University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Developing Frameworks to Quantify the Operational and Environmental
Performance of Energy Systems Within the Context of Climate Change
by
Measrainsey Meng
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Engineering (Environmental Engineering))
August 2020
Copyright 2020 Measrainsey Meng
Acknowledgements
The author acknowledges the nancial support of the National Science Foundation Graduate
Research Fellowships Program.
I would like to thank the following people, without whom I would not have been able to
complete my doctoral degree:
My advisor, Kelly Sanders, who sparked a light in me ve years ago when I felt as though I
had no direction. Thank you for taking a chance on me, for guiding me, and for your patience
throughout the process.
My committee members, George Ban-Weiss and Mohammed Beshir, whose knowledge and
insight steered me through my research.
My research group members, whose jokes, support, and supply of desserts allowed me to have
a PhD experience that was actually enjoyable.
My friends, who kept me grounded and reminded me of my humanity when I needed it most.
Finally, I would like to express my deepest gratitude to my family: my mom, my dad, and my
older sister. Thank you for always pushing me, believing in me, and cheering me on.
ii
Table of Contents
Acknowledgements ii
List Of Tables v
List Of Figures vi
Abstract viii
Chapter 1: Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Structure of document and resulting publications to date . . . . . . . . . . . . . . 4
Chapter 2: A data-driven approach to investigating the impacts of air tempera-
ture on the eciencies of coal and natural gas generators 6
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Preparing operational datasets for the regression analysis . . . . . . . . . . 9
2.2.2 Preparing climate datasets for the regression analysis . . . . . . . . . . . . 15
2.2.3 Regression analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.1 General implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.2 Data limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4.3 Implications for future work . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Chapter 3: Spatially allocating lifecycle water use for US coal-red electricity
across producers, generators, and consumers 34
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.1 Water for coal energy produced in mining provinces . . . . . . . . . . . . . 39
3.2.2 Water for coal-red electricity generated in balancing authorities . . . . . . 42
3.2.3 Water for coal-red electricity consumed in balancing authorities . . . . . . 43
3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3.1 Uncertainties and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3.2 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
iii
Chapter 4: Integrating water, energy, and climate modeling to assess vulnerabil-
ities to the US Southwest power grid 53
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.1 Study Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.2 Reservoir Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.3 Modeling system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.1 Model simulations of the climate and water systems . . . . . . . . 60
4.2.3.2 Model simulations of electric system . . . . . . . . . . . . . . . . . 63
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3.1 San Juan River Basin thermoelectric generation . . . . . . . . . . . . . . . 64
4.3.2 San Juan River Basin hydroelectric generation . . . . . . . . . . . . . . . . 68
4.3.3 WECC-wide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.1 Impacts of climate change and water constraints on basin-level electricity
generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.2 Impacts of local electricity generation changes on grid-wide electricity gen-
eration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Chapter 5: Conclusion 75
Reference List 79
Appendix A
Supplemental Information for Chapter 2: A data-driven approach to investigating the
impacts of air temperature on the eciencies of coal and natural gas generators . . 87
A.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A.1.1 Data ltering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A.1.2 Final set of units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
A.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Appendix B
Supplemental Information for Chapter 3: Spatially allocating lifecycle water use for US
coal-red electricity across producers, generators, and consumers . . . . . . . . . . 94
B.1 Computational Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
B.2 Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Appendix C
Supplemental Information for Chapter 4: Integrating water, energy, and climate model-
ing to assess vulnerabilities to the US Southwest power grid . . . . . . . . . . . . . 102
C.1 San Juan River Basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
C.2 WECC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
iv
List Of Tables
3.1 Water withdrawn and water consumed at each life cycle stage for coal-red elec-
tricity consumption in the US in 2017 . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1 List of power plants in San Juan River Basin . . . . . . . . . . . . . . . . . . . . . 58
4.2 Scenarios studied in analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
A.1 Lower and upper limits placed on heat rates, based on the EPA's assumptions for
the Platform model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A.2 Results of change in eciency per 1

C increase in ambient air temperature are
dierentiated by cooling type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
B.1 Water withdrawn and water consumed at each life cycle stage for coal-red elec-
tricity consumption in the US in 2017 . . . . . . . . . . . . . . . . . . . . . . . . . 95
B.2 Water withdrawn and water consumed at each life cycle stage for coal-red elec-
tricity consumption in the US in 2017 . . . . . . . . . . . . . . . . . . . . . . . . . 95
B.3 Balancing authority codes and corresponding names. . . . . . . . . . . . . . . . . . 96
B.4 Water consumption and withdrawal intensities for cooling coal power plants, by
balancing authority . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
v
List Of Figures
1.1 Diagram conceptually visualizing linkages between energy, water, and climate systems 3
1.2 Diagram describing chapters in dissertation . . . . . . . . . . . . . . . . . . . . . . 5
2.1 (a) Hourly heat rates for natural gas combined-cycle electricity generating units
are plotted and categorized by duct burner. (b) Hourly heat rates for natural gas
combined-cycle electricity generating units are plotted and categorized our pro-
posed classication methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 The nal number of electricity generating units included in the analysis are charac-
terized by (top) cooling type, (middle) fuel type and prime mover technology, and
(bottom) climate zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Regression results for generating unit eciency change per 1

