decile 5 is higher in the hottest
months, likely because there are large disparities in if and how households use AC during
comparatively mild summers.
Figure 3-5 Electricity load profiles differentiated by climate zone and season indicate that the hottest regions represent a
disproportionate amount of electricity usage in SCE, particularly during the hottest months. Note: The hottest climate zones are
denoted in red, moderate in yellow, and the coolest in blue.
3.5 Discussion
This paper highlights that there is a big breakdown in the magnitude of electricity use across
customers. We found that the top 20% of electricity customers use over 40% of the daily
electricity load during weekdays in the hottest months while the bottom half of electricity use
less than 25% of all electricity use. During milder months (MAMJ), the largest consuming
households show the disproportionally high usage compared to the average user, suggesting that
these households have more discretionary loads. For Californian utilities such as SCE, most
demand response efforts have targeted requests or incentives for load reductions during the
period spanning 4pm to 9pm, when solar generation availability decreases and general residential

54
54
electricity demand climbs towards peak net load. However, during the hottest months of the year,
the disproportionality of electricity use during critical 4-9pm period is greater than those hours
compared to mild and winter months. However, during the hottest months of the year, large users
use a disproportionate amount of electricity use during critical 4-9pm period, suggesting that
these customers are the most strategic for the utility to target for leveraging reductions during DR
events. If utilities can get these customers to participate in DR activities, such as precooling, that
shift loads from these late afternoon hours (when the grid is dependent on expensive gas
generators) to the midday (when the grid is served primarily by low-cost variable renewable
energy generators), there will be valuable cost, emissions, and reliability benefits for the utility
and customer base.321
In general, higher electricity usage is associated with higher income households and lower
electricity with lower income households, while use intensity (energy use that is normalized by
household square footage) is generally higher in low-income households.322–325 These findings
approximately match our analysis, but as found in the aforementioned studies, these findings are
not universally true. Within the top decile bin, about 16% of households were in census tracts
with an average household income greater than $150,000. [See Supplemental 8] About 3.1% of
households within the top decile bin of electricity use were in census tracts where the average
household income was less than the living wage for an individual in California.326 Thus, there
likely exists some high electricity consuming households with low household income who could
be disproportionately impacted by new demand response initiatives, such as time-of-use rates.
Low-income households may have less efficient appliances and low insulation/weatherization
interventions,
327,328 higher energy loads because of higher occupancy,329 or require electricitydependent medical equipment.330 In the lowest decile bin, 13% of users lived in census tracts

55
55
where the average household income was less than a living wage for an individual in California.
These customers are likely to be the most disproportionately impacted by rate changes yet
contribute little to the diurnal residential load profile, and potentially lack the flexibility to adjust
their electricity consuming behaviors during peak time.
There are big trends in electricity use within California that could markedly shift the trends
noted in this analysis over time. California has aggressive electrification efforts, including a goal
to phase out gasoline powered vehicles by 2035. (The state currently leads the United States in
electric vehicle registrations, with EV’s constituting one-fifth of California’ new car sales in
2023.
331,332) EV chargers increase the electricity use of a household an average of between 3 to
10 kWh on the days vehicles are being charged, with higher power draws concentrated during
specific hours.333,334 Furthermore, the state plans to implement a ban on natural-gas appliances
by 2030, with many municipalities already banning or discouraging the use of gas-fueled
appliances, such as heaters, washers, dryers, and stoves. Currently the warmest months in SCE
show disproportionate electricity consumption, but trends towards electrification will shift these
trends diurnally and seasonally. For example, compared to the rest of the United States,
Californian households are less likely to have appliances that use electricity with only 28% of
space heaters (vs 40% nationwide) and 21% of water heaters (US: 47%) using electricity in
2023. 335 These appliances have the possibility of increasing and/or shifting the small morning
peak currently seen in the Spring and Winter load profiles.
Climate change will also affect household electricity use, particularly in regions that
currently have relatively low AC penetration. AC penetration across the region is estimated at
69%, with the coastal areas of California having the lowest air conditioning penetration.315 While
our analysis shows that the cool, coastal climate zones (CZ6 and CZ8) have generally flat

56
56
electricity load profiles, households (which tend to be wealthier than average) within these
climate zones have the potential to grow significantly especially in the hottest months, due to
relatively low rates of existing air conditioning penetration that will likely increase as
temperature increases due to climate change.
336 (Even in these more mild climate zones,
temperatures have recently hit record highs that have exceeded 100 degrees Fahrenheit.337)
Using data-driven methods, researchers and utilities will be able to track in near-real time
electricity use behaviors.
Although smart meter data continues to grow more available, they are not yet ubiquitous and
their distribution to third party researchers is often prohibited by utilities due to privacy
concerns. However, as they become increasingly available, data-driven load profile studies like
this one can offer critical insight into the actual distribution of electricity usage among
customers, which cannot be generated from traditional bottom-up and top-down methods used to
characterize residential load profiles. While physics-based models, like the Department of
Energy’s ResStock, utilize detailed conditional probability tables to characterize variables such
as square footage, building construction, and heating and cooling appliances to generate regional
load profiles across the United States for current and future scenarios,338 this benefit is
constrained by the models' reliance on accurately representing the highly variable behaviors of
individual users. Past research in customer behavior using customer segmentation and social
science approaches have shown that customers may have dramatically different electricity use
behaviors despite similar demographics, appliances, and building designs.292,316,339,340 The results
of this paper suggest that even when in the same service area, different customers have different
load profiles because of differences such as by climate zones or electricity use bins. Therefore,
data-driven models are essential in highlighting the divergences in behavioral patterns and the

57
57
actual variances in electricity usage behaviors. Data-driven methods continue to grow in their
capabilities (e.g., using smart meter data to deduce the presence and usage of appliances using
non-intrusive load monitoring techniques341,342). As these methods improve, they can improve
the utility of bottom-up physics-based models seeking to develop forecasted load profile
scenarios by providing more accurate representations of actual electricity consumption
behaviors.
3.6 Conclusion
Developing insights into the patterns and magnitude of electricity usage in the residential
sector is key for utility managers and policy makers to ensure the long-term reliability of the
grid. To date, other studies developing load profiles of the residential sector have been limited by
the size of available datasets. By using a statistically representative dataset of the Southern
California Edison service area, we were able to characterize the load curves of statistically
significant subpopulations of households across the SoCal Edison service area by characteristics
such as total annual electricity usage, season, and climate zone.
This study provided a novel analysis of the contribution of groups of individual customers to
the overall electricity curve and disparities in electricity use. We found that a small fraction of
electricity users consumes an outsized proportion of overall and peak electricity use. By contrast,
the combined electricity use of the bottom half of household electricity consumers contributed to
less than one-quarter of the electricity use. The disparities in electricity usage across SCE
households are higher in the mid-day, especially in Spring since loads are not as dependent on
high HVAC loads. Hotter months have higher disparities in electricity use during peak hours than
cooler months because of wide differences in cooling-driven loads.

58
58
The methods provided in this paper give quick insights for utilities to understand how
residential electricity may change across days, seasons, and climate zones, and how different
groups across those spatio-temporal distinctions contribute to the overall residential sector
electricity usage. The region-specific trends identified in this study for Southern California are
likely to change due to a warming climate, as well as climate mitigation policies promoting deep
decarbonization and electrification.343,344 However, researchers can build off the framework
presented in this study to analyze diurnal changes in load profiles, both retrospectively to gain
insight into how electricity load profiles may shift due to external circumstances (e.g. weather or
local events) or to prospectively track trends and shifts in the residential load due to evolving
factors affecting electricity consuming behaviors.

59
59
Chapter 4 - Characterizing Individual and Collective
Peaking Behaviors in the Residential Electricity Sector
4.1 Introduction
In electricity grids with high penetration of variable renewable energy technologies (i.e.
solar and wind), utilities must dispatch a large amount of conventional electricity sources around
sunset and into the mid-evening hours, as solar availability decreases, and household electricity
pushes up demand.345 In California, this so-called “duck curve” challenge not only increases the
amount of conventional fossil fuel power plants needed to operate during these hours, increasing
costs and emissions, but it also puts strain on the electricity grid as peaking power plants must
rapidly ramp up production to meet grid needs, posing the threat of potential power shortages if
enough generation cannot come online fast enough.
263
Grid managers and utilities are exploring demand side management (DSM) strategies,
which try to reduce strain on electricity systems by encouraging customers to use less electricity
or to shift their electricity use to times of lower net demand.346 DSM strategies in the residential
sector are simultaneously promising and difficult to implement. While residential customers
represent 39% of end-use electricity consumption in the United States, they represent
approximately 87% of end-use customers.290 DSM has generally been implemented first in
other end-use sectors, such as commercial, industrial, agriculture, and transportation because
loads in other sectors are more centralized across fewer customers and are generally considered
more predictable.
293 Thus, gaining insights into the behavior of residential sector customers can
help policymakers and utilities understand when electricity is shiftable and can help support the
design of more targeted DSM strategies to reduce strain on the electricity system during system
peak hours.

60
60
While residential electricity loads have been a major area of research interest for over a
half century, the paradigm shift away from analyzing annual trends in electric energy demand to
understanding the diurnal behaviors of electricity usage has only gained major attention in the
past two decades.295,296 Residential electricity is dependent on numerous social, economic, and
physical pressures including the composition of household members (i.e. number, age,
composition, education), socioeconomic status, and building type.347,348 Even so, past studies
have found that even when households are controlled for demographically similar conditions,
consumption behaviors can vary significantly.292,349 This variability is a key challenge in
understanding how programs in the residential sector can be best design to meet the dual goal of
meeting the electricity needs of daily life, while optimizing for efficient and low emissions
energy systems.
Initial studies of residential electricity use behavior have focused on taking direct
measurements of electricity use to acquire data on electricity usage in individual households.
Many utilities led studies also implemented pricing experiments to understand elasticity of
residential electricity use.269,350,351 Because scaling such behavioral studies up is difficult,
modeling approaches have also been appealing.293 For example, recent models have combined
time-use surveys to create stochastic models of occupant behaviors in combination with physicsbased building energy models at scale to make long term forecasting decisions or simulate the
impact of new policies.352 However, while these models can be useful in developing an
understanding of how typical electricity is used, they still have key uncertainties relating to
whether those behaviors translate to electricity usage.

61
61
Smart meter infrastructure, which allows utilities to collect near-real-time measurements
of electricity use, has grown dramatically over the past two decades and enabled new methods to
track fine-grained behaviors in the residential sector. One major paradigm used to understand the
electricity behaviors of residential customers is a clustering-based approach. An early example of
this analysis type is by Kwac and Rajagopal (2014) who use a K-means based clustering
approach to create groups of load shapes to characterize groups of behaviors.316 Numerous
papers have applied and refined these clustering methodologies to identify groups of customers
who use electricity similarly and identify candidates for potential DSM programs.353–355 A key
component in understanding how to implement residential DSM is understanding key energy
intensive behaviors such as peaking behavior, i.e. the tendency for electricity customers to
consume at high rates for brief periods during the day.
356 Gunkel et. al used residential electricity
smart meter data for 720,000 households in Denmark to analyze the role of heat pumps and EV’s
on the residential load and their potential for coincident occurrences.357
Two major theoretical paradigms to conceptualizing how residential behaviors play a role
in shaping the collective residential load peak. The dominant ABC paradigm of electricity use
suggests that values and attitudes (A), which are believed to drive the kinds of behavior (B) that
individuals choose (C) to adopt, lead to individuals deciding to coincidentally use
electricity.358,359 By contrast, social practice theory contends that the collective peak is formed
from socially shared and loosely orchestrated practices that influence peaking behaviors, such as
work-schedules, holidays, and HVAC usage.347,360 By understanding the difference in variability
within individual customers, and the differences in the residential sector as a whole, utility
managers and policy makers can better understand whether to target individuals and their

62
62
individual choices during peak hours or key social practices that may influence peaking
behavior.
This paper seeks to fill gaps in existing knowledge by analyzing the electricity peaking
behaviors of Southern California residential customers using electricity smart meter data. We
propose and calculate metrics for peaking timing and peak magnitude for each of the 218 million
customer-day records in our dataset, representing more than 150,000 customers, each of whom
could have up to five years of electricity use. We then analyze these metrics to answer three
major research questions:
1. What is the distribution of peaking behaviors in terms of magnitude and timing
observed in Southern California? How do seasonality and geography impact these
differences?
2. What are the observable trends in the timing and magnitude of peak behaviors of
individuals in the residential sector?
3. How much do the behaviors of individuals in the residential sector vary in time and
magnitude between days? What is the impact of seasonality on this difference?
The remainder of this paper is structured as follows. We first describe the data that were
used for this analysis, the metrics calculated for each customer-day record, and the aggregate
statistics that we calculated to measure peak behaviors. We then discuss the collective behaviors
observed in our dataset and then explore the characteristics of individual customer peaking.
Finally, we discuss the implication of these results in the context of current rate policies of the
Southern California Edison service area.

63
63
4.2 Methodology
4.2.1 Data Description
Smart Meter Dataset
Hourly residential smart meter data was provided by Southern California Edison, a large
investor-owned utility serving the greater Los Angeles area. This dataset included 200,000
households, randomly selected so that the dataset is statistically representative of 5% of the
SCE’s 4.5 million residential utility customers in their service area which spans the Greater Los
Angeles area. Data spanned from 2015-2016 and 2018-2020.
Hourly smart meter data was provided as interval data, with each data point containing
the electricity used in each hour of the clock, measured in kWh. All measurements were done in
local time. In Southern California, daylight savings time is observed. Thus, on days where an
hour is gained (i.e. spring forward), all user records were excluded because they were missing
the 2 am to 3 am record. On days, where fall back occurs, user days are included, but the 2 am to
3 am period represents both the first and second 2 am to 3 am time periods experienced by users.
All other days are recorded based on the clock hours that are experienced by users.
Filtering of Smart Meter Data
To be analyzed in our study, a day of data had to contain 24 records of hourly electricity
data and the total electricity consumed in that day had to be positive. Each year of customer data
was considered for inclusion if it contained at least 60 days of data in each of the four seasons
(DJF, MAM, JJA, SON), and contained a cumulative of at least electricity use of 20 kWh of
electricity.

64
64
Each residential smart meter in our dataset was associated with the street address of each
household. We matched houses in this dataset to the county building assessor datasets of the four
counties that comprised the largest proportion of our homes. In our analysis, we found 124,687
exact houses matched parcels listed in the county assessor’s data of Riverside, Los Angeles, San
Bernadino, and Orange counties. Addresses that matched to locations without residential zoning
of our dataset (~1.5%) were excluded. Accounts with more than 100,000 kWh were manually
checked to ensure these houses were metered to single households.
Our final dataset consisted of 218 million customer-days of hourly records from 159,858
customer records across 599,139 customer-years. All analysis of the electricity data of customers
was stored on USC’s Center for High-Performance Computing with a highly secure HPC Secure
Data Account, to remain in line with the security and confidentiality requirements of SCE.
Climate Zone Data
SCE encompasses many different climate zones, each with unique weather patterns that
impact electricity use. Cool climate zones include CZ 6 (coastal) and CZ 8 (inland coast). By
contrast, the two climate zones in the valley, CZ 9 (Inland Valley) and CZ 10 (Interior Valley),
are classified as moderate climate zones, with CZ 10 having summers that are classified as hot.
CZ 14 (Low Desert) and CZ 15 (High Desert) are the hottest and driest climate zones in the
region. While these climate zones are meant to encompass the unique climatological differences
of the region, they also encapsulate the differences in demography and the built environment that
each of these geographical regions entails. Through the large area spanned by our dataset, the
climate zones in our dataset also represents a variety of urban, suburban, and rural areas
featuring census tracts with a diversity of economic and racial groups.

65
65
Income Categories
Each of our customers was associated to a census tract through their address via
geocoding. We used the 2016 ACS 5-year estimates of average household incomes for each
census tract. The median census-tract average household income in our dataset was
approximately 77,000 dollars. We developed three bins of census tract based on the federal
guidelines for income levels (Low-Moderate: <80% of median income, Middle: >80% and
<120% of median income, Upper: > 120% of median income).361
Electricity Use Quintile
Each customer was also placed into a energy-use bins based on their based on their
average daily load, with Q1 representing the 20% of users with the least average load and Q5
representing the 20% of users with the highest average load. Notably, while there is a correlation
between these two demographic groups, about more than 12% of low-moderate income users are
in the highest quintile of electricity demand and more than 15% of upper-income users are in the
lowest quintile of energy-use.
4.2.2 Daily Peaking Metrics
In this paper, we characterize four metrics to measure the timing and magnitude of the
peaking behavior. These metrics are summarized in Table 1.