C increase in tem-
perature (=T ), characterized by cooling system type . . . . . . . . . . . . . . . 23
2.4 Regression results for generating unit eciency change per 1

C increase in tem-
perature (=T ), characterized by cooling system type and climate zone . . . . . 24
3.1 Methodology schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2 Fuel and water consumption by mining province . . . . . . . . . . . . . . . . . . . 45
3.3 Flows of energy, water consumption, and water withdrawal from mining province
to balancing authority . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.4 Fuel and water consumption by balancing authority . . . . . . . . . . . . . . . . . 48
3.5 Coal-red electricity and water withdrawals exchanged between balancing authority 49
4.1 Map of San Juan River basin and power plants . . . . . . . . . . . . . . . . . . . . 57
4.2 Daily generation for thermoelectric and hydroelectric power plants in the San Juan
River Basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Annual generation of power plants in San Juan River Basin and annual costs (nor-
malized and absolute) for thermal power plants . . . . . . . . . . . . . . . . . . . . 67
vi
4.4 Dierence in annual generation and generation cost in WECC from base scenario . 70
4.5 Annual generation in WECC balancing authorities for all scenarios . . . . . . . . . 71
A.1 Generating units analyzed for each year are mapped and characterized by distance
to nearest NOAA weather station . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
A.2 Generating units analyzed for each year are mapped and characterized by fuel,
prime mover, and cooling system type . . . . . . . . . . . . . . . . . . . . . . . . . 89
A.3 Regression results for generating unit eciency change per 1

C increase in tem-
perature (=T ), characterized by fuel, prime mover, and cooling system type . 91
A.4 Residual standard error (RSE) values from regression models plotted and charac-
terized by cooling type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
A.5 Regression results for generating unit eciency change per 1

C increase in tem-
perature (=T ), characterized by cooling system type and nameplate capacity . 93
A.6 Regression results for generating unit eciency change per 1

C increase in temper-
ature (=T ), characterized by cooling system type and generating unit operation
year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
B.1 Electricity generated, water consumed, and water withdrawn within each balancing
authority . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
B.2 Net generation, cooling water consumed, and cooling water withdrawn for electric-
ity generated and electricity delivered . . . . . . . . . . . . . . . . . . . . . . . . . 101
C.1 Annual water withdrawals and consumption . . . . . . . . . . . . . . . . . . . . . . 102
C.2 Annual CO
2
, NO
x
, and SO
2
emissions in San Juan River Basin . . . . . . . . . . . 103
C.3 Annual generation in balancing authorities. . . . . . . . . . . . . . . . . . . . . . . 104
C.4 Annual cost in balancing authorities. . . . . . . . . . . . . . . . . . . . . . . . . . . 105
C.5 Annual CO
2
emissions in balancing authorities. . . . . . . . . . . . . . . . . . . . . 106
C.6 Annual NO
x
emissions in balancing authorities. . . . . . . . . . . . . . . . . . . . . 107
C.7 Annual SO
2
emissions in balancing authorities. . . . . . . . . . . . . . . . . . . . . 108
C.8 Dierence in annual WECC emissions . . . . . . . . . . . . . . . . . . . . . . . . . 109
vii
Abstract
The United States (US) energy system is responsible for numerous environmental and climate
impacts throughout the various lifecycle stages. The processes of extracting fossil fuels and con-
verting fossil fuels into electric power emit greenhouse gases that lead to the warming of the
climate. Additionally, both primary fuel production and electric power production require the
withdrawal and consumption of water resources. Changes in water resources and the climate, such
as reduced water availability, increased water temperatures, and increased ambient temperatures,
can constrain energy production. The complex interdependencies between climate, water, and
energy systems are important to understand, especially since climate change is expected to exac-
erbate the tensions between these systems. This body of work develops empirical and integrated
modeling frameworks to understand the operations and environmental impacts of energy systems
under climate change and climate variability. The ndings emphasize the importance of under-
standing the spatio-temporal contexts of environmental impacts, particularly water consumption,
incurred throughout the dierent lifecycle stages of energy consumption. There are decouplings
between where energy is produced, where energy is consumed, and where environmental impacts
occur. Another emphasis in this body of work is that changes in electricity generation in one
local area has cascading impacts on other areas in the interconnected grid. As the electricity grid
continues to experience various transitions, the vulnerabilities of the electricity grid are going to
change as well.
viii
Chapter 1
Introduction
1.1 Motivation
The United States (US) energy system { including primary energy extraction, processing, rening,
and conversion to secondary forms { is responsible for a number of climate and environmental
impacts throughout each of these lifecycle steps. The US electricity sector alone is the source
of 28% of the country's greenhouse gas emissions [1]. Thermoelectric power generators, which
produce a little over 80% of all electricity generation, are responsible for 50% of sulfur dioxide
(SO
2
) emissions, and 10% of nitrogen oxide (NO
x
) emissions nationally [2]. The US energy system
is also dependent on water resources. In 2014, the US commercial energy system consumed an
estimated 1:6 10
10
m
3
of water, approximately 10% of total US water consumption that year
[63]. Water needed for cooling thermal power plants represents 40% of US freshwater withdrawals
and 3% of U.S. freshwater consumption [3]. Conventional hydroelectric generation, which makes
up approximately 6% of US electricity generation, also requires large quantities of water to spin
turbines [4].
Fluctuations in ambient temperature, reductions in water availability, and increases in water
temperature can impact the performance and reliability of energy systems. For power plants that
require water for cooling, the lack of cooling water can force a curtailment of power production
1
or a complete shutdown. High intake water temperatures can also reduce power plant operating
eciencies, reduce maximum generation capacity, and force a generator to shut down. After water
is withdrawn and used for cooling, water discharged back to a body of water that exceeds certain
temperatures can cause curtailment of power or plant shutdown in order to comply with water
quality standards and environmental laws. Thermoelectric generators that do not use wet cooling
systems are more impacted by ambient air temperatures. For example, dry cooled-generator
eciency losses are generally around 2%-3% in moderate climates but can increase to 10% in
areas with high ambient air temperatures [5]. For gas-red generators, high inlet air temperature
decreases generation eciency and capacity [6]. The linkages between energy, water, and climate
systems are visualized in Figure 1.1. Note that the linkages shown are only the ones relevant to
this body of work and are not meant to be comprehensive of all interdependencies between energy,
climate, and water systems.
2
CLIMATE
ENERGY
WATER
Burning fossil fuels for electricity
releases greenhouse gas emissions.
Increased ambient temperatures reduce the
efficiencies of thermal power plants.
Extraction and processing of primary fuel,
cooling thermal power plants, and spinning
turbines for hydropower need water.
Reduced water availability and increased water
temperature constrain power production.
Climate change affects the hydrological cycle.
Figure 1.1: Diagram conceptually visualizing linkages between energy, water, and climate systems.
Note that not all linkages between the three systems are shown in the diagram; only the linkages
that are studied in the works presented in this document.
1.2 Research questions
The goal of this work is to create frameworks to assess the impact of climate conditions on the
operational and environmental performance of energy systems. Specically, I seek to answer the
following research questions:
A. Can the relationship between generator eciency and ambient temperature be derived em-
pirically?
3
B. Where is water consumed for each life cycle stage of coal-red electricity consumption, and
how does water consumption vary between where energy is produced versus where energy is
consumed?
C. How do climate change and water constraints aect local electricity generation, and what
are the cascading impacts on regional electricity generation?
The methodologies developed for each research question can be classied into one of two
groups: (a) data-driven approaches to assess historical trends in the water-energy-climate nexus
or (b) integrated modeling approaches to simulate projected trends in the water-energy-climate
nexus. A visualization of the research questions is shown in Figure 1.2.
1.3 Structure of document and resulting publications to
date
This document is organized into three additional chapters, each one corresponding to a research
question listed above, and a concluding chapter. Chapter 2 creates a methodology to link multi-
ple datasets and devise a regression model relating generator eciency to ambient temperature.
Chapter 3 develops a framework to spatially resolve the water consumption impacts of coal-red
electricity consumption. Chapter 4 introduces a framework that integrates data from climate,
hydrological, and water models at dierent scales with an electricity grid dispatch model. The
works described in Chapters 2 and 3 have been published, and the work in Chapter 4 is being
prepared for publication:
Chapter 2: M. Meng and K.T. Sanders. (2019). \A data-driven approach to investigate the
impact of air temperature on the eciencies of coal and natural gas generators." Applied
Energy, 25, 113486.
4
Chapter 3: M. Meng, E. Grubert, R.A.M. Peer, and K.T. Sanders. (2020). \Spatially
allocating life cycle water use for US coal-red electricity across producers, generators, and
consumers." Energy Technology, 2020, 1901497.
Chapter 4: M. Meng, J. Macknick, V.C. Tidwell, K.E. Bennett and K.T. Sanders. \Inte-
grating water, energy, and climate modeling to assess vulnerabilities to the U.S. Southwest
power grid." In preparation to be submitted to Environmental Research Letters.
PRESENT
Data-driven approaches to
assess historical trends
Integrated modeling approaches
to simulate projected trends
Can the relationship between
generator efficiency and ambient
temperature be derived
empirically?
Where is water consumed for each
life cycle stage of coal-fired
electricity consumption, and how
does water consumption vary
between where energy is produced
versus where energy is consumed?
How do climate change
and water constraints
affect local electricity
generation, and what are
the cascading impacts on
regional electricity
generation?
ENERGY CLIMATE WATER
A
B
C
Chapter 2
Chapter 3
Chapter 4
Figure 1.2: Diagram describing the chapters in this dissertation. The dotted vertical line repre-
sents the present day, splitting up the research by historical analyses on the left and projected
simulations on the right. The letters (A, B, and C) correspond with the research gap listed in 1.2
that each study addresses. The colors represent the topics that each study is focused on (green
= energy, orange = climate, blue = water).
5
Chapter 2
A data-driven approach to investigating the impacts of air
temperature on the eciencies of coal and natural gas
generators
The content included in this Chapter is published in: M. Meng and K.T. Sanders. (2019). \A
data-driven approach to investigate the impact of air temperature on the eciencies of coal and
natural gas generators." Applied Energy, 25, 113486.
2.1 Introduction
Thermal power production, accounting for 83% of total electricity generation, contributed to 28%
of greenhouse gas (GHG) emissions [7], 50% of sulfur dioxide (SO
2
) emissions, 10% of nitrogen
oxide (NO
x
) emissions [2], 40% of freshwater withdrawals [3], and 3% of freshwater consump-
tion [3] nation-wide in 2017. Coal and natural gas-fueled generators alone contributed 74% of
thermoelectric generation (62% of total electricity generation) [7], and thus, a signicant fraction
of these environmental impacts associated with the U.S. power sector. The GHG emissions, air
pollutants, and cooling water usage associated with power production depend on the eciency
of the electricity generating unit (EGU), which is in
uenced by factors such as fuel type, prime
6
mover, cooling system, and pollution controls. Less ecient EGUs generally require more fuel
and cooling water to generate one unit of electricity.
The impacts of climate variability and/or climate change on the operational capacity and
eciency of electricity infrastructure has been studied in past analyses. Several studies in the
literature have analyzed the in
uence of climate change on power infrastructure using thermo-
dynamic modeling. Studies utilizing thermodynamic models to assess the impact of temperature
increases on gas turbine eciency found that a 1