66
66
Table 4-1: Summary of Peaking Behavior Metrics
Metric Name Units Definition Statistics Calculated
Daily Electricity
Load
kWh Total electricity consumed in
the 24-hour record
Mean, Standard Deviation
Peak Hour Hour of
the Day
Hour of the day with the highest
electricity use.
Mean, Variance, Mode,
Mode Frequency
Absolute Peak
Height
kWh Electrical energy used during
the hour with the peak hour
Mean, Standard Deviation
Relative Peak
Height
% of
Daily
Usage
Proportion of the Daily
Electricity Load used during the
Peak Hour
Mean, Standard Deviation
*We calculate a discrete circular mean to account for the temporal nature of the peak hour. For
example, a peak at �� (1:00 am to 2:00 am) is temporally only two hours after Hour 23 (11:00 pm
to 12:00 am), it would be evaluated as a 22-hour difference in an arithmetic mean.
Each customer-day is represented by a set of 24-hourly electricity measurements, �(�, �)
for a given customer, � on day, �. (Equation 1)
�(�, �) = [�!, �", … , �#, … , �$%]
7
The Peak Hour is the hour of the day with the highest electricity use. (Equation 2)
Peak Hour (�, �) = �� (�, �) 8
Two measures for the height of the peak height are used. Absolute Peak Height measures
the electricity used in the peak hour, measured in kilowatt-hours. (Equation 3)
Absolute Peak Height (�, �) = �� (�, �) 9
Relative Peak Height measures the percent of the total daily electricity load is
used during the peak hour. (Equation 4 and Equation 5)
Daily Load (�, �) = ,�#
$%
#
10
Relative Peak Height (�, �) = max �(�, �) 11

67
67
4.2.3 Aggregate Statistics
To analyze differences in the peaking behaviors throughout various temporal and
geographic slices, we calculated the mean and standard deviation of the Peak Hour, Relative
Peak Height, and Absolute Peak Height. [See Table 4-1] For time of day, the arithmetic mean is
inaccurate due to the circular nature of a clock. Instead, we define the distance between two
hours as their shorter distance on the clock (e.g. there are 5 hours between hour 23 and hour 4,
instead of 19 hours). By calculating the mean using this definition of distance, the mean
minimizes the variance of a set of times, such as when calculating the mean Peak Hour for an
individual customer. For Peak Hour, we also calculate the Mode Peak Hour (i.e. hour of the
clock that the Peak Hour is most frequently), and the Mode Peak Hour Frequency (i.e. the
percent of records the Mode Peak Hour occurs in a given set).
We analyze two aggregations for summary statistics in our analysis – customer-day
aggregations and individual customers. Customer-day aggregations were calculated by
combining all daily records observed. We perform this statistical analysis both on the overall
dataset, but on subsets based on the type of day (i.e. weekend vs. weekday), season (i.e. Fall,
Winter, Spring, Summer), and by climate zone. For this analysis, we consider each season by the
three month periods: Winter: December, January, February (DJF), Spring: March, April, May
(MAM), Summer: June, July, August (JJA), and Fall: September, October, November (SON).
In individual customer analysis, we calculate the mean and standard deviation for each of
our metrics for each customer. In our analysis, statistics that were based on individual customers
will be prefixed by “Customer” (e.g. Customer Mean Peak Height, Customer Mode Peak Hour
Frequency, etc… ). Analysis by individual customer also allows us to hypothesis test specifically
whether each customer’s peaking metrics change significantly between season to season.

68
68
4.3 Results
In this section, we discuss the results of our analysis of peaking behaviors. We first
discuss the analysis of customer-day aggregations of peaking behaviors between different
temporal breakdowns (i.e. weekend vs. weekday, and by season) and for the climate zone subsets
of our dataset. We then discuss the individual behaviors observed in the data and the consistency
of users in our dataset.
4.3.1 Summary of Collective Peaking Behaviors
Table 2 shows a summary of average customer-day peaking behaviors by season and type
of day (i.e. weekday vs. weekend). The most common Peak Hour in our dataset was a peak at
hour 19 (19:00-20:00) with about 10% of all records peaking at this time. The mean peak time
was hour 17 (17:00-18:00). The Absolute Peak Height has high variability in our dataset, largely
due to the right-skew of very high consumption households. The mean Relative Peak Height
remains consistent at about 10% of the overall electricity throughout the year. Thus, absolute
peaks change dramatically over seasonal differences, but largely track the change in mean Daily
Load. This suggests that higher Absolute Peak Heights are largely driven by increases in
electricity load, and not “peakier” behaviors.
Peaking behaviors appear to be similar on weekdays vs weekends. Weekdays have
slightly lower Daily Loads and lower electricity Absolute Peak Heights than weekends, with the
mean Peak Hour occurring a little earlier in the afternoon on weekends. However, seasonal
differences in peaking behaviors are much more pronounced than type of day. For example,
summertime electricity Absolute Peak Heights are 51% larger than those in winter. Because of
the relatively small differences in weekday and weekend electricity peaks, the remainder of the

69
69
study will focus on weekday electricity peaks, allowing the analysis to focus on factors that have
a larger impact on peaking behaviors, like seasonal impacts and inter-customer variability.
Table 4-2 Summary of Peaking Metrics for Customer-days
n Daily Load
(kWh)
Peak Hour
(Hour of the Day)
Abs.
Height
(kWh)
Rel. Height
(% of daily
electricity
load)
Records
(107
)
Mean Std Mode Mode
Freq
Clock
Mean
Clock
Var
Mean Std Mean Std
Overall 21.8 19.8 19.7 19 10% 17 4.1 2.0 1.8 10.2% 4.2%
By type of day
Weekend 6.20 20.3 20.1 19 9% 17 4.1 2.0 1.8 10.1% 4.1%
Weekday 15.6 19.7 19.6 19 10% 18 4.1 2.0 1.8 10.2% 4.2%
By Season
DJF 5.4 17.6 17.1 18 11% 18 4.5 1.7 1.5 10.1% 4.1%
MAM 5.5 16.3 16.0 19 11% 18 4.4 1.6 1.5 10.1% 4.1%
JJA 5.5 25.7 24.0 16 9% 17 3.6 2.6 2.2 10.3% 4.3%
SON 5.5 19.8 19.3 18 10% 17 3.9 2.0 1.8 10.2% 4.2%
By Season and
Weekend
Weekend DJF 1.5 17.9 17.2 18 10% 18 4.5 1.7 1.5 9.9% 4.0%
MAM 1.6 16.6 16.1 20 10% 17 4.4 1.7 1.5 10.0% 4.1%
JJA 1.6 26.2 24.5 16 9% 16 3.7 2.6 2.2 10.2% 4.3%
SON 1.6 20.6 20.1 18 9% 17 3.9 2.1 1.9 10.1% 4.2%
Weekday DJF 3.9 17.5 17.1 18 11% 18 4.5 1.7 1.5 10.1% 4.1%
MAM 3.9 16.2 15.9 19 11% 18 4.4 1.6 1.5 10.1% 4.1%
JJA 3.9 25.6 23.8 20 9% 17 3.5 2.6 2.2 10.4% 4.4%
SON 3.9 19.5 19.0 19 10% 17 3.9 2.0 1.8 10.3% 4.2%
The mean Peak Hour in fall and summer is hour 17, which is an hour earlier in the day
than in winter or spring. While the Peak Hour variance of these two seasons is lower than that of
winter or spring, the mode Peak Hour frequency decreases. Thus, seasonal differences in Peak
Hour variance is more a reflection of less likelihood of peaking behaviors occurring in the
morning and late evening, rather than a concentration of electricity use in a few hours in the
afternoon. (Figure 1) Importantly, more frequent peaks in each hour do not necessarily correlate
with a higher peak in that hour. While most Peak Hours occur between hours 17 and 20, peaks

70
70
that occur between hour 13 and hour 16 constitute peaks with higher Absolute Peak Heights,
likely due in large part to increased air conditioning use in these hours in the summer.
Figure 4-1 The top panel shows the percent of all customer weekdays with specific peak time. The middle and bottom
panels show the mean relative peak height and mean absolute peak height for all customer weekdays by peak time. Results are
shown by season and for the full study period. Peak times are more likely to occur earlier in the day in summer months, where we
also see higher absolute, but not relative, peaks.
While summertime afternoon peaks are the most prominent, we observe some morning
Peak Hours , especially in winter and spring when loads are less driven by HVAC loads. During
weekdays in these months, customers with morning peaks (i.e. peaks between hours 5 and hour
9) consume a larger proportion of their electricity in the morning and have a slightly higher
absolute peak height, although those behaviors still consume less electricity than peak heights in
summer and fall when hot temperature drive higher cooling loads.

71
71
Differences in Peak Behaviors by Climate Zone
Figure 4-2 shows the differences in peaking behaviors between climate zones between
weekdays in summer and winter. In the winter, the peaking timing in the cooler climate zones
and hotter climate zones are relatively consistent, with both having most Peak Hours occurring in
the evening (i.e between 17 and 21). However, in the summer, the cool climate zones a plurality
of customer-day records have Peak Hours in the early evening, while the hot climate zones had a
much higher proportion of Peak Hours occur in the early evening and late afternoon. The
summertime afternoon peaks in these hotter climate zones are not only more frequent, but also
have higher Absolute Peak Heights than those of Peak Hours occurring later.
4.3.2 Peak Behaviors by Individual Customers
The distribution of weekday customers behaviors by their customer mean Absolute and
Relative Peak Heights throughout all four seasons are relatively stable. (Figure 3) The
Figure 4-2. Comparison of Peak Time, Relative Peak Height, and Absolute Peak Height between winter and summer months across climate zones
(ordered from coolest to hottest by CDDs or mean summer max temperature…). Summertime peaks, especially in hotter climate zones, occur earlier
in the day than winter peaks and are higher in magnitude, showing that peaking behaviors can be driven by temperature.

72
72
distribution of customer mean Relative Peak Heights is relatively flat throughout seasons. This
again suggests that peaking behaviors are driven more by an increase in a customer’s overall load
rather than an increase in the electricity used in one specific hour. The customer means and
customer standard deviations of Absolute Peak heights increase in summer. The lower quartile of
electricity customer peak behaviors remains relatively consistent across all months suggesting
that for some percentage of the electricity customers there is little seasonality in their peak
heights.
For each customer, we performed a Welch’s t-test to see if there was a statistical
difference in the peak height of their electricity peaks. We found that 89% of customers had a
statistically significant difference (α=0.05) in absolute peak height between summer and winter
months, with 68% of customers with statistically significant increases in peak heights in summer
and 21% with statistically significant decreases. By contrast, while 81% of customers also had
statistically significant differences in relative peak height, 41% of customers in our dataset had a
Figure 4-3 Boxplot of Weekday Customer Mean (Left) and Customer Standard Deviations (Middle) of Relative Peak
Height (Bottom) and Absolute Peak Height (Top). Outliers were excluded using the 1.5 IQR rule. The right most panel
shows the magnitude of change from DJF to JJA in relative peak heights and absolute peak heights, in increasing order,
for all customers in our dataset. While the customer-day analysis showed higher magnitude summer peaks, analysis of
individual customers reveals that some customers have lower magnitude peaks in the summer.

73
73
statistically significant increases in relative peak height and 40% having a decrease in relative
peak heights. Thus, while about a fifth of customers decrease their customer-mean Absolute Peak
Heights in the summer, the general trend is an increase in the median customer-mean Absolute
Peak Heights.
Peak Hours and Peak Hour Consistency
Weekday mean and mode Peak Hours of customers are more consistent than those
observed in customer-day aggregations of peaking behaviors. The mode Peak Hour frequency is
about 10% of weekdays in any given season. (Figure 1) By contrast, as many as 18% of
customers share mean weekday peak times and as many as 21% of customers share modal peak
times. (Figure 3) Customers themselves, on average, have their modal peak hour occur on 17%
of all days. Thus, even though customer behaviors are generally more consistent than collective
behaviors, customers generally do not peak at the same hour of every day.
In non-summer months, the mode weekday Peak Hour and mean weekday Peak Hour
occurred in the late afternoon and early evening (between hours 17 and 21) for over 60% of
customers in our dataset. During the summer, mean Peak Hours shifted earlier in the afternoon,
with over 60% of customers having mean weekday peak times between hours 14 and 18. For
those customers, the customer-mean Peak Hour variance was smaller than customers peaking at
similar times in other months. Thus, while customers were not very consistent in the specific
Peak Hour they peaked in, the average customer Peak Hour variance of summertime decreased,
implying customers were more likely to have afternoon peaks in general. Furthermore, summer
months see a higher number of customers with customer-mean Peak Hours occurring in the
afternoon, but customer-mode Peak Hour times split between early afternoon and later evening.

74
74
This split may indicate that some customers have behaviors that are more sensitive to seasonal
changes, and thus have a higher frequency of behaviors that appear in the afternoon, while others
are less sensitive and thus appear to remain peaking in the evening.
Figure 4-4 Comparison of Mean Peak Time, Peak Time Variance, Mode Peak Time, and Mode Frequency across seasons for all
customers. Peaking behaviors are less variable when examined within a customer than in the customer-day analysis that revealed
large between-customer variation. However, internal variation in peaking behaviors is still significant.
Customer differences by demographics
Figure 5 compares the key summertime and wintertime peaking metrics of customers by
the census-tract income level and usage-quintile bin. All three income groups (Low-Moderate,
Middle, and Upper) have relatively similar Absolute Peak Heights throughout the winter months.
Overall, customers in Upper-Income census tracts and customers who use more electricity seem
to have the highest summertime loads and the most consistency in their peak timing. Similarly,

75
75
larger average users, have more similar wintertime customer-mean Peak Heights than in summer
and have generally lower customer Peak Hour variances.
Figure 4-5 Boxplots of Customer Mean Absolute Peak Height, Customer Mean Relative Peak Height, and Customer Peak
Variance for distinct census tract income levels (Top) and quintiles of user electricity consumption (Bottom). Customers in
higher-income census tracts and those with higher levels of average load have similar peak magnitudes in winter, but much larger
magnitude peaks in summer months.
4.4 Discussion
The peaking behavior of customers in California is set to change with California’s aggressive
electronification efforts. The state currently leads the United States in electric vehicle
registration, with EVs constituting one-fifth of California’ new car sales in 2022.
331,332 EV
chargers increase the electricity use of a household an average of between 3 to 10 kWh on the
days vehicles are being charged, with higher power draws, between power draws of more than
1kW concentrated during specific hours.333,334 This will dramatically shift peaking behaviors,
especially in more temperate months where average power draw rates hover around 2 kW (see
Table 1).

76
76
Furthermore, the state plans to implement a ban on natural gas appliances by 2030, with
many municipalities already banning or discouraging the use of gas-fueled appliances, such as
heaters, washers, dryers, and stoves. As shown in Gunkel (2023), heat pumps have higher
variability towards peaking behavior than electric vehicles.357 While, unlike Denmark, Southern
California shows disproportionate electricity consumption in warm months, the trends towards
electrification will shift these trends diurnally and seasonally. We already see some morning
peaking rates in DJF and this trend is likely to increase as California electrifies (Figure 1).
Compared to the rest of the United States, Californian households are less likely to have
appliances that use electricity with only 28% of space heaters (vs 40% nationwide) and 21% of
water heaters (US: 47%) using electricity in 2023. 335 These appliances have the possibility of
increasing and/or shifting the small morning peak currently seen in the Spring and Winter load
profiles.
4.4.1 Direct Implications for Time of Use:
Another trend that is likely to shift the peaking behavior of customers is the implementation
of time-of-use rates in California. Time-of-use rates attempt to shift electricity use away from
hours with high demand by providing lower rates during low-demand hours and higher rates
during high-demand hours. Since 2016, a rule making effort has been put into place an effort to
analyze the electricity load of the California Independent System Operator (CAISO) and develop
analyses to develop new TOU designs,270 with a pilot study commissioned in 2018 proposing the
implementation of default TOU pricing for all residential electricity customers.271 Time-of-use
rates have been implemented as the default pricing scheme for customers in the three largest
utilities in the state (PG&E 272 and SDG&E273 since 2019 and SCE since 2021.274) However,
while there has been a concerted effort to develop such rates and understand the impact of such

77
77
pricing rates on decreasing the load at peak hours, understanding the load shape of the residential
sector remains a key gap in the literature.
Figure 4-6 The proportion of customers by best rate structure (top) for customers divided by electricity usage quintile
(left) and neighborhood income (right) assuming behaviors from 2015, 2016, and 2018, shows time of use rates are better rates
for the highest users of electricity. The pricing differences of these rates (bottom), on average are within one cent of one another.
In the SCE service area, time-of-use rates have been available since late 2018. (Figure 6)
While our data cannot show the impact that a new rate structure may have on electricity users,
we can assess the financial impact of time-of-use rates assuming past behaviors using current
rate structures. We found that the top 20% of electricity users are the only bin of electricity use to
have a majority of a customer benefit from time of use rates over traditional tiered rates, and on
average, these customers benefit from a about 1 cent less per kWh. Thus, these programs are
targeted at the biggest electricity users, who we have shown also have the highest peak demands
and have the most consistency in their electricity use such that they could implement an effective
change in their electricity behaviors.
However, these users are also more likely to live in upper-income neighborhoods, and
would be rewarded for maintaining current behaviors their behaviors are not adequately
incentivized of the new rate. Survey data of California respondents have also shown that low-

78
78
income and home-tech users more willing to shift, while those with routine lifestyles less so.
362
California households have been shown to be poor reporters of their own peaking behavior.363
Thus, even if a customer does have a financial incentive and ability to shift their electricity load,
they may not be able to correctly identify and implement that behavioral change. Furthermore,
our work found that the largest residential peak behaviors were in the hottest climate zones in the
summer. (Figure 2) One-in-ten residential electricity customers have reported keeping their
house at an unsafe temperature in order to afford an electricity bill.364 Disincentivizing air
conditioning use could have catastrophic impacts in the case of an extreme heat event.
Overall new DSM strategies, such as these implementation of time of use rates, could
exacerbate existing challenges in energy justice and accessibility.365 Furthermore, DSM
programs may require households to make costly investments in energy efficiency measures or to
adopt new technologies, which could be financially out of reach for some households. Many
households may also be unable to participate, either by physically or logistically living in
buildings without individual metering, or by having inelastic demand requirements, such as for
medical equipment that must continuously operate.366 Long term monitoring of electricity
behaviors using smart meter data will be critical to ensuring that these new strategies do not
unduly punish electricity customers who may need the most help.