C increase in temperature correlates to an
eciency decrease of approximately 0.1% [8, 9]. Maulbetsch and DiFilippo studied natural gas
combined-cycle power plants in four dierent environments and found that recirculating tower
cooled power plants and dry-cooled power plants experienced capacity reductions between 0.3-
0.5% and 0.7% per 1

C increase in air temperature, respectively [10]. Erdem and Sevilgen showed
that the electricity generation of a gas turbine can decrease by 1.67-6.22% when temperatures
surpass 15

C [6].
Other analyses have utilized an integrated modeling approach to assess changes in the electric-
ity grid due to climate change. Van Vliet et al. applied a hydrology-electricity modeling framework
to predict that during the period spanning 2031-2060, approximately 4.4-16% and 6.3-19% of ther-
moelectric generation in the U.S. and Europe, respectively, will be lost [11]. They also found that
by the 2050s, thermal power plants globally could experience capacity reductions of 7-12% [12].
Sathaye et al. used temperature projections from general circulation models to estimate changes
in natural gas power plant capacity in California and found that 1.1-4.6% of peak capacity could
be lost by the end of the century [13]. Bartos and Chester combined climate, hydrology, and
thermodynamic power plant models to estimate a 1-3% reduction in summer generating capacity
by mid-century in the Western U.S., with reductions up to 7-9% when analyzed within the con-
text of a ten-year drought scenario [13]. Cook et al. used a combination of regression, climate,
and thermodynamic modeling to nd that a 1

C rise in cooling water temperature can lead to
a 0.15%-0.5% decrease in power output [14]. Liu et al. assessed the impact of climate change
7
and thermal discharge regulations on thermoelectric generators in the U.S. using a regional Earth
system model and a thermoelectric power generation model. The study found that by the 2060s,
climate change alone could reduce generation capacity by 2-3%, but environmental regulations
could actually raise capacity reductions up to 12% if power plant operators are forced to curtail op-
eration when water discharge temperatures exceed legal limits [15]. Miara et al. used the coupled
Water Balance Model and Thermoelectric Power and Thermal Pollution Model (WBM-TP2M)
to project changes in U.S. thermoelectric power plant capacity due to water constraints and cli-
mate change and found that under modeled contemporary climate scenarios, once-through-cooled
power plants experience the greatest reduction in capacity [16].
Only a handful of studies have statistically modeled the relationship between generator e-
ciency and climatological parameters based on real-world empirical data [17, 18], and only one
has been a peer-reviewed journal publication [19]. Henry and Pratson used regression modeling
to develop a relationship between climate parameters (water and air temperature) and generator
eciency [19]. The study indicated that once-through cooled generators will experience a change
of -0.11% to -0.05% in eciency per 1

C increase in intake water temperature and an eciency
change of -0.02% to +0.05% per 1

C increase in intake ambient air temperature based on a sam-
ple of 20 once-through power plants. It found that recirculating cooling systems had an eciency
change ranging from -0.06% to +0.04% per 1

C increase in wet-bulb temperature based on a
sample size of 19 power plants. Because Henry and Pratson's study only analyzed water-cooled
thermal power plants, natural gas red combustion generators and dry-cooled generators were not
included in their analysis. Furthermore, their limited samples of 20 once-through-cooled and 19
recirculating-cooled power plants might not be sucient for capturing variability across dierent
fuels, prime movers, cooling systems, and local climate.
With predicted climate change expected to increase air temperatures, it is important to un-
derstand how generators react to changes in climatic parameters as changes in the performance
of generators will consequently aect pollutant emissions, GHG emissions, and water usage. As
8
much of the prior body of work described above utilizes theoretical physics-based models to estab-
lish predicted capacity reductions, to date there is a lack of data-driven studies that quantify the
historical relationship between power plant eciency and changes in temperature for a statisti-
cally signicant population of power plants. Here we investigate how the eciencies of real-world
thermoelectric generators respond to changes in ambient air temperature, as the operational and
climatic variables aecting the performance of operational power plants are typically too complex
to capture in purely theoretical models. We also look at how the responses vary across fuel types,
prime mover, and cooling type. To do so, we utilize regression modeling, applied to over one
thousand EGUs (representing 618 unique power plants) to assess and critique the capability of
federally available datasets to support such an analysis. To the best of the authors' knowledge,
this will be the rst study to look at the impact of air temperature on generator eciency based on
a statistically representative set of electricity generating units. We also comment on how federal
datasets could be improved for usability to facilitate more data-driven studies in the future.
2.2 Methods
2.2.1 Preparing operational datasets for the regression analysis
A series of datasets characterizing each EGU in terms of its technical conguration and histor-
ical operational data were cleaned, ltered, and processed to prepare the data required for the
regression analysis. In this study, the term "EGU" is used to represent a coal steam (CL-ST)
generator, natural gas combustion (NG-GT) generator, natural gas steam generator (NG-ST), or
natural gas combined-cycle (NG-CC) system. Data from the years 2008 to 2017 were considered.
Each EGU's fuel type, prime mover, cooling technology, and combined heat and power (CHP)
status were characterized according to the U.S. Energy Information Administration's (EIA) Form
860, which provides information on every EGU over 1 MW nationally [20]. The EIA 860 Form
assigns a unit code to each generator for identication. For coal steam, natural gas combustion
9
and natural gas steam EGUs, no unit code is usually given, as each generator is its own unit.
For natural gas combined-cycle systems, multiple generators share a single unit code. Thus, each
coal steam, natural gas combustion and natural gas steam EGU refers to a single generator,
whereas an natural gas combined-cycle EGU refers to all generators that share the same unit
code. Nameplate capacity for each generator was aggregated to the EGU level. Generators are
linked to its respective boiler(s), and each boiler is linked to its respective cooling system (if any
is used).
This analysis focused on EGUs that utilize natural gas and coal as fuel and excluded all
combined heat and power units. For coal steam, natural gas steam, and natural gas combined-
cycle EGUs, only those with once-through without cooling pond (ON), once-through with cooling
pond (OC), recirculating with cooling pond (RC), recirculating with towers (RT) and dry (DRY)
cooling were considered. (Cooling systems classied as "other" or "hybrid" in the EIA dataset
were excluded from the study.) Natural gas combustion EGUs, which do not utilize wet or dry
cooling, were labeled as having a cooling system of "NONE (GT)". Units with one or more of
the same type of cooling system across an entire year were kept in the nal dataset, but instances
where a unit had 1) more than one cooling type within the year of analysis and/or 2) changed,
added or removed a cooling system within the course of a single year of analysis were removed.
For example, if an EGU and its respective cooling system began operation in May 2009, only data
from 2010 and onward for that EGU was used. Similarly, if a generator that began operation
before 2008 switched from a once-through cooling system to a recirculating cooling system in
August 2012, the entire year of 2012 is removed from that EGU's data.
Hourly gross load (MWh) and heat input (MMBtu) data were obtained from Environmental
Protection Agency's (EPA) Air Markets Program Data (AMPD) [21]. The AMPD provides
continuous emissions monitoring (CEM) data for generators that are required to monitor under the
EPA's Clean Air Markets Programs [21]. The AMPD classies collection methods for heat input
data into one of the following categories: "Measured", "Calculated", "Substitute", "Measured
10
and Substitute", "Not Applicable", "Undetermined", "Unknown Code". For this analysis, we
used only observations where the heat input was labeled as "Measured".
While the power plant codes between the two datasets were fairly consistent in the years an-
alyzed, the generator/boiler IDs had much greater variance. A script was developed to detect
naming patterns and inconsistencies to match as many generators as possible. While most gener-
ators were matched with this method, a few had to be matched manually. Diculties in joining
the EIA and EPA datasets are elaborated further in the discussion.
After matching units between the EIA and EPA datasets, the heat rate (HR), eciency (),
and capacity factor (CF ) at every hour were calculated Equations 2.1, 2.2 and 2.3, respectively.
The heat input is the primary energy put into generating the electricity load. The heat rate
(with units of MMBtu/MWh) is the amount of energy used to generate one unit of electricity
and is calculated by dividing the heat input by the gross load. The eciency is the percentage
of primary energy that gets converted into electrical energy and can be calculated by multiplying
the inverse of the heat rate by a conversion factor. Capacity factor refers to the percentage of the
total nameplate capacity at which the EGU is operating over a period of time. In this analysis,
we dene the change in capacity factor, also referred to here as the ramping rate, as the dierence
in capacity factor from one hour to the next. The AMPD contained observations where measured
heat rate were unreasonable or nonsensible. Lower and upper limits on hourly heat rate (for
all non-zero gross load hours) were applied based on the limits developed by the EPA for their
Power Sector Modeling Platform [22]. The limits are dependent on the fuel and prime mover
conguration of the generating unit and are provided in Table A.1 in the appendix.
HR =
Hourly Heat Input [MMBtu]
Hourly Gross Load [MWh]
(2.1)
=
Hourly Gross Load [MWh]
Hourly Heat Input [MMBtu]