79
79
Chapter 5 - Conclusion
The research summarized in this dissertation fills existing research gaps regarding the
management of and climate-change induced extreme heat events in the context of extreme events
and the corresponding electricity use during the. First, this work characterizes the key policy
decisions that occurred to manage extreme heat during the COVID-19 pandemic. Next, this work
studies the diurnal patterns of individual residential electricity use through the use of a large
smart meter dataset. Previous efforts to characterize seasonal impacts of the residential sector
have been relatively limited in number. This body of work reveals the key role that some
customers play over others. The contributions of each chapter are summarized in the following
paragraphs.
In Chapter 2, this work analyzes the strategies and interventions used to manage
compound COVID-19-extreme heat events in the 25 most populous cities of the United States.
Heat adaptation strategies employed prior to the COVID-19 pandemic were not adequate to meet
the co-occurring compound hazard of COVID-19-EHE. This comprehensive literature review
found that cities not only implemented masking and social distancing requirements to cooling
resources, but also reduced the availability of cooling resources compared to years past. This was
especially true for alternative cooling resources provided by cities, such as pools and beaches,
largely due to budget cuts.
In Chapter 3, characterizes the daily hourly load profiles of approximately 160,000
residential electricity customers across the Southern California Edison (SCE) service area during
the period spanning 2016 to 2019 across weekends, weekdays, seasons, and climate zones. We
present a methodology based upon the Gini index developed for calculating the residential load
profile from high-spatiotemporal resolution smart meter data at scale and assessing asymmetric

80
80
disparities in hourly loads between households. This work found that the top 20% of households
consumed about 40% of the daily electricity load while the bottom half of consumers used less
than 25% of the daily electricity load. This study also found that customers in mild climates
show less seasonal variability in their load profiles than those in hot climates.
In Chapter 4, I calculate the daily peaking metrics for over 200 million daily records of
residential electricity use. This work found that residential electricity peaks shift earlier into the
afternoon and were significantly higher in the summer, but that these shifts were most prominent
in hot climate zones. This analysis also found that the temporal variance in peak hours is lower in
summer. However, this study also found that the average customer only had a mode frequency of
their peak hour in less than 25% of days. This study was one of the first to characterize the
consistency of diurnal behaviors of individual customers use between days at scale.
5.1 Limitations
In Chapter 2 of this dissertation, we discussed the responses by cities to address extreme
heat management in response to COVID-19 challenges. The work presented in Chapter 2 focuses
on large municipal responses, with a specific focus on the actions of cities to provide individual
residents with the resources needed for extreme heat events. There are several limitations in
response to how cities may be prepared for future compound climate and infectious disease
events.
5.1.1 Limitations to Cooling Center Analysis
First, this work only focused on metropolitan areas and the accessibility of cooling
resources in cities. Some cities may be better adapted to climate extremes than others. The
physical infrastructure in which people live plays a large role in their vulnerability to extreme
heat events. Extreme heat is most dangerous for those who live in older buildings, on upper

81
81
floors, and for those who do not have air conditioning.239 The distribution of residential buildings
can vary by city or geographic region. Furthermore, even within the same city, the temperatures
experienced by individuals may be dramatically different depending on the urban landscape of
their surrounding area.252 Thus, there is likely a large degree of variability between not only
residents of different cities but also residents within different areas of the same city. To account
for this our analysis only explored the differences in resources available between two time
periods, however, deeper geospatial analysis is needed to tell whether these cooling resources
were truly less available.
Second, our study only focuses on the number of cooling resources that were available to
residents. This analysis did not consider accessibility, capacity, or utilization of cooling center
locations. Public cooling resources are largely targeted toward providing cooling resources for
those who may need it the most. Often, most vulnerable individuals may have no way of
accessing a cooling center if it is too far away — some estimates have shown that cooling centers
can be at 1/3 of capacity. Furthermore, there are no studies that have accessed the capacity of
cooling centers or their utilization at scale. Even if cooling resources are available, they still may
not be utilized by the population. Estimates show that only about 20% of residents use cooling
centers at any point.367 Cooling centers, like libraries, are more effective at drawing in the heat
vulnerable when they have programming and resources that are appealing.247
5.1.2 Limitations to Energy Analysis
In Chapters 3 and 4, we studied the electricity use of residential patterns based on smart
meter data.

82
82
Limitations on the construction of the load profiles
Our analysis of electricity loads considers all customer days as contributing equally to the
overall electricity curve. Thus, this curve represents the average curve expected over time.
However, each customer can have one electricity pattern on each given day. Thus, on some days
the shape of the curve may look different than the average. Our analysis does not explore
potential differences in the load shapes between different days within a season. While we did
perform robustness checks on days of high temperature or high total load to see that the overall
shape of the curve did not vary dramatically, a more robust exploration of this analysis could
potentially elucidate patterns within customers.
Limitations on analysis of peaking behavior
Our analysis of peaking behaviors is based on three metrics, absolute peak height,
relative peak height, and peak timing, intended characterize a single hour in a user’s electricity
load every day. We found high levels of variability in the individual behaviors of customers
based on this hourly electricity use. However, this metric may not directly translate into hourly
electricity use behaviors in general. First, customers may not necessarily perceive their electricity
use on an hourly level. Most customers will probably not think of an action as occurring at 4:15
pm or 3:55 pm as dramatically different times, but because of our data, these hours of use appear
as two distinct periods.
Furthermore, there are key limitations to the extent to which our analysis can be
translated into behavioral change. Customers, in our analysis, are households that could contain
multiple people. For example, even if a customer had the elasticity to shift electricity from one
period to another, that does not mean that the customer has the capability or awareness to be able
to shift that electricity use. This limitation could potentially be ameliorated by data from the

83
83
utility of the available rates, which could better elucidate the elasticity of electricity as a response
to recent changes.
5.2 Policy Recommendations and Future Work
Cooling Center Placement and Evaluation
In Chapter 2, we found that cities around the country deprioritized cooling resources
when faced with a concurrent global pandemic. Many of the closures observed in 2020 occurred
due to budget shortfalls rather than physical limitations. These budget shortfalls led to major cuts
in recreational and parks budgets, which during emergencies may seem cosmetic to customers,
but may play a larger role in providing support systems to those who are most vulnerable. This
work suggests that cities need to consider prioritizing funding programs traditionally deemed
“non-essential,” as they can play a critical role in public health during emergency events. To
support policymakers in prioritizing these resources, future work can explore methods to create
accurate assessments of cooling resources, especially regarding the spatial access and capacity of
cooling resources.
This work alludes to larger shifts in emergency response to compound extreme events.
Generally, distributed emergency response resources, such as cooling centers, vaccination sites,
and shelters, can help reduce these risks by providing essential services to affected populations.
However, the optimal allocation of these limited resources is a complex spatial decision problem
that involves multiple criteria and stakeholders. Furthermore, identifying vulnerable populations
during emergency response scenarios is often limited by the spatial resolution (i.e. data may be
collected at a neighborhood, zip-code, city, county, etc.). Models that can quickly identify
vulnerabilities and support the dynamic allocation of resources can help emergency responders

84
84
iterate through potential placements of resources quickly, calculate potential accessibility, and
estimate demand.
Analysis of Daily Loads
In Chapter 3, this work found that electricity consumption peaks during hotter months,
particularly in warmer climate zones, with high-consuming households having a notable impact
on peak load hours. By classifying households into decile bins based on their typical daily
consumption, the research identifies disparities in load contribution and highlights the
importance of understanding differences in consumption patterns for implementing effective
demand management strategies. This study provides valuable insights for shaping more equitable
and efficient energy policies. Because of the outsized influence of large electricity consumers on
the shape of the electricity curve, utilities need to put a concerted effort into shaping their
policies to target shifts in the electricity behavior of these customers.
This paper focused on the improved level of decision-making in aggregate over seasonal
period. This work found that there were large seasonal differences in electricity use, especially
with a high summertime electricity usage peak. However, this analysis utilized all customer days
to construct the residential load profile. In this way, we assume that all customer behaviors were
equally likely to create the average load profile. This assumption is most appropriate for
assessing seasonal differences in load profiles as described in Chapter 3, largely because of the
large sample sizes selected for each of these days.
Future work that explores the impact of temperature or other external stimuli should
analyze the daily load. The average load of a given day can be constructed from the residential
electricity data for that day and multiple days could then be averaged to get the expected load for

85
85
a given day. Because the size of the temporal samples is much smaller, averaging by day first
will prevent any issues of some customers potentially being over or under-represented in our
dataset.
Time-of-Use Pricing Elasticity
In Chapter 4, we explored time-of-use rate pricing assuming existing behaviors before
time-of-use rates became default electricity rates. This work found that there was limited pricing
pressure for time-of-use rates and that few to no customers had a pricing structure. However,
designing rate structures requires not only balancing incentives for short-term load shifting, but
also accounting for social equity, public health, long-term technology adoption, and/or energy
efficiency initiatives. Recent work by Borenstein (2024) has found that a large proportion of the
difference in consumption of those who would be in the top 20% of electricity users can be
explained by the number of occupants in the house, gross consumption inclusive of distributed
solar, differences in climate, electrified appliances, and the demographics of the household.368
Thus, while changes to pricing structures based on the amount of consumed electricity has the
potential to benefit the grid, pricing structures also can be regressive in nature. This highlights
the need for future work to increase its focus on the hourly elasticity of customer electricity
demand.
Our analysis was constrained by the absence of corresponding billing data to complement
the smart meter data. This limitation restricts the ability to conduct investigations into the
elasticity of electricity use. Future research endeavors to integrate these tags into the analysis. A
major benefit of the current dataset is that it encompasses almost five years of data which can
provide us with a strong counterfactual to the electricity use before a switch to time-of-use rates.
Such data will allow us to assess elasticity across different income groups, enabling a nuanced

86
86
understanding of how energy consumption behaviors vary among socioeconomic segments. This
approach is particularly pertinent for addressing equity concerns and facilitating the development
of targeted interventions that promote both efficiency and fairness in energy use.
5.2.1 Contributions of this work
Collectively, these three studies inform the ways we approach how we adapt to a
changing climate. The two studies on residential electricity developed metrics that can be
calculated quickly and at scale to characterize key features of residential electricity usage. This
body of work found that summertime electricity load profiles were higher overall, and that a
small proportion of customers consumed a disproportionate quantity of electricity. Customers
whose mean peak hour occurred in the afternoon during times, when the load was at its highest,
were also generally less variable in the timing of their peaks. This suggests that initiative to
change behaviors in the summertime will largely have to change routine behaviors. However, as
found during the COVID-19 pandemic, when cities face other extreme events, they may choose
to deprioritize the very tools that may serve best to change routine behaviors such as local
beaches, parks, and other recreational programs.

87
87
References:
1. NOAA National Centers for Environmental Information. Weather Related Fatality and Injury
Statistics. Weather.gov (2022).
2. IPCC. Summary for policymakers: Climate Change 2022: Impacts, Adaptation and Vulnerability. in
Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation:
Special Report of the Intergovernmental Panel on Climate Change vol. 9781107025 3–33 (].
Cambridge University Press, Cambridge, UK and New York, NY, USA, 2022).
3. Luber, G. & McGeehin, M. Climate Change and Extreme Heat Events. American Journal of
Preventive Medicine 35, 429–435 (2008).
4. Nik, V. M., Perera, A. T. D. & Chen, D. Towards climate resilient urban energy systems: A review.
National Science Review 8, (2021).
5. Jin, A. S. et al. Building resilience will require compromise on efficiency. Nature Energy 6, 997–
999 (2021).
6. Perkins, S. E., Alexander, L. V. & Nairn, J. R. Increasing frequency, intensity and duration of
observed global heatwaves and warm spells. Geophysical Research Letters 39, 1–5 (2012).
7. Centers for Disease Control and Prevention Centers for Disease Control and Prevention National
Center for Health Statistics. National Vital Statistics System, Mortality 2018-2021 on CDC
WONDER Online Database, released in 2021. Deaths from X30 (Exposure to excessive natural
heat), 2018-2021. http://wonder.cdc.gov/ucd-icd10-expanded.html (2022).
8. Johnson, M. G. et al. Identifying heat-related deaths by using medical examiner and vital statistics
data: Surveillance analysis and descriptive epidemiology - Oklahoma, 1990-2011. Environmental
Research 150, 30–37 (2016).
9. Wong, K. V., Paddon, A. & Jimenez, A. Review of World Urban Heat Islands: Many Linked to
Increased Mortality. JOURNAL OF ENERGY RESOURCES TECHNOLOGY-TRANSACTIONS OF
THE ASME 135, (2013).
10. Bandala, E. R. et al. Extreme heat and mortality rates in Las Vegas, Nevada: inter-annual variations
and thresholds. International Journal of Environmental Science and Technology (2019)
doi:10.1007/s13762-019-02357-9.
11. U.S. Energy Information Administration (EIA). Electricity Explained - Use of electricity. U.S.
Energy Information Administration (EIA) https://www.eia.gov/energyexplained/electricity/use-ofelectricity.php (2022).
12. Uejio, C. K. et al. Intra-urban societal vulnerability to extreme heat: The role of heat exposure and
the built environment, socioeconomics, and neighborhood stability. Health & Place 17, 498–507
(2011).
13. Meehl, G. A. & Tebaldi, C. More intense, more frequent, and longer lasting heat waves in the 21st
century. Science 305, 994–997 (2004).

88
88
14. Vaidyanathan, A., Malilay, J., Schramm, P. & Saha, S. MMWR - Heat-Related Deaths — United
States, 2004–2018. (2004).
15. Heaviside, C., Macintyre, H. & Vardoulakis, S. The Urban Heat Island: Implications for Health in a
Changing Environment. Current environmental health reports 4, 296–305 (2017).
16. Abbinett, J. et al. Heat Response Plans: Summary of Evidence and Strategies for Collaboration and
Implementation Climate and Health Technical Report Series Climate and Health Program, Centers
for Disease Control and Prevention Heat Response Plans: Summary of Evidence and Str. (2020).
17. Bernard, S. M. & McGeehin, M. A. Municipal Heat Wave Response Plans. American Journal of
Public Health 94, 1520–1522 (2004).
18. Klinenberg, E. Heat Wave: A Social Autopsy of Disaster in Chicago. (2002).
19. U.S. EPA. Excessive Heat Events Guidebook. United States Environmental Protection Agency 1–60
(2016).
20. White-Newsome, J. L. et al. Strategies to reduce the harmful effects of extreme heat events: A fourcity study. International Journal of Environmental Research and Public Health 11, 1960–1988
(2014).
21. Sheridan, S. C. A survey of public perception and response to heat warnings across four North
American cities: An evaluation of municipal effectiveness. International Journal of Biometeorology
52, 3–15 (2007).
22. Hayden, M. H. et al. Adaptive capacity to extreme heat: Results from a household survey in
Houston, Texas. Weather, Climate, and Society 9, 787–799 (2017).
23. Kalkstein, A. J. & Sheridan, S. C. The social impacts of the heat-health watch/warning system in
Phoenix, Arizona: Assessing the perceived risk and response of the public. International Journal of
Biometeorology 52, 43–55 (2007).
24. Hayden, M. H., Brenkert-Smith, H. & Wilhelmi, O. V. Differential adaptive capacity to extreme
heat: A Phoenix, Arizona, case study. Weather, Climate, and Society 3, 269–280 (2011).
25. Semenza, J. C. et al. Public perception and behavior change in relationship to hot weather and air
pollution. Environmental Research 107, 401–411 (2008).
26. Kilbourne, E. M., Choi, K. & Jones, T. S. Risk Factors for Heatstroke. JAMA 247, 3332 (1982).
27. Widerynski, S. et al. Use of cooling centers to prevent heat-related illness : summary of evidence
and strategies for implementation. (2017).
28. Hansen, A. et al. Perceptions of heat-susceptibility in older persons: Barriers to adaptation.
International Journal of Environmental Research and Public Health 8, 4714–4728 (2011).
29. Lane, K. et al. Extreme Heat Awareness and Protective Behaviors in New York City. Journal of
Urban Health: Bulletin of the New York Academy of Medicine 91,.