3:412 [MMBtu]
[MWh]
(2.2)
CF =
Hourly Gross Load [MWh]
Nameplate Capacity [MW ] 1 hr
(2.3)
11
Upon inspection of heat rates, it became apparent that some natural gas combined-cycle
EGUs in the EPA dataset reported operational data for only the gas turbine part of the natural
gas combined-cycle EGUs, while others reported data representative of the entire unit [21]. All
natural gas combined-cycle EGUs with duct burners, a technology added to heat recovery steam
generators to increase their high-pressure steam output, should report generation (i.e., gross load)
from the entire unit (i.e., both the steam cycle and gas cycle generation), whereas units without
duct burners are not mandated to report steam generation. (The heat input of all components of a
natural gas combined-cycle EGU is reported regardless due to emissions monitoring regulations.)
More information on the reporting procedures can be found in the EPA Emissions Monitoring
Policy Manual [23] and the Clean Air Markets Emissions Collection and Monitoring Plan System
Reporting Instructions [24]. Figure 2.1a illustrates a density plot of all reported heat rates from
2008 to 2017 of natural gas combined-cycle EGUs, separated according to whether the natural gas
combined-cycle EGU has a duct burner as reported by the EIA Form 860 [20]. All duct burner
classications are determined by year; thus, a natural gas combined-cycle EGU can be classied
as having a duct burner in some years but not others. The majority of calculated heat rates for
the EGUs with reported duct burners range from 5.5 to 10 MMBtu/MWh (Figure 2.1a, left),
which is consistent with the average heat rate of natural gas combined-cycle EGUs [25]. However,
the density plot shown in for the units without duct burners (Figure 2.1a, right) is bimodal, with
about half of the heat rates ranging from 5.5 to 10 MMBtu/MWh and the other half ranging from
10 15 MMBtu/MWh (i.e., consistent with just the gas turbine part of the natural gas combined-
cycle EGU). This bimodal distribution indicates for natural gas combined cycle generating units
that do not have duct burners, some operators report generation from both the steam and gas
cycles, whereas other operators only report the gas generation.
Consequently, natural gas combined-cycle EGUs had to be recategorized to determine whether
the entire system's generation or just the gas cycle generation was reported in the EPA dataset to
prevent misinterpretation of the calculated heat rates. Natural gas combined-cycle EGUs reported
12
as having a duct burner in the EIA 860 form were automatically marked as reporting both steam
and gas generation. More analysis was required for EGUs that did not report a duct burner.
For each natural gas combined-cycle EGU, the 90th quantile of the hourly heat rates within each
calendar year of study was calculated. A density plot of the 90th percentiles showed a similar
bimodal distribution to the units without duct burners in Figure 2.1a, with a split between 5.5-10
MMBtu/MWh and 10-15 MMBtu/MWh. Accordingly, all natural gas combined-cycle EGUs that
had a 90th percentile heat rate value equal to or less than 10 MMBtu/MWh were categorized as
"Reporting Total Combined Cycle Generation". A natural gas combined cycle generating unit
with a 90th percentile heat rate above 10 MMBtu/MWh was marked as "Not Reporting Total
Combined Cycle Generation", meaning that only the gas generation of the combined-cycle EGU
was reported.
The revised classication of units is illustrated in the density plot of heat rates shown in Figure
2.1b. In comparing Figure 2.1a and Figure 2.1b, the bimodal distribution in Figure 2.1b is much
less prominent for units not reporting steam generation, with the vast majority of heat rates
falling in between 10-15 MMBtu/MWh. As for the units reporting steam generation, there is a
unimodal distribution, with most heat rates ranging 5.5-10 MMBtu/MWh.
There are still some observations that do not fall within the larger distribution, which is
most likely due to increased heat rate (decreased eciency) during hours where the generating
unit is rapidly increasing or decreasing its output, commonly referred to as a period of ramping.
(The ramping rate refers to an increase or decrease in an EGU's electrical output per unit time.)
Since ramping would likely obscure any impact of temperature on unit eciency, we attempted
to eliminate any operational data occurring during hours of signicant ramping. The change in
capacity factor from one hour to the next (i.e., CF ) was used as a proxy to detect ramping up
or down during generation, in eorts to remove these periods from skewing the eciency analysis.
The CF at each hour t (CF
t
) was calculated by subtracting the capacity factor at said hour
t (CF
t
) from the capacity factor at the previous hour (CF
t1
). For hours where operational data
13
(a)
(b)
Figure 2.1: (a) Hourly heat rates for natural gas combined-cycle electricity generating units
are plotted and categorized by whether each unit has a duct burner or not for each operating
year, based on EIA Form 860 data [20]. The density plot for units with duct burners show a
unimodal distribution between 5.5 to 10 MMBtu/MWh. The plot for units without duct burners
show a bimodal distribution, with one distribution ranging from 5 to 10 MMBtu/MWh and
another ranging from 10 to 15 MMBtu/MWh. (b) Hourly heat rates for natural gas combined-
cycle electricity generating units are plotted and categorized by whether each unit is reporting
steam generation based on our proposed classication methodology. Units with duct burners
were automatically classied as reporting steam generation, while units without duct burners
were classied based on the 90th percentile of their hourly heat rates each year.
were missing (either due to the EPA dataset not having the data or due to the data being ltered
out), the CF could not be calculated for the following hour, and a value of "NA" (not available)
was assigned to CF . All hours with a CF of "NA" were removed from the analysis. After
calculating the ramp rate at every hour for all units, hours with a gross load of "0" MW were also
removed.
14
When plotting the eciency versus capacity factor for natural gas combined-cycle EGUs,
two clusters were often found (e.g., one smaller cluster of lower hourly capacity factor values
and a larger cluster of higher hourly capacity factor values). Having two clusters would skew
the regression, so the smaller cluster containing fewer hourly capacity factor values (since these
values were less representative of the typical operation of the EGU) were removed. Density-
based clustering, through the dbscan package in R [26], to implemented detect two clusters of
capacity factor for every natural gas combined-cycle unit. Afterwards, the cluster with the lowest
count of observations was removed from the dataset. (On average, approximately 74% of each
NG-CC unit's data were part of the larger cluster.) By analyzing the most prominent cluster of
operational capacity factor values, each regression considered hours at which the EGUs generated
electricity within similar operational conditions, thus reducing the impact of other operational
characteristics on eciency variability (e.g., operating at a high versus low capacity factor).
To remove outliers in the operational data, observations where the eciencies and capacity
factors were not within the 5th and 95th percentiles (of each respective generating unit) were