89
89
30. Fraser, A. M. et al. Household accessibility to heat refuges: Residential air conditioning, public
cooled space, and walkability. Environment and Planning B: Urban Analytics and City Science 44,
1036–1055 (2017).
31. Martinez, G. S., Linares, C., de’Donato, F. & Diaz, J. Protect the vulnerable from extreme heat
during the COVID-19 pandemic. Environmental Research 187, 109684 (2020).
32. Phillips, C. A. et al. Compound climate risks in the COVID-19 pandemic. Nature Climate Change
10, 586–588 (2020).
33. Hoeppe, P. Trends in weather related disasters - Consequences for insurers and society. Weather
and Climate Extremes 11, 70–79 (2016).
34. Kron, W., Löw, P. & Kundzewicz, Z. W. Changes in risk of extreme weather events in Europe.
Environmental Science & Policy 100, 74–83 (2019).
35. Bedford, J. et al. A new twenty-first century science for effective epidemic response. Nature 575,
130–136 (2019).
36. US Census Bureau. Annual Estimates of the Resident Population for Incorporated Places of 50,000
or More, Ranked by July 1, 2019 Population: April 1, 2010 to July 1, 2019. (2019).
37. National Weather Service. National Weather Service on Twitter. https://www.weather.gov/twitter
(2021).
38. Ebi, K. L., Teisberg, T. J., Kalkstein, L. S., Robinson, L. & Weiher, R. F. Heat watch/warning
systems save lives: Estimated costs and benefits for Philadelphia 1995-98. Bulletin of the American
Meteorological Society 85, 1067–1073 (2004).
39. Casanueva, A. et al. Overview of existing heat-health warning systems in Europe. International
Journal of Environmental Research and Public Health 16, (2019).
40. Olson, C. & Toprani, A. 2020 Health Alert #19: Help Prevent Heat-Related Illness and Death
Among Heat-Vulnerable at-Vulnerable People During the COVID-19 Pandemic. (2020).
41. Dangerous Heat Wave Begins in NYC. https://www.ny1.com/nyc/allboroughs/weather/2020/07/18/week-long-heat-wave-begins-in-nyc.
42. Olson, C. 2020 Health Alert #25 Heat Event July 26, 2020 to July 27 , 2020: Help Prevent HeatRelated Illness and Death.
43. Guthartz, S. 2020 Health Alert #27 Heat Event August 10, 2020 to August 12, 2020: Help Prevent
Heat-Related Illness and Death. (2020).
44. Office of The Mayor of New York City. Mayor de Blasio Announces COVID-19 Heat Wave Plan
to Protect Vulnerable New Yorkers | City of New York. https://www1.nyc.gov/office-of-themayor/news/350-20/mayor-de-blasio-covid-19-heat-wave-plan-protect-vulnerable-new-yorkers.
45. NYC Cooling Centers and Cool It! NYC Initiatives Available Tuesday to Help New Yorkers Beat
the Heat. https://www1.nyc.gov/site/em/about/press-releases/20200727_pr_nycem_cooling-centersremain-open-on-Tuesday.page.

90
90
46. Schuerman, M. NYC’s Cooling Center Network Strands Nearly A Third Of City’s Seniors -
Gothamist. Gothamist -WNYC https://gothamist.com/news/nycs-cooling-center-network-strandsnearly-a-third-of-citys-seniors (2019).
47. New York City Agencies, Residents Faced With Sweltering Sunday, First Heat Wave Of The Year
– CBS New York. https://newyork.cbslocal.com/2020/07/19/new-york-city-agencies-residentsbrace-for-sweltering-sunday-first-heat-wave-of-the-year/.
48. Get Cool NYC: Mayor de Blasio Updates New Yorkers on COVID-19 Summer Heat Plan | City of
New York. https://www1.nyc.gov/office-of-the-mayor/news/433-20/get-cool-nyc-mayor-de-blasionew-yorkers-covid-19-summer-heat-plan.
49. New York Senate. S08113 - Relates to Issuing a Moratorium on Utility Termination of Services
during Periods of Pandemics and/or State of Emergencies.
50. Fry, H. As Heat Wave Sizzles in Southern California, Out-of-towners urged to Avoid Beaches. Los
Angeles Times https://www.latimes.com/california/story/2020-04-23/out-of-town-beachgoersurged-to-stay-home-as-heat-wave-sizzles-across-southern-california.
51. Heat wave continues in Southern California, and it’s not cooling down at night – Daily News.
https://www.dailynews.com/2020/05/06/heat-wave-continues-in-southern-california-and-its-notcooling-down-at-night/.
52. Temperatures Climb by 20 Degrees During SoCal Hot Streak – NBC Los Angeles.
https://www.nbclosangeles.com/weather-news/weather-red-flag-wildfires-wind-forecast/2376641/.
53. Los Angeles is in the grips of a heat wave this weekend - Los Angeles Times.
https://www.latimes.com/california/story/2020-07-11/los-angeles-in-the-grips-of-a-heat-wave-thisweekend.
54. Lloyd, J. & De Leon, B. LA County Smashes an All-Time Heat Record Over Labor Day Weekend –
NBC Los Angeles. https://www.nbclosangeles.com/weather-news/weather-heat-wave-recordswoodland-hills-wildfires-power-outages-labor-day/2424621/.
55. Fall heat wave breaks records, prompts statewide flex alert - Los Angeles Times.
https://www.latimes.com/california/story/2020-10-01/fall-heat-wave-breaks-records-promptsstatewide-flex-alert-to-conserve-energy.
56. Davenport, F. A. Coronavirus (COVID-19) - Creating a Heat Emergency Response Plan for Los
Angeles County (Item No 23, Agenda of August 4, 2020). http://ceoiacounty.gov (2020).
57. Southern California prepares for heat wave amid Stay Home order, with or without air conditioning
- ABC7 Los Angeles. https://abc7.com/heat-wave-covid-los-angeles-coronavirus/6127436/ (2020).
58. County of Los Angeles Department of Public Health. Cooling Center Guidance During Covid-19.v3
(5-2-20) Guidance for Operating Cooling Centers During COVID-19. (2020).
59. Locations and Hours: Los Angeles Cooling Centers Are Open – NBC Los Angeles.
https://www.nbclosangeles.com/weather-news/cooling-centers-los-angeles-heat-hot-weathertemperatures/2412957/.

91
91
60. LOS ANGELES COUNTY COOLING CENTERS. www.emergency.lacity.org/heat (2020).
61. Cooling Center FAQs Cooling Centers and COVID-19.
http://publichealth.lacounty.gov/eh/climatechange/ExtremeHeat.htm.
62. Reynolds, C. Public splash pads and pools are opening in L.A., but it’s a trickle, not a torrent - Los
Angeles Times. https://www.latimes.com/travel/story/2020-07-10/la-travel-swimming-poolssouthern-california (2020).
63. LADWP and LA City Council President Nury Martinez Announce $50 Million In Emergency
Relief for Low-Income Households Impacted by COVID-19. https://www.ladwpnews.com/ladwpand-la-city-council-president-nury-martinez-announce-50-million-in-emergency-relief-for-lowincome-households-impacted-by-covid-19/.
64. Mayor Garcetti orders new restrictions on evictions, announces indefinite moratorium on water and
power shutoffs in fight against COVID-19 | Office of Los Angeles Mayor Eric Garcetti.
https://www.lamayor.org/mayor-garcetti-orders-new-restrictions-evictions-announces-indefinitemoratorium-water-and-power.
65. Heat, holiday to close L.A. coronavirus testing centers - Los Angeles Times.
https://www.latimes.com/california/story/2020-09-03/heat-holiday-to-close-l-a-coronavirus-testingcenters.
66. Chicago Weather: Heat Advisory In Effect; Temperatures Will Feel Like 105 – CBS Chicago.
https://chicago.cbslocal.com/2020/07/18/chicago-weather-heat-advisory-in-effect-temperatureswill-feel-like-105/.
67. City of Chicago. Mayor Lightfoot and OEMC Expand Cooling Resources for Residents to Beat
Extreme Heat This Week. City of Chicago
https://www.chicago.gov/city/en/depts/mayor/press_room/press_releases/2020/july/ExpandingCooli
ngResources.html.
68. City of Chicago :: Cooling Areas.
https://www.chicago.gov/city/en/depts/fss/provdrs/serv/svcs/dfss_cooling_centers.html.
69. 25 years after deadly heat wave, COVID-19 gives Chicago a new summer foe - Chicago Tribune.
https://www.chicagotribune.com/coronavirus/ct-1995-heat-wave-25th-anniversary-covid-20200711-
4kyiclqpjjhwxkp3rg4m45zyva-story.html.
70. New Program Offers ComEd Customers Additional Bill Assistance During Pandemic | ComEd - An
Exelon Company. https://www.comed.com/News/Pages/NewsReleases/2020-11-09.aspx.
71. Heat advisory for Houston area today, triple-digit ‘feels like’ temps in forecast.
https://www.houstonchronicle.com/news/houston-weather/article/Heat-advisory-for-Houston-areatoday-15326843.php.
72. ‘Y’all, it’s hot’: Heat index could hit 110 in Houston; high temps expected through Tuesday.
https://www.houstonchronicle.com/news/houston-texas/houston/article/Y-all-it-s-hot-Heat-indexcould-hit-15401491.php.

92
92
73. Heat advisory starts at 10 a.m. for Houston today. https://www.chron.com/news/houstonweather/article/Heat-advisory-starts-at-10-a-m-for-Houston-today-15521793.php.
74. City of Houston and Reliant Launch 2020 Beat the Heat Program to Help Houstonians ‘Stay Cool in
Your Home’. https://www.houstontx.gov/mayor/press/2020/beat-the-heat.html.
75. Houston cooling centers, a/c units part of Beat the Heat program | khou.com.
https://www.khou.com/article/news/health/coronavirus/beat-the-heat-cooling-centers-houstonsummer-heat-relief-program/285-813b029e-abdd-4296-8fc3-e6f1dbc8cf4e.
76. Houston Will Have 11 Cooling Centers During The Hot Summer Months – Houston Public Media.
https://www.houstonpublicmedia.org/articles/news/city-of-houston/2018/06/15/291181/houstonwill-have-11-cooling-centers-during-the-hot-summer-months/.
77. McCord, C. Houston pools closed this summer.
https://www.khou.com/article/news/health/coronavirus/houston-pools-splash-pads-will-be-closedthis-summer-due-to-covid-19-pandemic/285-b7b95bc2-25f3-4722-b945-ac2ee5fdef42.
78. Houston to swap cooling centers for A/C units amid pandemic’s summer spike.
https://www.houstonchronicle.com/news/houston-texas/houston/article/Houston-to-swap-coolingcenters-for-A-C-units-15391888.php.
79. Texas’ Moratorium On Electricity Shutoffs Comes To An End In October – CBS Dallas / Fort
Worth. https://dfw.cbslocal.com/2020/09/11/texas-moratorium-electricity-shutoffs-end/.
80. Extreme heat impacts COVID-19 testing sites, causes some to close early.
https://www.click2houston.com/news/local/2020/07/10/extreme-heat-impacts-covid-19-testingsites-causes-some-to-close-early/.
81. Delony, D. & Bennett, A. Harris County coronavirus testing closed due to hot weather. KHOU 11
https://www.khou.com/article/news/health/coronavirus/harris-county-public-health-coronavirustesting-sites-closing-early-due-to-extreme-heat/285-475e075e-3f2f-4eb5-9f85-4522a0322246
(2020).
82. NWS Phoenix. Heat Safety. National Weather Service https://www.weather.gov/psr/HeatSafety
(2020).
83. Phoenix to open convention center for homeless heat relief; more support is needed, advocates say.
https://amp.azcentral.com/story/news/local/phoenix/2020/05/29/phoenix-open-convention-centerhomeless-heat-relief-more-support-needed-advocates-say/5275105002/.
84. Excessive heat warning: Heat relief stations activated in Phoenix | 12news.com.
https://www.12news.com/article/news/heat-releif-stations/75-f8dd97ce-db58-45a0-a863-
9ee42fa5a906.
85. Heat of the Day Respite Center Opens at PHX Convention Center. City of Phoenix
https://www.phoenix.gov/newsroom/article/1287 (2020).
86. Statement about City Pools for Summer 2020. https://www.phoenix.gov/newsroom/parks-andrecreation/1227.

93
93
87. Arizona Corporation Commission bans electricity shutoffs in summer.
https://amp.azcentral.com/story/money/business/energy/2019/06/20/arizona-corporationcommission-bans-electricity-shutoffs-summer/1505457001/.
88. Wood, A. R. Philadelphia weather has heat wave on tap, 100 degrees possible. The Philadelphia
Inquirer https://www.inquirer.com/weather/heat-wave-philadelphia-deaths-weather-20200717.html
(2020).
89. Summer Arrives and Philadelphia Heat Plan Is Still Work-in-Progress – NBC10 Philadelphia.
https://www.nbcphiladelphia.com/news/local/summer-arrives-and-philly-heat-plan-is-still-work-inprogress/2443678/.
90. Philadelphia Department of Public Health Declares First Heat Health Emergency of 2020 |
Department of Public Health | City of Philadelphia. https://www.phila.gov/2020-07-20-
philadelphia-department-of-public-health-declares-first-heat-health-emergency-of-2020/.
91. Horn, T. Find Your Local Cooling Center and Protect Yourself From the Summer Heat. NBC
Philadelphia https://www.nbcphiladelphia.com/news/local/protect-yourself-from/113866/ (2019).
92. Jaramillo, C. & Meyer, K. Philadelphia opens cooling centers for heat emergency - WHYY.
https://whyy.org/articles/need-a-respite-from-the-brutal-heat-philly-offers-10-public-coolingcenters/.
93. Philadelphia pools won’t open in 2020 amid COVID-19 crisis | PhillyVoice.
https://www.phillyvoice.com/philadelphia-pools-closed-summer-2020-covid-19-budget-cutcoronavirus-parks-recreation/.
94. Utilities can now cut off non-paying customers, Pa. agency says, but there are safeguards for the
poor. https://www.inquirer.com/business/pennsylvania-pa-utilities-shutoff-ban-lifted-puc-viruscoronavirus-covid-20201008.html.
95. LIHEAP Recovery Crisis Program. https://www.dhs.pa.gov/providers/Providers/Pages/LIHEAPRecovery-Crisis-Program.aspx?mc_cid=6821fdbfc9&mc_eid=e55f922782.
96. Metro Health Issues Heat Advisory Level III and Ozone Action Day - The City of San Antonio -
Official City Website.
https://www.sanantonio.gov/gpa/News/ArtMID/24373/ArticleID/18830/Metro-Health-Issues-HeatAdvisory-Level-III-and-Ozone-Action-Day.
97. Metro Health Issues Heat Advisory - The City of San Antonio - Official City Website.
https://www.sanantonio.gov/gpa/News/ArtMID/24373/ArticleID/19155/Metro-Health-Issues-HeatAdvisory.
98. Excessive heat warning issued for San Antonio as temperatures hit 106 degrees.
https://www.mysanantonio.com/news/local/article/Excessive-heat-warning-issued-for-San-Antonioas-15403943.php.
99. NWS warns this will be the ‘hottest weekend of the year’ in San Antonio.
https://www.mysanantonio.com/news/local/article/NWS-warns-this-will-be-the-hottest-weekend-of15481104.php.

94
94
100. NWS Austin/San Antonio on Twitter. More dangerous heat across the region today. A Heat
Advisory has been posted for much of the area from 12 - 8 PM. Even areas outside the advisory will
see temperatures near 100 degrees, so there isn’t any place nec.
https://twitter.com/NWSSanAntonio/status/1299653422841503745 (2020).
101. NWS Austin/San Antonio on Twitter. (20) NWS Austin/San Antonio on Twitter: ‘A Heat Adivsory
will be in effect again on Tuesday. Peak afternoon Heat Index values will range from 108 to 114.
Stay hydrated and in air-conditioning if possible. https://t.co/t8AUCNwiec’ / Twitter.
https://twitter.com/NWSSanAntonio/status/1300644551766335491 (2020).
102. City of San Antonio Opening Cooling Centers. The City of San Antonio
https://www.sanantonio.gov/gpa/News/ArtMID/24373/ArticleID/19051/City-of-San-AntonioOpening-Cooling-Centers (2020).
103. Ibañez, D. Cooling centers to open across San Antonio. KSAT.com
https://www.ksat.com/news/local/2020/06/10/cooling-centers-to-open-across-san-antonio/.
104. Carden, A. Cooling centers open to public for heat relief | WOAI. News 4 San Antonio
https://news4sanantonio.com/news/local/cooling-centers-open-to-public-for-heat-relief (2019).
105. Pools and Splash Pads to Remain Closed - The City of San Antonio - Official City Website.
https://www.sanantonio.gov/gpa/News/ArtMID/24373/ArticleID/19184/Pools-and-Splash-Pads-toRemain-Closed.
106. Catholic Charities, City’s Department of Human Services come together for Project Cool to collect,
distribute box fans to seniors - The City of San Antonio - Official City Website.
https://www.sanantonio.gov/gpa/News/ArtMID/24373/ArticleID/19017/Catholic-CharitiesCity%E2%80%99s-Department-of-Human-Services-come-together-for-Project-Cool-to-collectdistribute-box-fans-to-seniors.
107. Neda Iranpour. Hot weather expected in San Diego County until Sunday, Cool Zones closed.
https://www.cbs8.com/article/weather/heat-wave-throughout-san-diego-county-causes-concernamid-coronavirus-outbreak/509-b08110c3-10a7-4085-998f-9452bb2519c3.
108. Heat Wave to Peak in San Diego County on Wednesday - Times of San Diego.
https://timesofsandiego.com/life/2020/05/06/heat-wave-to-peak-in-san-diego-county-onwednesday/.
109. Heat Wave To Continue Through Thursday In San Diego County | KPBS.
https://www.kpbs.org/news/2020/jun/09/heat-wave-to-continue-through-thursday-in-san/.
110. Record-High Temps to Sweep San Diego As Heat Wave Peaks – NBC 7 San Diego.
https://www.nbcsandiego.com/news/local/record-high-temps-to-sweep-san-diego-as-heat-wavepeaks/2376219/.
111. Scorching heat continues in San Diego County | FOX 5 San Diego.
https://fox5sandiego.com/weather/scorching-heat-continues-in-san-diego-county/.
112. Bravo, C. Sweltering Heat Expected for Labor Day Weekend – NBC 7 San Diego. NBC San Diego
https://www.nbcsandiego.com/news/local/sweltering-heat-expected-for-labor-dayweekend/2397033/ (2020).

95
95
113. County of San Diego. Consumer Version Excessive Heat Response Plan. www.LiveWellSD.org.
(2020).
114. Are San Diego Cool Zones Closed? | cbs8.com. https://www.cbs8.com/article/weather/forecast/sandiego-cool-zones-closed-amid-heat-coronavirus-pandemic/509-902f8395-640b-457e-a1fe39cc98508c08.
115. Cool Zones plan on the horizon, free fan program launches.
https://www.sandiegouniontribune.com/news/liveblog/coronavirus-live-updates-may-1.
116. Cool Zones. https://www.sandiegocounty.gov/hhsa/programs/ais/cool_zones/.
117. County of San Diego. Cool Zones 2020. (2020).
118. Cool Zone Sites 2019. County of San Diego (2019).
119. San Diego County to public pool visitors: not yet | San Diego Reader.
https://www.sandiegoreader.com/news/2020/sep/01/stringers-san-diego-city-pools-swimmers-notyet/.
120. As of May 4: What’s open and closed this week: Beaches, parks and trails in Southern California -
Los Angeles Times. https://www.latimes.com/travel/story/2020-05-01/whats-open-and-closed-thisweekend-beaches-parks-and-trails-in-southern-california.
121. Francisco, S. EMERGENCY AUTHORIZATION AND ORDER DIRECTING UTILITIES TO
IMPLEMENT EMERGENCY CUSTOMER PROTECTIONS TO SUPPORT CALIFORNIA
CUSTOMERS DURING THE COVID-19 PANDEMIC. https://www.sdge.com/coronavirus (2020).
122. Heat advisory issued for Dallas-Fort Worth as humidity pushes heat index values into the 100s.
https://www.dallasnews.com/news/weather/2020/06/29/high-temps-humidity-to-push-heat-indexvalues-into-the-100s-this-week-in-dallas-fort-worth/.
123. With highs in the 100s expected, heat advisory for Dallas-Fort Worth extended through Sunday
night. https://www.dallasnews.com/news/weather/2020/07/07/oppressive-summer-heat-set-fordallas-fort-worth-with-highs-in-the-triple-digits-on-the-horizon/.
124. DFW Weather: Much Of North Texas Under Excessive Heat Warning – CBS Dallas / Fort Worth.
https://dfw.cbslocal.com/2020/08/14/dfw-weather-north-texas-excessive-heat-warning/.
125. Dangerous heat expected this weekend in North Texas | wfaa.com.
https://www.wfaa.com/article/weather/dangerous-heat-this-weekend-dallas-fort-worth-august/287-
4f17300f-aa9d-49bc-95e4-7db2b220fa77.
126. Jimenez, J. Despite triple-digit heat in the forecast, Dallas has not set up cooling stations.
https://www.dallasnews.com/news/weather/2020/07/10/despite-triple-digit-heat-in-the-forecastdallas-has-not-set-up-cooling-stations/.
127. The Salvation Army debuts cooling stations as Dallas-Fort Worth hits triple-digit temperatures.
https://www.dallasnews.com/news/weather/2020/07/12/the-salvation-army-debuts-cooling-stationsas-dallas-fort-worth-hits-triple-digit-temperatures/.

96
96
128. City of Dallas - Office of Emergency Management. City of Dallas Heat Advisory Response.
https://dallascityhall.com/departments/officeemergencymanagement/Pages/Heat-Advisory-CoolingStations.aspx (2020).
129. Where to find free splash pads, spraygrounds and fountains. https://www.dallasnews.com/artsentertainment/things-to-do/2020/07/31/where-to-find-free-splash-pads-spraygrounds-and-fountains/.
130. Some Pools Close Due to Pandemic, But Other Aquatic Activities Remain Open – NBC 5 DallasFort Worth. https://www.nbcdfw.com/news/coronavirus/some-pools-close-due-to-pandemic-butsome-aquatic-activities-remain-open/2405277/.
131. Bay Area weather: Memorial Day temps reach record highs.
https://www.mercurynews.com/2020/05/25/bay-area-weather-memorial-day-temperatures-setrecord-highs/.
132. Heat wave hits Bay Area as shelter in place, curfews force people to stay indoors.
https://www.sfchronicle.com/bayarea/article/Heatwave-hits-Bay-Area-as-shelter-in-place15311810.php.
133. Bay Area heat wave: Record temps in San Jose, Oakland.
https://www.mercurynews.com/2020/08/14/bay-area-heat-wave-arrives-record-temperaturesexpected/.
134. Favro, M. Heat Wave Leaves Bay Area Scrambling for Ways to Cool Down. NBC Bay Area
https://www.nbcbayarea.com/news/local/heat-wave-leaves-bay-area-residents-scrambling-for-waysto-cool-down/2297266/ (2020).
135. San Jose Cooling Centers Open to Help Beat the Heat – NBC Bay Area.
https://www.nbcbayarea.com/news/local/san-jose-cooling-centers-open-to-help-beat-theheat/1960542/.
136. Pools to remain closed in Santa Clara County amid heat wave. San Jose Spotlight
https://sanjosespotlight.com/pools-to-remain-closed-in-santa-clara-county-amid-heat-wave/ (2020).
137. PG&E Reminds Customers of Ongoing Support Available to Help with the COVID-19 Pandemic
Impacts | PG&E.
https://www.pge.com/en/about/newsroom/newsdetails/index.page?title=20201207_pge_reminds_cu
stomers_of_ongoing_support_available_to_help_with_the_covid-19_pandemic_impacts.
138. Williams, H., Mikell, J. & Ruiz, M. Triple-digit heat and West Texas dust. KVUE
https://www.kvue.com/article/weather/austin-weather-central-texas-small-rain-chances/269-
4e13b043-ee20-4d7b-bec8-e1fc99c238e1.
139. Stay cool: Heat advisory in effect this weekend.
https://www.statesman.com/story/weather/2020/07/11/stay-cool-heat-advisory-in-effect-thisweekend/113753954/.
140. Villalpando, R. 5 things to know about Austin’s excessive heat this week. Austin AmericanStatesman https://www.statesman.com/story/news/local/2020/07/13/5-things-to-know-aboutaustinrsquos-excessive-heat-this-week/113866854/.

97
97
141. Plohetski, T. FORECAST: Heat advisory in effect Sunday. Austin American-Statesman
https://www.statesman.com/story/news/2020/08/30/forecast-heat-advisory-in-effectsunday/42347065/ (2020).
142. These Free Pools in Austin are Open Summer 2020. Austin.com https://austin.com/free-pools-inaustin/ (2020).
143. Dangerous heat continues today, tomorrow.
https://www.news4jax.com/weather/2020/09/04/dangerous-heat-continues-through-sunday/.
144. Harding, A. & Piggott, J. Jacksonville pools to reopen in phases; mayor outlines plan for summer
camps. https://www.news4jax.com/news/morning-show/2020/05/27/mayor-of-jacksonville-tomake-an-announcement-on-city-pools-summer-camps/.
145. Duval County beach restrictions lifted; zoo to reopen.
https://www.news4jax.com/news/local/2020/05/05/mayor-of-jacksonville-expected-to-announceplans-to-reopen-city/.
146. Minor, T. How to apply for Jacksonville’s emergency rental and utility assistance program. News 4
Jax https://www.news4jax.com/news/local/2021/03/26/how-to-apply-for-jacksonvilles-emergencyrental-and-utility-assistance-program/.
147. JEA: Pandemic reprieve for disconnections ends July 7.
https://www.news4jax.com/news/local/2020/06/05/jea-pandemic-reprieve-for-disconnections-endsjuly-7/.
148. Fort Worth Designates Two Community Centers as Cooling Stations – NBC 5 Dallas-Fort Worth.
https://www.nbcdfw.com/news/local/fort-worth-designates-two-community-centers-as-coolingstations/2407065/.
149. Fort Worth swimming pools will be closed amid coronavirus | Fort Worth Star-Telegram.
https://www.star-telegram.com/news/coronavirus/article242716286.html.
150. Heat advisory issued for Thursday for much of Ohio | WSYX.
https://abc6onyourside.com/news/local/heat-advisory-issued-for-thursday-for-much-of-ohio.
151. Heat Advisory issued for Ohio on Sunday, July 19th | WSYX.
https://abc6onyourside.com/news/local/heat-advisory-issued-for-ohio-on-sunday-july-19th.
152. Fenter, M. It’s going to be hot: Lack of pools, community centers add to heat-wave woes. The
Columbia Dispatch https://www.dispatch.com/story/weather/2020/07/02/itrsquos-going-to-be-hotlack-of-pools-community-centers-add-to-heat-wave-woes/112738984/ (2020).
153. Here’s when Ohio utilities will resume service shutoffs for unpaid bills - cleveland.com.
https://www.cleveland.com/open/2020/07/heres-when-ohio-utilities-will-resume-service-shutoffsfor-unpaid-bills.html.
154. Heat advisory issued for central North Carolina | myfox8.com. https://myfox8.com/news/heatadvisory-issued-for-central-north-carolina/.

98
98
155. Charlotte weather: Heat index dangers, storm in forecast | Charlotte Observer.
https://www.charlotteobserver.com/news/weather-news/article244654847.html.
156. Helping Residents Stay Cool During Extreme Heat. Charlotte Mecklenburg Emergency
Management Office https://www.mecknc.gov/news/Pages/County-Activates-Plan-to-HelpResidents-Stay-Cool-During-Extreme-Heat-.aspx (2020).
157. Looking for a place to escape the heat? Here’s some cooling stations around the area.
https://www.wbtv.com/2019/07/20/looking-place-escape-heat-heres-some-cooling-stations-aroundarea/ (2020).
158. Nicco, M. & Patel, S. Bay Area sees 110 degrees as region hits record temperatures during
scorching heat wave. ABC7 San Francisco https://abc7news.com/heat-wave-bay-area-advisorywarning-excessive-watch/6370708/.
159. Buchmann, A. & Hwang, K. Bay Area temps could hit 112 over Labor Day weekend - fire weather
watch to follow heat warning. San Francisco Chronicle
https://www.sfchronicle.com/bayarea/article/Bay-Area-temps-could-hit-113-over-Labor-Day15541695.php (2020).
160. Nicco, M. Bay Area braces for intense heat, critical fire weather as Glass Fire continues to rage in
Napa and Sonoma Counties. ABC7 news https://abc7news.com/red-flag-warning-heat-advisory-bayarea-smoke-fire/6672950/.
161. List: Bay Area Cooling Centers Open for Hot Labor Day Weekend – NBC Bay Area. NBC Bay
Area https://www.nbcbayarea.com/news/local/list-bay-area-cooling-centers-open-for-hot-labor-dayweekend/2357207/ (2020).
162. 2020 - 25 HSP Cooling Centers. San Francisco Department of Public Health (2020).
163. San Francisco Department of Emergency Management. City and County of San Francisco Cooling
Centers for Saturday September 2, 2017 | Department of Emergency Management.
https://sfdem.org/article/city-and-county-san-francisco-cooling-centers-saturday-september-2-2017
(2020).
164. San Francisco Allows Swim Team To Practice Despite COVID-19 Health Order Banning Pool Use.
CBS SF Bay Area https://sanfrancisco.cbslocal.com/2020/07/13/sf-allows-high-school-swim-teamto-practice-despite-covid-19-health-order-banning-pool-use/.
165. Bay Area Beaches Remain Open During Hot Weekend – CBS San Francisco. CBS SF Bay Area
https://sanfrancisco.cbslocal.com/2020/08/15/bay-area-beaches-remain-open-during-hot-weekend/
(2020).
166. Mack, J. L. Indianapolis weather: Extreme heat, severe storms target Central Indiana. Indianapolis
Star https://www.indystar.com/story/news/local/indianapolis/2020/07/19/indianapolis-weatherextreme-heat-severe-storms-target-central-indiana/5468154002/ (2020).
167. Sims, C. Indy Parks splash pads, parks to reopen with mask guidelines in place. Indy Star
https://www.indystar.com/story/entertainment/events/2020/07/09/indianapolis-indy-parks-reopenspraygrounds-splash-parks/5404317002/ (2020).

99
99
168. Sims, C. Indy Parks will open these 5 Indianapolis pools — with some restrictions. Indy Star
https://www.indystar.com/story/entertainment/2020/06/05/indy-parks-open-indianapolis-pools-june13-indiana-swimming/3153646001/.
169. Gibson, L. Ban on shutting off utilities extended six more weeks. Indianapolis Star
https://www.indystar.com/story/news/environment/2020/06/29/commission-extends-utility-shutoffmoratorium-denies-utilities-request-recover-lost-revenue/3278411001/ (2020).
170. Some Western Washington cities hit over 95 degrees Monday. King 5
https://www.king5.com/article/weather/weather-blog/western-washington-seattle-hot-weatherforecast/281-2b91cd79-6978-4a13-8147-ad1bd95bf0a7 (2020).
171. The Seattle area is in a ‘short-lived heat burst.’ Here’s how high temps could get and when things
may cool off. | The Seattle Times. https://www.seattletimes.com/seattle-news/weather/the-seattlearea-is-in-a-short-lived-heat-burst-heres-how-balmy-it-could-get-and-when-things-cool-off/.
172. King County Modified Phase 1 (and 2) Seattle park and recreation guidance and the reopening of
many amenities - Parkways. https://parkways.seattle.gov/2020/06/12/modified-phase-1-reopening/.
173. Morrison, K. Seattle Parks’ lifeguarded swimming areas now open - Parkways. Seattle Parks and
Recreation https://parkways.seattle.gov/2020/07/01/opening-of-lifeguarded-swim-areas-on-holddue-to-rise-in-covid-19-cases-in-king-county/.
174. PROCLAMATION BY THE GOVERNOR AMENDING AND EXTENDING EMERGENCY
PROCLAMATIONS 20-05 AND 20-23, et Seq. 20-23.11 Ratepayer Assistance and Preservation of
Essential Services.
175. Kirk, A. What health officials are saying on coronavirus, swimming pools. 9 News
https://www.9news.com/article/news/health/coronavirus/colorado-swimming-pools-coronavirus/73-
57bdd2ff-ac2d-42a6-b629-ca3745e212fa (2020).
176. Bote, T. Moratorium on utility disconnections expires; customers needing help encouraged to make
payment arrangements or apply for assistance. Colorado Public Utilities Commission
https://www.colorado.gov/moratorium-utility-disconnections-expires-customers-needing-helpencouraged-make-payment.
177. Thomas, M. Dangerous heat, humidity Monday in DC region with heat index values near 105
degrees. FOX 5 Washington DC https://www.fox5dc.com/news/dangerous-heat-humidity-mondayin-dc-region-with-heat-index-values-near-105-degrees (2020).
178. Mayor Bowser Activates Heat Emergency Plan for June 5, 2020. Government of the District of
Columbia https://mayor.dc.gov/release/mayor-bowser-activates-heat-emergency-plan-june-5-2020
(2020).
179. Schultz, K. Cooling centers DC near me: Here’s a list of whats open. WKYC
https://www.wkyc.com/article/news/local/dc/heat-advisory-dc-washington-dc-virginia-excessiveheat/65-75ee3d5f-1a70-4d39-a5de-c807ba9e28fc (2020).
180. Extreme heat closes coronavirus testing centers in DC | wusa9.com.
https://www.wusa9.com/article/news/local/dc/heat-emergency-coronavirus-cooling-centers-dcmontgomery-county/65-8b2227e6-f6be-4e9b-8010-8c63970866c7.