omitted from the analysis. The distribution of CF indicated that most values fell between -0.1
and 0.1 (i.e. signifying a change of operational output within 10% of the previous hour). Data
in hours where the values of CF fell outside of these bounds were omitted from the regression
models, so that ramping did not skew the eciency analysis. (On average, 96%, 87% and 80%
of the data records considered for CL-ST and NG-CC, NG-ST, and NG-GT generation units,
respectively, fell within range and were kept for the analysis.)
2.2.2 Preparing climate datasets for the regression analysis
Local Climatological Data (LCD) from the National Oceanic and Atmospheric Administration
(NOAA) were used to obtain ambient temperature [27]. The NOAA dataset includes hourly (and
sub-hourly) dry-bulb temperature (T
db
), wet-bulb temperature (T
wb
), and relative humidity data
from thousands of weather stations around the country. For instances where multiple observations
15
were given within the same hour at a weather station, the average of each indicator (temperature,
relative humidity, etc.) for that hour at was calculated.
The longitude and latitude of each power plant were retrieved from the EIA 860 form [20]
and each power plant was matched to the nearest NOAA weather station for each year of EGU
operation. The reason for assigning a weather station for each year is to account for instances
where a weather station was installed closer to a power plant within the 2008-2017 timespan.
Most power plants were within 20 km (12 mi) of a NOAA station. However, there were some
plants where the nearest NOAA station was more than 30 km (18.6 mi) away, with a handful
reaching up to 50 km (31 mi). (A map of the distance from each EGU to the nearest weather
station is provided in Figure A.1 in the appendix.) For this analysis, we did not place a lter
on records based on the maximum distance between power plant and weather station, but we
acknowledge that the accuracy of climate data can decrease with increasing distance between the
unit and weather station.
After assigning a weather station to each generating unit, climate data from NOAA was merged
with operational data from the AMPD set. Any missing hourly data in the climate or operational
datasets were removed from the analysis. To remove outliers in the climate data, the 1st and 99th
percentiles for the wet-bulb and dry-bulb temperatures, respectively, were identied for each EGU.
For EGUs with recirculating tower cooling systems, observations where the wet-bulb temperature
did not fall within the 1st and 99th percentile wet-bulb temperatures were removed. For EGUs
with all other cooling systems, the same methodology was applied for dry-bulb temperature. The
reasoning behind why wet-bulb and dry-bulb temperatures were applied based on on cooling
system is explained in Section 2.2.3.
The impact of the local climate on the response of each EGU to temperature changes was
also analyzed. County-level climate zones, dened by the U.S. Department of Energy's (DOE)
Building America Program, were utilized for data sorting purposes [28]. These county-level classi-
cations, based on climate regions developed by the Pacic Northwest National Laboratory for the
16
International Energy Conservation Code, re
ect one of seven climate zones: Marine, Very Cold,
Cold, Mixed-Humid, Hot-Humid, Mixed-Dry, and Hot-Dry. Each EGU was assigned a climate
zone based on the county it is located in.
2.2.3 Regression analysis
We developed a regression model to relate EGU eciency with operational variables and ambient
air temperature. An exponential relationship was observed between hourly capacity factor and
eciency data. For the relationship between eciency and CF , a symmetrical relationship
was found: for negative values of CF , as the magnitude increased, the eciency of the EGU
decreased. Conversely, as positive CF values grew in magnitude, eciency values decreased,
suggesting that ramping an EGU up has a similar impact on eciency as ramping operations
down. Thus, CF was tted as linear spline, with a knot at CF = 0. We acknowledge that
some generating unit types, such as natural gas combustion units, are more susceptible to ramping
events than other unit types, such as coal steam units. However, by ltering out hours where
CF were below -0.1 and above 0.1, we aimed to remove instances where ramping would be the
prominent in
uencing variable on eciency. Furthermore, a priori analysis showed that having
CF versus not having CF , as well as keeping unltered versus lter CF values, did not
have signicant impacts on the regression results.
For regressing eciency versus temperature, choosing between dry-bulb and wet-bulb temper-
ature is important. For natural gas combustion EGUs, the eect of humidity on the eciency is
minimal [29], and therefore the dry-bulb temperature, was chosen as the dependent variable. Hu-
midity typically does not have a signicant eect on the eciency of wet-cooled generating units,
with the exception of units cooled with recirculating towers. In a recirculating cooling tower,
steam is cooled to a temperature approaching wet-bulb temperature [29, 30, 31, 32]. Thus, for
EGUs with recirculating cooling towers, the eciency was predicted using wet-bulb temperature.
For dry-cooled generating units, the evaporation process is not used, and heat is transferred only
17
to ambient air [32]. The steam is cooled to a temperature that reaches the dry-bulb temperature,
and therefore the eciencies of EGUs with dry cooling were predicted using dry-bulb temperature
[33]. Once-through cooling systems and recirculating with cooling pond systems do not rely on
the evaporative process (since water itself is used to remove heat), and as such, the eciencies of
EGUs utilizing once-through cooling and recirculating cooling with ponds were regressed against
dry-bulb temperature.
Each hourly temperature value was binned to account for more drastic decreases in eciency
at higher ambient temperatures. The 50th, 75th, and 90th percentiles of temperature values (wet-
bulb for recirculating tower-cooled and dry-bulb for all other cooling types) were calculated for
each generating unit. Then, each hourly observation was assigned a temperature bin depending on
if temperature value fell into one of the following categories: below the 50th percentile, between the
50th and 75th percentiles, between the 75th and 90th percentiles, and above the 90th percentile.
Before performing the regression, any EGU with less than 1,000 hourly observations was
removed from the analysis, as having a small number of observations can result to a skewed
regression t. Furthermore, we noticed that even if an EGU had more than 1,000 observations,
it is possible that the number of observations per temperature category could be small (i.e., an
EGU might only experience temperatures above the 90th percentile for 80 observations). Thus,
any EGU that has less than 250 observations in any temperature category was exluded from the
analysis as well. In the end, over one thousand generating units (over the span of 2008 to 2017),
representing 618 unique power plants, were analyzed (Figure 2.2) Figure A.2 in the appendix
provides maps of each EGU analyzed in the study by year.
Because of the combination of linear and non-linear relationships between eciency and the
various variables we examined, non-linear least squares (NLS) analysis was utilized. The regression
models shown in Equations 2.4 and 2.5 were used, where a (%/