100
100
181. Department of Energy & Environment. Receive Assistance With Your Utility Bills (LIHEAP).
District of Columbia https://doee.dc.gov/liheap.
182. Commission Updates in Response to the COVID-19 Public Health Emergency. District of Columbia
Public Service Commission https://dcpsc.org/Coronavirus.aspx (2020).
183. Williams, M. Heat advisory issued for much of Massachusetts for Sunday, Monday - masslive.com.
MassLive https://www.masslive.com/weather/2020/07/heat-advisory-issued-for-much-ofmassachusetts-for-sunday-monday.html (2020).
184. Epstein, D. Forecast: Extreme Heat Through Tuesday. WBUR
https://www.wbur.org/news/2020/07/27/forecast-extreme-heat-through-tuesday (2020).
185. Heat advisory in effect for Boston, Worcester Co., most of eastern Mass. WCVB
https://www.wcvb.com/article/heat-advisory-issued-for-boston-most-of-easternmassachusetts/33557241 (2020).
186. PUBLIC HEALTH COMMISSION. Mayor Walsh declares heat emergency, opens 20 BCYF
community centers as cooling centers. City of Boston https://www.boston.gov/news/mayor-walshdeclares-heat-emergency-opens-20-bcyf-community-centers-cooling-centers (2020).
187. Boston to open cooling centers to deal with expected heat wave. Boston 25 News
https://www.boston25news.com/news/local/boston-open-cooling-centers-deal-with-expected-heatwave/FTRLPOIBCFBWPN44W2ZWMAUVEI/ (2020).
188. Sobey, R. Beating the heat: Residents escape to Boston cooling centers. Boston Herald
https://www.bostonherald.com/2019/07/21/beating-the-heat-residents-escape-to-boston-coolingcenters/ (2019).
189. Slane, K. Massachusetts beaches can reopen next week. Here are the rules you’ll need to follow if
you go. Boston.com https://www.boston.com/culture/coronavirus/2020/05/18/massachusettsbeaches-reopen-rules-coronavirus (2020).
190. Gilvarg, C. Baker-Polito Administration Extends Utility Shutoff Moratorium | Mass.gov. Press
Release https://www.mass.gov/news/baker-polito-administration-extends-utility-shutoff-moratorium
(2020).
191. Weather Authority Alert: Heat Advisory today, rain/ storm chances return | KTSM 9 News.
https://www.ktsm.com/weather/weather-authority-alert-heat-advisory-today-rain-storm-chancesreturn/.
192. Miner, K. Weather Warn Day: Dangerous heat is on the way. 4 News
https://cbs4local.com/news/local/dangerous-heat-is-on-the-way (2020).
193. Parker, J. El Paso soars to 109 again as Excessive Heat Warning is in effect for 1st time in 25 years
- KVIA. KVIA https://kvia.com/weather/weather-news/2020/07/10/el-paso-soars-to-109-again-withexcessive-heat-warning-until-sunday/ (2020).
194. Parker, J. El Paso hits 109 degrees to blow past old high temp record during heat advisory. KVIA
https://kvia.com/weather/weather-news/2020/07/30/el-paso-hits-109-degrees-to-blow-past-old-hightemp-record-during-heat-advisory/ (2020).

101
101
195. Borunda, D. El Paso, Juárez in heat advisory as temperatures top 100 degrees again. El Paso Times
https://www.elpasotimes.com/story/news/local/2020/08/11/el-paso-juarez-under-heat-advisorytemps-top-100-again/3344484001/ (2020).
196. Heat Advisory August 20-21 – El Paso Office of Emergency Management. El Paso City-County
Office of Emergency Management https://elpasoready.org/2020/08/21/heat-advisory-august-20-21/
(2020).
197. El Paso cooling centers extends hours of operation to help residents face extreme heat. KFOX 14
https://kfoxtv.com/news/local/el-paso-opens-cooling-centers-to-help-residents-face-extreme-heat
(2020).
198. Bock, H. El Pasoans head outdoors despite extreme heat. KFOX 14
https://kfoxtv.com/news/local/el-pasoans-head-outdoors-despite-extreme-heat (2020).
199. Extreme Weather Task Force launches 2020 Summer Safety Campaign, Fan Drive. El Paso Herald
Post https://elpasoheraldpost.com/extreme-weather-task-force-launches-2020-summer-safetycampaign-fan-drive/ (2020).
200. NWSNashville. After hitting 98 DEGREES yesterday, the NEW EDITION of the forecast would
suggest we’re not in for just 5 SECONDS OF SUMMER. If you’re one of the NEW KIDS ON THE
BLOCK, TAKE THAT to heart by getting yourself NSYNC with a heat index over 1-0-THE
JACKSON 5 and drink lots of water. Tweet
https://twitter.com/NWSNashville/status/1284805066147328001 (2020).
201. Metro Parks’ Phase 2 Re-openings. Nashville https://www.nashville.gov/News-Media/NewsArticle/ID/9801/Metro-Parks-Phase-2-Reopenings.aspx (2020).
202. Jorge, K. Wave Country’s wave pool reopening this weekend in Nashville. FOX 17 Nashville
https://fox17.com/news/local/metro-parks-wave-country-reopening-this-weekend (2020).
203. TPUC Adopts Lifting Service Disconnection Motion. Tennessee Public Utility Commission
https://www.tn.gov/tpuc/news/2020/8/10/tpuc-adopts-lifting-service-disconnection-motion.html
(2020).
204. City of Detroit safely opens recreation centers to provide relief from extreme heat and protection
against COVID-19 | City of Detroit. City of Detroit https://detroitmi.gov/news/city-detroit-safelyopens-recreation-centers-provide-relief-extreme-heat-and-protection-against-COVID-19 (2020).
205. LIST: Cooling Centers Open In Detroit During Extreme Heat. CBS Detroit
https://detroit.cbslocal.com/2019/07/18/list-cooling-centers-open-in-detroit-during-extreme-heat/
(2020).
206. Hall, C. All public pools in metro Detroit, Detroit to be closed indefinitely. Detroit Free Press
https://www.freep.com/story/news/local/michigan/2020/05/28/public-pools-metro-detroit-detroitclosed/5277240002/ (2020).
207. Hall, C. Michigan pools, splash pads closed, beaches open Memorial Day weekend. Detroit Free
Press https://www.freep.com/story/news/local/michigan/2020/05/22/pool-splash-pad-beachmichigan-coronavirus/5241085002/ (2020).

102
102
208. Rahman, N. Michigan families vulnerable as economic safeguards expire. Detroit Free Press
https://www.freep.com/story/news/local/michigan/2020/08/11/economic-safeguards-expire-uistimulus-checks-covid/3297793001/ (2020).
209. Sweetman, C. Oklahoma pools, aquatic centers reopening amid pandemic | KFOR.com Oklahoma
City. KFOR https://kfor.com/news/oklahoma-pools-aquatic-centers-reopening-amid-pandemic/
(2020).
210. Oklahoma Gas & Electric. OG&E to Resume Service-Disconnects Sept. 6. Press Release
https://ogeenergy.gcs-web.com/news-releases/news-release-details/oge-resume-service-disconnectssept-6 (2020).
211. Denn, J. PSC Approves Plan to Help Con Edison’s Low-Income Customers Emergency Summer
2020 Cooling Bill Relief Program Provides Bill Credits to Low-Income Bill Discount Program
Customers in NYC and Westchester County Financial Support Totaling Up to $70.56 Million.
(2020).
212. Jaramillo, C. Philadelphia Parks and Rec expands Playstreets with cooling kits -WHYY. WHYY
https://whyy.org/articles/100-extremely-hot-philly-blocks-are-getting-super-soakers-and-other-toysto-beat-the-heat/ (2020).
213. Granberry, M., Norimine, H. & Jaramilo, C. Dallas furloughs 472 city employees, cutting heavily
into Park & Rec, libraries and the arts. The Dallas Morning News (2020).
214. Narciso, D. Columbus library board approves staff furloughs. The Columbus Dispatch (2020).
215. Fernandez, R. San Jose, Menlo Park and Palo Alto plan service cuts as Covid-19 impacts municipal
budgets. Silicon Valley Business Journal
https://www.bizjournals.com/sanjose/news/2020/05/28/coronavirus-roundup-covid-san-jose-budgetcuts.html.
216. LEFRAK, M. D.C. Opens 14 Cooling Centers Around The City : NPR. NPR (2020).
217. KVIA ABC-7. El Paso operating cooling centers for triple-digit heat wave - KVIA. KVIA (2020).
218. Allnutt, B. Heatwave plus COVID-19 equal a strain on city resources. Planet Detroit (2020).
219. Alpert, E. L.A. suffered deadly heat, yet chairs sat empty at its cooling centers. Los Angeles Times
https://www.latimes.com/california/story/2020-09-19/la-deadly-heat-empty-cooling-centers (2020).
220. Levin, S. & Ho, V. Thousands of people pack California beaches despite coronavirus concerns. The
Guardian https://www.theguardian.com/us-news/2020/apr/27/california-beaches-coronavirusorange-county.
221. Reimann, N. Maskless Parties And Crowded Beaches Across U.S. As Coronavirus Spikes Over
Holiday Weekend. Forbes https://www.forbes.com/sites/nicholasreimann/2020/07/05/masklessparties-and-crowded-beaches-across-us-as-coronavirus-spikes-over-holidayweekend/?sh=4e4119046997.
222. Quansah, R. et al. Household Energy Insecurity and COVID-19 Have Independent and Synergistic
Health Effects on Vulnerable Populations. Public Health | www.frontiersin.org 1, 609608 (2021).

103
103
223. Memmott, T., Carley, S., Graff, M. & Konisky, D. M. Sociodemographic disparities in energy
insecurity among low-income households before and during the COVID-19 pandemic. Nature
Energy 6, 186–193 (2021).
224. Berry, C. RECS: One in three U.S. households faced challenges in paying energy bills in 2015.
Energy Information Administration
https://www.eia.gov/consumption/residential/reports/2015/energybills/ (2015).
225. Graff, M. & Carley, S. COVID-19 assistance needs to target energy insecurity. Nature Energy 5,
352–354 (2020).
226. Health and Human Services. LIHEAP DCL 2020-10 CARES Act Supplemental Funding Release
FFY20. (2020).
227. Frank, D. A. et al. Heat or eat: The Low Income Home Energy Assistance Program and nutritional
and health risks among children less than 3 years of age. Pediatrics 118, e1293–e1302 (2006).
228. Coronavirus Aid, Relief, and Economic Security Act. (116th Congress, 2020).
229. Voosen, P. Global temperatures in 2020 tied record highs. Science 371, 334–335 (2021).
230. Fox, L. 2020 AZ Heat Season Recap. Arizona Department of Health Services
https://www.youtube.com/watch?v=DssqMUYDS7w (2020).
231. Oster, A. M., Caruso, E., DeVies, J., Hartnett, K. P. & Boehmer, T. K. Transmission Dynamics by
Age Group in COVID-19 Hotspot Counties — United States, April–September 2020. MMWR.
Morbidity and Mortality Weekly Report 69, 1494–1496 (2020).
232. Liu, Z. et al. Near-real-time monitoring of global CO2 emissions reveals the effects of the COVID19 pandemic. Nature Communications 11, 1–12 (2020).
233. Kaplan, S. Heat wave, wildfires and drought show deadly realities of climate change - The
Washington Post. New York Times https://www.washingtonpost.com/climateenvironment/2021/07/03/climate-change-heat-dome-death/ (2021).
234. Silberner, J. Heat wave causes hundreds of deaths and hospitalisations in Pacific north west. BMJ
374, n1696 (2021).
235. Wilhelmi, O. V., Howe, P. D., Hayden, M. H. & O’Lenick, C. R. Compounding hazards and
intersecting vulnerabilities: experiences and responses to extreme heat during COVID-19.
Environmental Research Letters 16, 084060 (2021).
236. Serna, J., II, R.-G. L. & Miller, L. Bay Area fires fueled by dry lightning and extreme heat. Los
Angeles Times https://www.latimes.com/california/story/2020-08-19/destructive-bay-area-firesfueled-by-rare-mix-of-intense-dry-lightning-and-extreme-heat (2020).
237. Money, L. & Serna, J. Massive August fire is now the largest in California history - Los Angeles
Times. Los Angeles Times https://www.latimes.com/california/story/2020-09-10/massive-augustfire-now-largest-in-california-history-at-471-000-acres-and-counting (2020).

104
104
238. Smith, H. August Complex fire burns record-breaking 1 million acres - Los Angeles Times. Los
Angeles Times https://www.latimes.com/california/story/2020-10-05/august-complex-fire-burnsrecord-breaking-1-million-acres (2020).
239. McKeever, A., Morris, R. & Jacobs, B. The West Coast had the world’s most polluted cities in
September. National Geographic https://www.nationalgeographic.com/science/article/west-coasthad-worlds-most-polluted-cities-this-month (2020).
240. Browning, K., Cowan, J., Fortin, J., Fountain, H. & Tompkins, L. Scope of California Wildfires Is
Staggering. New York Times https://www.nytimes.com/2020/08/24/us/california-fires.html (2020).
241. Graff, A. & De Guzman, D. North Bay fires rage, Angwin and St. Helena Hospital ordered to
evacuate. SFGATE https://www.sfgate.com/california-wildfires/article/LNU-Lightning-ComlpexNapa-County-wildfire-15494914.php (2020).
242. Samenow, J. & Cusick, A. Excessive heat warning in areas ravaged by Hurricane Laura, power
outages - The Washington Post. The Washington Post
https://www.washingtonpost.com/weather/2020/08/31/hurricane-laura-heat-power-outages/.
243. Haines, M. How Louisiana Medical Team Protected COVID-19 Patients as Hurricane Laura Roared
In | Voice of America - English. https://www.voanews.com/usa/how-louisiana-medical-teamprotected-covid-19-patients-hurricane-laura-roared.
244. Cardona, S., Felipe, N., Fischer, K., Sehgal, N. J. & Schwartz, B. E. Vaccination Disparity:
Quantifying Racial Inequity in COVID-19 Vaccine Administration in Maryland. Journal of Urban
Health 2021 98:4 98, 464–468 (2021).
245. Centers for Disease Control and Prevention. April 11, 2020 Interim Guidance | COVID-19 and
Cooling Centers. https://www.cdc.gov/coronavirus/2019-ncov/php/cooling-center.html (2020).
246. Gray, K. These Wi-Fi Buses Bring School to Students for Remote Learning. The New York Times
https://www.nytimes.com/interactive/2020/12/17/us/school-bus-remote-learning-wifi.html (2020).
247. Klinenberg, E. Palaces for the People: How Social Infrastructure Can Help Fight Inequality,
Polarization, and the Decline of Civic Life. (2018).
248. Ebi, K. L. & Semenza, J. C. Community-Based Adaptation to the Health Impacts of Climate
Change. American Journal of Preventive Medicine 35, 501–507 (2008).
249. Chen, M., Ban-Weiss, G. A. & Sanders, K. T. The role of household level electricity data in
improving estimates of the impacts of climate on building electricity use. Energy and Buildings 180,
146–158 (2018).
250. How many smart meters are installed in the United States, and who has them? U.S. Energy
Information Administration https://www.eia.gov/tools/faqs/faq.php?id=108&t=3 (2020).
251. Buckee, C. O. et al. Aggregated mobility data could help fight COVID-19. Science 368, 145–146
(2020).

105
105
252. Vargo, J., Stone, B., Habeeb, D., Liu, P. & Russell, A. The social and spatial distribution of
temperature-related health impacts from urban heat island reduction policies. Environmental
Science & Policy 66, 366–374 (2016).
253. Vahmani, P., Sun, F., Hall, A. & Ban-Weiss, G. Investigating the climate impacts of urbanization
and the potential for cool roofs to counter future climate change in Southern California.
ENVIRONMENTAL RESEARCH LETTERS 11, (2016).
254. Haines, A., Kovats, R., Campbell-Lendrum, D. & Corvalan, C. Climate change and human health:
impacts, vulnerability, and mitigation. The Lancet 367, 2101–2109 (2006).
255. International Energy Agency. Renewables 2022. https://www.iea.org/reports/renewables2022/executive-summary (2022).
256. US Energy Information Administration. Annual Energy Outlook 2023.
https://www.eia.gov/outlooks/aeo/pdf/AEO2023_Narrative.pdf.
257. Kroposki, B. et al. Achieving a 100% Renewable Grid: Operating Electric Power Systems with
Extremely High Levels of Variable Renewable Energy. IEEE Power and Energy Magazine 15, 61–
73 (2017).
258. Bird, L., Milligan, M. & Lew, D. Integrating Variable Renewable Energy: Challenges and
Solutions. https://www.osti.gov/biblio/1097911 (2013) doi:10.2172/1097911.
259. California Independent System Operator. Understanding the ISO.
https://www.caiso.com/about/Pages/OurBusiness/Default.aspx.
260. California Independent System Operator. 2022 Annual Report on Market Issues and Performance.
http://www.caiso.com/Documents/2022-Annual-Report-on-Market-Issues-and-Performance-Jul-11-
2023.pdf (2023).
261. CAISO. What the duck curve tells us about managing a green grid.
https://www.caiso.com/Documents/FlexibleResourcesHelpRenewables_FastFacts.pdf (2016).
262. Denholm, P., O’Connell, M., Brinkman, G. & Jorgenson, J. Overgeneration from Solar Energy in
California. A Field Guide to the Duck Chart. https://www.osti.gov/biblio/1226167 (2015)
doi:10.2172/1226167.
263. As solar capacity grows, duck curves are getting deeper in California.
https://www.eia.gov/todayinenergy/detail.php?id=56880.
264. Balasubramanian, S. & Balachandra, P. Effectiveness of demand response in achieving supplydemand matching in a renewables dominated electricity system: A modelling approach. Renewable
and Sustainable Energy Reviews 147, 111245 (2021).
265. Gelazanskas, L. & Gamage, K. A. A. Demand side management in smart grid: A review and
proposals for future direction. Sustainable Cities and Society 11, 22–30 (2014).
266. Reports on Demand Response and Advanced Metering | Federal Energy Regulatory Commission.
https://www.ferc.gov/power-sales-and-markets/demand-response/reports-demand-response-andadvanced-metering.

106
106
267. Albadi, M. H. & El-Saadany, E. F. Demand Response in Electricity Markets: An Overview. in 2007
IEEE Power Engineering Society General Meeting 1–5 (2007). doi:10.1109/PES.2007.385728.
268. Lee, M. State OKs higher electric rates for most consumers. San Diego Union-Tribune (2015).
269. Herter, K. & Wayland, S. Residential response to critical-peak pricing of electricity: California
evidence. Energy 35, 1561–1567 (2010).
270. Decision Adopting Policy Guidelines to Assess Time Periods for Future Time-of-Use Rates and
Energy Resource Contract Payment. Cal. PUC Lexis 335 P.U.R.4th 337 vol. 2017 19* (2017).
271. George, S., Bell, E., Savage, A. & Messer, B. California Statewide Opt-in Time-of-Use Pricing
Pilot. https://www.cpuc.ca.gov/-/media/cpuc-website/files/legacyfiles/o/6442453144-opt-in-toureport.pdf.
272. Brown, T. K. PG&E is changing how your electricity bill works — here’s everything you need to
know. San Francisco Chronicle
https://www.sfchronicle.com/local/environment/article/MicroClimates-PG-E-is-changing-howyour-16037469.php (2021).
273. SDG&E rolling out ‘time of use’ rates next month and what it means for you. San Diego UnionTribune https://www.sandiegouniontribune.com/business/energy-green/sd-fi-sdge-timeofuse20190221-story.html (2019).
274. Column: How you pay for electricity is changing. Here’s what you need to know. Los Angeles
Times https://www.latimes.com/business/story/2021-09-24/column-edison-energy-rate-switch
(2021).
275. Linder, K. F., Dickson, C. T. & Inglis, M. R. Hourly Electric-Load Model (HELM). Volume 1.
Design, Development, and Demonstration. Final Report. https://www.osti.gov/biblio/5947823
(1985).
276. Baroiant, S., Barnes, J., Chapman, D., Keates, S. & Phung, J. California Investor-Owned Utility
Electricity Load Shapes.
277. Siano, P. Demand response and smart grids—A survey. Renewable and Sustainable Energy Reviews
30, 461–478 (2014).
278. Gastwirth, J. L. The Estimation of the Lorenz Curve and Gini Index. The Review of Economics and
Statistics 54, 306–316 (1972).
279. Beccali, M., Cellura, M., Lo Brano, V. & Marvuglia, A. Forecasting daily urban electric load
profiles using artificial neural networks. Energy Conversion and Management 45, 2879–2900
(2004).
280. Hippert, H. S., Bunn, D. W. & Souza, R. C. Large neural networks for electricity load forecasting:
Are they overfitted? International Journal of Forecasting 21, 425–434 (2005).
281. Sandels, C., Widén, J. & Nordström, L. Forecasting household consumer electricity load profiles
with a combined physical and behavioral approach. Applied Energy 131, 267–278 (2014).

107
107
282. Lindberg, K. B., Seljom, P., Madsen, H., Fischer, D. & Korpås, M. Long-term electricity load
forecasting: Current and future trends. Utilities Policy 58, 102–119 (2019).
283. Phurailatpam, C., Rajpurohit, B. S. & Wang, L. Planning and optimization of autonomous DC
microgrids for rural and urban applications in India. Renewable and Sustainable Energy Reviews 82,
194–204 (2018).
284. Iria, J., Heleno, M. & Cardoso, G. Optimal sizing and placement of energy storage systems and onload tap changer transformers in distribution networks. Applied Energy 250, 1147–1157 (2019).
285. Hajiaghapour-Moghimi, M., Azimi-Hosseini, K., Hajipour, E. & Vakilian, M. Residential Load
Clustering Contribution to Accurate Distribution Transformer Sizing. in 2019 International Power
System Conference (PSC) 313–319 (2019). doi:10.1109/PSC49016.2019.9081518.
286. Chen, C. S., Hwang, J. C. & Huang, C. W. Application of load survey systems to proper tariff
design. IEEE Transactions on Power Systems 12, 1746–1751 (1997).
287. Ramos, S., Vale, Z., Santana, J. & Rodrigues, F. USE OF DATA MINING TECHNIQUES TO
CHARACTERIZE MV CONSUMERS AND TO SUPPORT THE CONSUMER- SUPPLIER
RELATIONSHIP.
288. Mahmoudi-Kohan, N., Moghaddam, M. P. & Sheikh-El-Eslami, M. K. An annual framework for
clustering-based pricing for an electricity retailer. Electric Power Systems Research 80, 1042–1048
(2010).
289. Panapakidis, I. P., Alexiadis, M. C. & Papagiannis, G. K. Load profiling in the deregulated
electricity markets: A review of the applications. in 2012 9th International Conference on the
European Energy Market 1–8 (2012). doi:10.1109/EEM.2012.6254762.
290. US Energy Information Administration. Electric Power Annual 2021. (2022).
291. Kavousian, A., Rajagopal, R. & Fischer, M. Determinants of residential electricity consumption:
Using smart meter data to examine the effect of climate, building characteristics, appliance stock,
and occupants’ behavior. Energy 55, 184–194 (2013).
292. Socolow, R. H. The twin rivers program on energy conservation in housing: Highlights and
conclusions. Energy and Buildings 1, 207–242 (1978).
293. Swan, L. G. & Ugursal, V. I. Modeling of end-use energy consumption in the residential sector: A
review of modeling techniques. Renewable and Sustainable Energy Reviews 13, 1819–1835 (2009).
294. Brandenburg, L. et al. Load Research Manual. Volume 2. Fundamentals of Implementing Load
Research Procedures. ANL/SPG--13(Vol.2), 6705684 http://www.osti.gov/servlets/purl/6705684/
(1980) doi:10.2172/6705684.
295. Proedrou, E. A Comprehensive Review of Residential Electricity Load Profile Models. IEEE Access
9, 12114–12133 (2021).
296. Grandjean, A., Adnot, J. & Binet, G. A review and an analysis of the residential electric load curve
models. Renewable and Sustainable Energy Reviews 16, 6539–6565 (2012).

108
108
297. Kang, X., An, J. & Yan, D. A systematic review of building electricity use profile models. Energy
and Buildings 281, 112753 (2023).
298. Parkash, B., Lie, T. T., Li, W. & Tito, S. R. Hierarchical Structure Based Energy Consumption
Forecasting in Top-Down Approach. Proceedings - 2022 7th Asia Conference on Power and
Electrical Engineering, ACPEE 2022 1732–1737 (2022) doi:10.1109/ACPEE53904.2022.9784036.
299. Fischer, D., Surmann, A. & Lindberg, K. B. Impact of emerging technologies on the electricity load
profile of residential areas. Energy and Buildings 208, 109614 (2020).
300. Parti, M. & Parti, C. The Total and Appliance-Specific Conditional Demand for Electricity in the
Household Sector. The Bell Journal of Economics 11, 309–321 (1980).
301. Aigner, D. J., Sorooshian, C. & Kerwin, P. Conditional Demand Analysis for Estimating Residential
End-Use Load Profiles. EJ 5, (1984).
302. Kavgic, M. et al. A review of bottom-up building stock models for energy consumption in the
residential sector. Building and Environment 45, 1683–1697 (2010).
303. Capasso, A., Grattieri, W., Lamedica, R. & Prudenzi, A. A bottom-up approach to residential load
modeling. IEEE Transactions on Power Systems 9, 957–964 (1994).
304. US Department of Energy. Advanced Metering Infrastructure and Customer Systems.
https://www.energy.gov/sites/prod/files/2016/12/f34/AMI%20Summary%20Report_09-26-16.pdf
(2016).
305. Every, J., Li, L. & Dorrell, D. Leveraging smart meter data for economic optimization of residential
photovoltaics under existing tariff structures and incentive schemes. APPLIED ENERGY 201, 158–
173 (2017).
306. Yildiz, B., Bilbao, J. I., Dore, J. & Sproul, A. B. Recent advances in the analysis of residential
electricity consumption and applications of smart meter data. Applied Energy 208, 402–427 (2017).
307. Wang, Y., Chen, Q., Hong, T. & Kang, C. Review of Smart Meter Data Analytics: Applications,
Methodologies, and Challenges. IEEE Transactions on Smart Grid 10, 3125–3148 (2019).
308. Mukai, T. et al. Smart meter-based home energy report: Design and operation of automatic
generation system of hourly usage indication message. ELECTRONICS AND COMMUNICATIONS
IN JAPAN 104, 74–86 (2021).
309. Sobrino, E., Santiago, A., Gonzalez, A., & IEEE. Forecasting the Electricity Hourly Consumption
of Residential Consumers with Smart Meters using Machine Learning Algorithms. in (2019).
310. Andersen, F., Gunkel, P., Jacobsen, H. & Kitzing, L. Residential electricity consumption and
household characteristics: An econometric analysis of Danish smart-meter data. ENERGY
ECONOMICS 100, (2021).
311. MacMackin, N., Miller, L. & Carriveau, R. Modeling and disaggregating hourly effects of weather
on sectoral electricity demand. ENERGY 188, (2019).

109
109
312. Sanchez-Lopez, M. et al. The diverse impacts of COVID-19 on electricity demand: The case of
Chile. INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS 138,
(2022).
313. Novan, K. & Smith, A. The Incentive to Overinvest in Energy Efficiency: Evidence from Hourly
Smart-Meter Data. JOURNAL OF THE ASSOCIATION OF ENVIRONMENTAL AND RESOURCE
ECONOMISTS 5, 577–605 (2018).
314. Ushakova, A. & Mikhaylov, S. Big data to the rescue? Challenges in analysing granular household
electricity consumption in the United Kingdom. ENERGY RESEARCH & SOCIAL SCIENCE 64,
(2020).
315. Chen, M., Sanders, K. T. & Ban-Weiss, G. A. A new method utilizing smart meter data for
identifying the existence of air conditioning in residential homes. Environmental Research Letters
vol. 14 (2019).
316. Kwac, J., Flora, J. & Rajagopal, R. Household Energy Consumption Segmentation Using Hourly
Data. IEEE TRANSACTIONS ON SMART GRID 5, 420–430 (2014).
317. Zethmayr, J. & Makhija, R. S. Six unique load shapes: A segmentation analysis of Illinois
residential electricity consumers. The Electricity Journal 32, 106643 (2019).
318. Gunkel, P. A., Klinge Jacobsen, H., Bergaentzlé, C.-M., Scheller, F. & Møller Andersen, F.
Variability in electricity consumption by category of consumer: The impact on electricity load
profiles. International Journal of Electrical Power & Energy Systems 147, 108852 (2023).
319. Gastwirth, J. L. A General Definition of the Lorenz Curve. Econometrica 39, 1037–1039 (1971).
320. Hernández, D. & Bird, S. Energy Burden and the Need for Integrated Low-Income Housing and
Energy Policy. Poverty & Public Policy 2, 5–25 (2010).
321. Mayes, S., Zhang, T. & Sanders, K. T. Residential precooling on a high-solar grid: impacts on CO2
emissions, peak period demand, and electricity costs across California. Environ. Res.: Energy 1,
015001 (2023).
322. Liddle, B. & Huntington, H. How prices, income, and weather shape household electricity demand
in high-income and middle-income countries. Energy Economics 95, 104995 (2021).
323. Chen, C., Xu, X., Adua, L., Briggs, M. & Nelson, H. Exploring the factors that influence energy use
intensity across low-, middle-, and high-income households in the United States. Energy Policy 168,
113071 (2022).
324. Alberini, A., Gans, W. & Velez-Lopez, D. Residential consumption of gas and electricity in the
U.S.: The role of prices and income. Energy Economics 33, 870–881 (2011).
325. U.S. Energy Information Administration. U.S. energy insecure households were billed more for
energy than other households -. https://www.eia.gov/todayinenergy/detail.php?id=56640.
326. Nadeau, C. A. MIT Living Wage Calculator. (2023).

110
110
327. Xu, X. & Chen, C. Energy efficiency and energy justice for U.S. low-income households: An
analysis of multifaceted challenges and potential. Energy Policy 128, 763–774 (2019).
328. Chen, C., Xu, X. & Day, J. K. Thermal comfort or money saving? Exploring intentions to conserve
energy among low-income households in the United States. Energy Research & Social Science 26,
61–71 (2017).
329. Brown, M. A., Soni, A., Lapsa, M. V., Southworth, K. & Cox, M. High energy burden and lowincome energy affordability: conclusions from a literature review. Prog. Energy 2, 042003 (2020).
330. Casey, J. A. et al. Trends from 2008–2018 in electricity-dependent durable medical equipment
rentals and sociodemographic disparities. Epidemiology 32, 327–335 (2021).
331. Bloomberg, D. H. Electric cars now make up a fifth of California’s auto sales. Los Angeles Times
https://www.latimes.com/business/story/2023-11-01/electric-cars-now-make-up-a-fifth-ofcalifornias-auto-sales (2023).
332. California leads the United States in electric vehicles and charging locations.
https://www.eia.gov/todayinenergy/detail.php?id=61082.
333. Muratori, M. Impact of uncoordinated plug-in electric vehicle charging on residential power
demand. Nat Energy 3, 193–201 (2018).
334. Burlig, F., Bushnell, J., Rapson, D. & Wolfram, C. Low Energy: Estimating Electric Vehicle
Electricity Use. AEA Papers and Proceedings 111, 430–435 (2021).
335. Davis, L. Three Facts about Electric Heating in California. Energy Institute Blog
https://energyathaas.wordpress.com/2023/05/08/three-facts-about-electric-heating-in-california/
(2023).
336. Chen, M., Ban-Weiss, G. A. & Sanders, K. T. Utilizing smart-meter data to project impacts of urban
warming on residential electricity use for vulnerable populations in Southern California.
Environmental Research Letters vol. 15 (2020).
337. Yee, G. Temperature records broken across Los Angeles area at end of spring heat wave. Los
Angeles Times https://www.latimes.com/california/story/2022-04-08/record-high-temperatures-losangeles-heat-wave (2022).
338. Wilson, E. J. H. et al. End-Use Load Profiles for the U.S. Building Stock: Methodology and Results
of Model Calibration, Validation, and Uncertainty Quantification.
https://www.osti.gov/biblio/1854582 (2022) doi:10.2172/1854582.
339. Firth, S., Lomas, K., Wright, A. & Wall, R. Identifying trends in the use of domestic appliances
from household electricity consumption measurements. Energy and Buildings 40, 926–936 (2008).
340. Rajabi, A. et al. A comparative study of clustering techniques for electrical load pattern
segmentation. Renewable and Sustainable Energy Reviews 120, 109628 (2020).
341. Schirmer, P. A. & Mporas, I. Non-Intrusive Load Monitoring: A Review. IEEE Transactions on
Smart Grid 14, 769–784 (2023).

111
111
342. Abubakar, I., Khalid, S. N., Mustafa, M. W., Shareef, H. & Mustapha, M. An overview of Nonintrusive load monitoring methodologies. in 2015 IEEE Conference on Energy Conversion
(CENCON) 54–59 (2015). doi:10.1109/CENCON.2015.7409513.
343. Gold, R. Status Report on Electrification Policy: Where to Next? Curr Sustainable Renewable
Energy Rep 8, 114–122 (2021).
344. Clean Energy, Jobs, and Affordability Act of 2022. SB-1020 (2022).
345. Jones-Albertus, B. Confronting the Duck Curve: How to Address Over-Generation of Solar Energy.
Energy.gov https://www.energy.gov/eere/articles/confronting-duck-curve-how-address-overgeneration-solar-energy (2017).
346. Gelazanskas, L. & Gamage, K. A. A. Demand side management in smart grid: A review and
proposals for future direction. Sustainable Cities and Society 11, 22–30 (2014).
347. Guo, Z. et al. Residential electricity consumption behavior: Influencing factors, related theories and
intervention strategies. RENEWABLE & SUSTAINABLE ENERGY REVIEWS 81, 399–412 (2018).
348. Sanquist, T. F., Orr, H., Shui, B. & Bittner, A. C. Lifestyle factors in U.S. residential electricity
consumption. Energy Policy 42, 354–364 (2012).
349. Winett, R. A., Neale, M. S. & Grier, H. C. Effects of Self-Monitoring and Feedback on Residential
Electricity Consumption. Journal of Applied Behavior Analysis 12, 173–184 (1979).
350. Faruqui, A. & Malko, J. R. The residential demand for electricity by time-of-use: A survey of
twelve experiments with peak load pricing. Energy 8, 781–795 (1983).
351. Wolak, F. A. Residential Customer Response to Real-time Pricing: The Anaheim Critical Peak
Pricing Experiment. (2007).
352. Chen, J. et al. Stochastic simulation of occupant-driven energy use in a bottom-up residential
building stock model. APPLIED ENERGY 325, (2022).
353. Alonso, A., Nogales, F. & Ruiz, C. Hierarchical Clustering for Smart Meter Electricity Loads Based
on Quantile Autocovariances. IEEE TRANSACTIONS ON SMART GRID 11, 4522–4530 (2020).
354. Cerquitelli, T. et al. Clustering-Based Assessment of Residential Consumers from Hourly-Metered
Data. in (2018).
355. Rhodes, J., Cole, W., Upshaw, C., Edgar, T. & Webber, M. Clustering analysis of residential
electricity demand profiles. APPLIED ENERGY 135, 461–471 (2014).
356. Kohlenberg, R., Phillips, T. & Proctor, W. A Behavioral Analysis of Peaking in Residential
Electrical-Energy Consumers1. Journal of Applied Behavior Analysis 9, 13–18 (1976).
357. Andreas Gunkel, P., Klinge Jacobsen, H., Bergaentzlé, C.-M., Scheller, F. & Møller Andersen, F.
Variability in electricity consumption by category of consumer: The impact on electricity load
profiles. International Journal of Electrical Power & Energy Systems 147, 108852 (2023).
358. Shove, E. Beyond the ABC: Climate Change Policy and Theories of Social Change. Environ Plan A
42, 1273–1285 (2010).

112
112
359. Strengers, Y. Peak electricity demand and social practice theories: reframing the role of change
agents in the energy sector. in The Global Challenge of Encouraging Sustainable Living 18–42
(Edward Elgar Publishing, 2013).
360. Powells, G., Bulkeley, H., Bell, S. & Judson, E. Peak electricity demand and the flexibility of
everyday life. Geoforum 55, 43–52 (2014).
361. California Department of Housing and Community Development. Income Limits.
https://www.hcd.ca.gov/grants-and-funding/income-limits.
362. Muttaqee, M. et al. Time of use pricing and likelihood of shifting energy activities, strategies, and
timing. Energy Policy 187, 114019 (2024).
363. Zanocco, C. et al. Assessing Californians’ awareness of their daily electricity use patterns. Nat
Energy 7, 1191–1199 (2022).
364. In 2020, 27% of U.S. households had difficulty meeting their energy needs - U.S. Energy
Information Administration (EIA). https://www.eia.gov/todayinenergy/detail.php?id=51979.
365. Calver, P. & Simcock, N. Demand response and energy justice: A critical overview of ethical risks
and opportunities within digital, decentralised, and decarbonised futures. Energy Policy 151,
112198 (2021).
366. White, L. V. & Sintov, N. D. Varied health and financial impacts of time-of-use energy rates across
sociodemographic groups raise equity concerns. Nat Energy 5, 16–17 (2020).
367. Derakhshan, S. et al. Smartphone locations reveal patterns of cooling center use as a heat mitigation
strategy. Applied Geography 150, 102821 (2023).
368. Borenstein, S. Energy Hogs and Energy Angels: What Does Residential Electricity Usage Really
Tell Us About Profligate Consumption? Working Paper at https://doi.org/10.3386/w32023 (2024).

113
113
Appendix
Appendix A Appendix - Supplemental Information for Chapter 2
A.1 Geographical distribution of Southern California Edison
dataset.
Figure A-1 Plot of customers included in the SCE dataset. Users are counted by 1 km by 1 km grid. Grid cells with less
than 5 households are suppressed in our analysis.
A.2 Data Filtering
A.2.1 Exclusion of 2020 Data
Data from 2020 was filtered out of our analysis. The average residential load
profile of our dataset was notably higher in 2020 for Spring (MAMJ) and the Hottest
Months (JASO) compared to similar seasons in our analysis. Figure shows the
summertime load profile by year. The weekday load profiles of 2020, especially in
MAMJ and JASO are notably higher during the mid-day, likely due to higher proportions
of work-from-home and school-from-home households during this time period.

114
114
Figure A-2 Weekday load profiles by season and year. NB: Winter NDJF includes the first two and last two months of
the year
A.2.2 Exclusion of Daylight Savings Rollback Days
In Southern California, daylight savings time is observed. Data recorded by SCE records
was based on the clock hour in our dataset. Thus, on days where an hour is gained (i.e. spring
forward), all user records were excluded because they were missing the 2 am to 3 am record. On
days, where fall back occurs, user days are included, but the 2am to 3am period represents both
the first and second 2am to 3am time periods experienced by users. All other days are recorded
based on the clock hour that are experienced by users.
A.2.3 Overview of Filtering Process
Table A-1 - Summary of Filtering Process. Data in each column represents the number of users, user-years, and userdays of data that remained after each given
Data Remaining
Users
UserYears User-Days
Initial Data 194,251 718,906 237,870,135
Removal of Non-zero User-Day records 194,058 718,112 237,171,041
Remove User-years if less than 200 days in that year 170,819 642,214 230,324,437
Remove User-year if cumulative electricity use <20
kwh 170,817 642,194 230,318,443
Manual Residential Removal 170,779 642,044 230,264,604
Remove 2020 Data 170,768 535,892 191,956,533
Merging with Climate Zone and Census Tract Data 169,487 533,136 191,016,682

115
115
Weather Data Filtering* 163,403 514,370 184,409,741
Final Dataset 163,403 514,370 184,409,741
A.3 Validation of Residential Zoning
Data provided by SCE was marked as residential by the utility. We performed an analysis
of houses using building assessor data to ensure that residential houses were included in the data.
We matched 146,400 houses to the parcels listed in the county assessor’s data of Riverside, Los
Angeles, San Bernadino, and Orange counties. Of these matched houses, only about 1.5% (2164)
were not explicitly zoned in residential parcels.
We also manually performed a google search on the areas near the residential houses of
the 75 largest annual consumers (i.e. houses with more than 100,000 kWh of electricity used in a
given hour). To preserve privacy, a random house number was chosen within the same 1000
house-numbers on the same street of the house, and then navigated through Google Streetview to
ensure that they were residential. There was a total of 75 total customers whose electricity was
more than 100,000 kWh of electricity. Of these customers, we found that 38 were either nonresidential or constituted multiple households metered at one site (e.g. trailer parks). In houses
with greater than 150,000 kWh of electricity usage 29 were considered non-residential with only
6 true single residence meters.
A.4 Breakdown of User Deciles
Table A-2 Distribution of borders of each electricity use decile bin. Data represent annually averaged daily
consumption.
Decile 1 Less than 6.648 kWh
Decile 2 (6.648, 9.172] kWh
Decile 3 (9.172, 11.412] kWh
Decile 4 (11.412, 13.644] kWh
Decile 5 (13.644, 16.035] kWh
Decile 6 (16.035, 18.694 ] kWh

116
116
Decile 7 (18.694, 21.877 ] kWh
Decile 8 (21.877, 26.197] kWh
Decile 9 (26.197, 33.416] kWh
Decile 10 Greater than 33.416 kWh
A.5 Summary of average load profiles by season and day-type
Table A-3 Summary statistics for the residential load profile by season and weekday/weekend. Peak hour represents
the hour of the load profile with the highest amount of electricity used, and minimum of electricity
Season Weekend
Peak
Hour
Peak
Height
(kWh)
Min
Hour
Min
Height
(kwh) Total Load (KW)
JASO Weekday 17 1.47 4 0.58 23.8
JASO Weekend 16 1.55 4 0.57 24.9
MAMJ Weekday 19 0.96 3 0.47 17.0
MAMJ Weekend 19 0.90 4 0.46 17.4
NDJF Weekday 19 0.98 3 0.52 17.0
NDJF Weekend 18 0.96 4 0.52 17.5
A.6 Lorenz Curve
Each customer is ordered based on the average daily electricity use of each customer.
Figure 3 shows the Lorenz curve for total daily electricity usage, as well as electricity consumed
between the 4pm to 9pm period. Electricity records are aligned based on their ranked typical
daily usage.

117
117
Figure A-3 Lorenz curve of electricity use for the daily load (left) and system peak 4pm to 9pm period (right). The
solid dark-blue line represents the Lorenz-curve. The dashed line represents a line of perfect equality, where each user uses the
same amount of electricity. The GINI index is the proportion of the area under the line of perfect equality that is contained by the
area between the line of perfect equality and the Lorenz curve.

118
118
A.7 Summary of average load profiles by user decile and season
Table A-4 Summary statistics of load profiles by usage decile and season on weekdays. The contribution columns
show the hour where that decile is contributing most or least to the overall curve and when those periods occur. The usage shows
the hour with the highest and lowest electricity use.
Season Deciles Contribution Usage
Daily Load
Max Min Max Min
Hour Percent Hour Percent Hour Value (kWh) Hour Value (kWh)
NDJF 0% to 10% 22 3% 11 2% 19 0.29 3 0.15 4.7
10% to 20% 21 5% 11 4% 19 0.46 3 0.23 7.7
20% to 30% 21 6% 11 5% 19 0.59 3 0.30 9.7
30% to 40% 21 7% 11 6% 19 0.70 3 0.35 11.5
40% to 50% 21 8% 11 7% 19 0.80 3 0.41 13.4
50% to 60% 20 10% 11 8% 19 0.91 3 0.47 15.3
60% to 70% 19 11% 11 10% 19 1.03 3 0.53 17.4
70% to 80% 16 12% 11 12% 19 1.18 3 0.61 20.3
80% to 90% 11 16% 21 14% 19 1.39 3 0.74 24.6
90% to 100% 11 29% 21 23% 19 2.22 3 1.26 41.3
JASO 0% to 10% 5 3% 14 2% 20 0.31 3 0.16 5.3
10% to 20% 6 5% 13 3% 19 0.54 3 0.26 9.0
20% to 30% 6 6% 12 4% 19 0.72 3 0.33 11.9
30% to 40% 6 7% 11 6% 17 0.91 4 0.39 14.8
40% to 50% 6 8% 11 7% 17 1.12 4 0.45 17.8
50% to 60% 16 10% 10 8% 17 1.37 4 0.52 21.3
60% to 70% 16 11% 9 10% 16 1.63 4 0.60 25.2
70% to 80% 15 14% 3 12% 16 1.95 4 0.69 30.0
80% to 90% 13 17% 4 15% 16 2.36 4 0.84 37.0
90% to 100% 10 29% 20 24% 17 3.50 4 1.43 59.5
MAMJ 0% to 10% 6 3% 13 2% 20 0.26 3 0.13 4.2
10% to 20% 6 5% 13 3% 20 0.42 3 0.20 6.9
20% to 30% 6 6% 13 5% 20 0.54 3 0.26 8.9
30% to 40% 6 7% 12 6% 19 0.64 3 0.31 10.7
40% to 50% 6 8% 12 7% 19 0.75 3 0.36 12.7
50% to 60% 6 9% 11 8% 19 0.87 3 0.41 14.7
60% to 70% 17 11% 11 10% 19 1.01 3 0.48 17.2
70% to 80% 16 13% 9 12% 19 1.18 3 0.55 20.5
80% to 90% 13 17% 6 15% 19 1.42 3 0.68 25.7
90% to 100% 11 31% 6 24% 13 2.40 3 1.18 44.2

119
119
A.8 Summary of Climate Zone Load Profiles by Decile
A.9 Usage decile breakpoint in different climate zones
Figure A-4 – Breakpoints of user decile bins by climate zones, where each red-dot represents the breakpoint in
electricity usage by the average daily electricity of users. For example, an electricity customer with an annual-daily average
electricity consumption of 30 kWh per day in Climate Zone 15 would be classified as CZ15-Decile 7 while a customer with the
same annual-daily average electricity consumption in Climate Zone 8 would be classified as CZ8-Decile 10.

Abstract (if available)

Abstract As climate-induced extreme events, such as extreme heat, continue to grow in frequency, duration, and intensity, managing these challenges is becoming increasingly difficult. First, this dissertation analyzes the challenges of managing compound climate-extreme events, by exploring municipal responses to compound global pandemic and extreme heat events during the 2020 COVID-19 pandemic. This body of work then explores electricity consumption behaviors in Southern California by analyzing a dataset of smart meter electricity records for approximately 200,000 homes in the greater Los Angeles area over a five-year span. This dissertation develops frameworks to construct of data-driven hourly load profiles of the residential electricity sector and analyzes individual peaking behaviors of residential customers. The results of this body of work can be used by grid operators, utilities, and policymakers, to better design electrification policies, rate structures, and heat action plans to protect communities from the impacts of climate change while preparing for future energy needs.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Residential electricity demand in the context of urban warming: leveraging high resolution smart meter data to quantify spatial and temporal patterns…

PDF

Investigating the role of climate in affecting residential electricity consumption through high spatiotemporal resolution observations

PDF

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

PDF

Beyond greenhouse gases and towards urban-scale climate mitigation: understanding the roles of black carbon aerosols and the urban heat island effect as local to regional radiative forcing agents

PDF

From tugboats to trees: investigating the coupled systems of urban air pollution and meteorology

PDF

Building occupancy modeling and occupancy-loads relationships for building heating/cooling energy efficiency

PDF

Evaluating the role of energy system decarbonization and land cover properties on urban air quality in southern California

PDF

Point cloud data fusion of RGB and thermal information for advanced building envelope modeling in support of energy audits for large districts

PDF

Thermally driven water treatment with membrane distillation: membrane performance, waste heat integration, and cooling analysis

PDF

Developing high-resolution spatiotemporal methods to model and quantify water use for energy

PDF

Integrating systems of desalination and potable reuse: reduced energy consumption for increased water supply

PDF

Examining exposure to extreme heat and air pollution and its effects on all-cause, cardiovascular, and respiratory mortality in California: effect modification by the social deprivation index

PDF

Impact of virtual reality (VR)-based training on construction robotics remote-operation

PDF

Analysis of sources and profiles of organic carbon in ambient particulate matter across fine and coarse sizes and introducing an optical technique for real-time urban dust measurement

PDF

Intelligent adaptive automation: activity-driven and user-centered building automation

PDF

Identifying and mitigating the effects of urban heat islands in California

PDF

Behavioral form finding using multi-agent systems: a computational methodology for combining generative design with environmental and structural analysis in architectural design

PDF

Robust and adaptive online reinforcement learning

PDF

Water desalination: real-time membrane characterization for performance prediction and system analysis for energetic enhancement

PDF

Towards generalized event understanding in text via generative models

Asset Metadata

Creator Jin, Andrew Shida (author)

Core Title Evaluating energy consuming behaviors and the sufficiency of urban systems in the context of extreme heat hazards

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Civil Engineering

Degree Conferral Date 2024-05

Publication Date 05/22/2024

Defense Date 04/04/2024

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

Tag Energy,extreme heat,OAI-PMH Harvest,residential electricity

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sanders, Kelly (committee chair), Becerik, Burçin (committee member), Linkov, Igor (committee member), Rose, Adam (committee member)

Creator Email asjin@usc.edu,me@andrewjin.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113954477

Unique identifier UC113954477

Identifier etd-JinAndrewS-13011.pdf (filename)

Legacy Identifier etd-JinAndrewS-13011

Document Type Dissertation

Format theses (aat)

Rights Jin, Andrew Shida

Internet Media Type application/pdf

Type texts

Source 20240522-usctheses-batch-1160 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

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

Repository Email cisadmin@lib.usc.edu