C) is the coecient relating the
change in eciency due to a unit degree Celsius increase in dry-bulb temperature (T
db
), and b
(%/

C) is the coecient corresponding to change in eciency due to a unit degree Celsius increase
18
in wet-bulb temperature (T
wb
). For both equations, the capacity factor is exponentially regressed
to a power of k, with a coecient of d (%) and the change in capacity factor is linearly t with
a coecient of g. The is a mean-zero random error term. The models used here are similar to
the ones developed by Henry & Pratson [19], who analyzed the impacts of water temperature and
dry-bulb temperature on once-through cooling systems and the impact of wet-bulb temperature
on recirculating cooling systems.

DRY;RC;OT;NONE
=aT
db
+dCF
k
+gCF + (2.4)

RT
=bT
wb
+dCF
k
+gCF + (2.5)
To ensure that there was not signicant collinearity amongst regression variables, we performed
diagnostic tests for the regression variables on a few dozen random generating units. We did not
see a strong correlation between CF and CF . Furthermore, there did not seem to be any
collinearity, as the variance in
ation factor (VIF) for all variables was typically below 5.
19
Figure 2.2: The nal number of electricity generating units included in the analysis are character-
ized by (top) cooling type, (middle) fuel type and prime mover technology, and (bottom) climate
zone. The decrease in number of units in 2012 is due to a large number of NOAA weather stations
missing temperature data in that year.
20
2.3 Results
We assumed a linear relationship between eciency and temperature, and it is possible some of
the complexities in the relationship between the two variables were lost. Rahman et al. found
that the heat rate of a natural gas combustion EGU increased (and thus, eciency decreased)
fairly linearly with increasing ambient temperature [34]. Henry and Pratson, using the EPA and
EIA datasets, also found a linear relationship between temperature and eciency [19]. While it is
possible that the relationship between temperature and EGU performance is nonlinear, Sathaye
et al. argued that it is likely that nonlinear relationships occur at temperatures outside of current
and future temperature projections [13].
The p-value of a regression model can be used to infer if there is indeed a relationship between
the predictor (ambient temperature) and the response (eciency), and a small p-value (usually
0.05 or less) indicates there is indeed an association between the two variables. The majority of
EGUs (1220) have p-values that are less than 0.05, while 101 EGUs have p-values greater than
0.05. Of the 101 EGUs, 33 are natural gas combustion turbine units, 37 are coal steam units,
14 are natural gas steam units, and 17 are natural gas combined-cycle units. For the natural gas
combustion turbine units, all 33 units have nameplate capacities below 200 MW. All coal and
natural gas steam units with p-value > 0.05 were installed before 1980 and most are cooled with
once-through cooling systems. We were unable to identify a trend for natural gas combined-cycle
units across nameplate capacity, installation year, climate zone, or number of data records per
unit. For analyses of results onward, EGUs where the p-value are above 0.05 will be excluded.
The eect of air temperature on EGU eciency is inconsistent, with most generating units
(872) experiencing a drop in eciency due to rising temperatures, while other generating units
(396) experiencing a rise in eciency instead. While the median change in eciency (regressing
using all temperatures) for each cooling type is -0.01%/

C (Table A.2), the range of responses
for each cooling type varies. Results for dry-cooled generating units are the most consistent,
21
having the smallest range of eciency response (-0.05%/

C to +0.03%/

C), whereas generating
units with with once-through cooling (no ponds) and recirculating towers have the widest range
of values (-0.09%/

C to +0.07%/

C).
Applying the regression model within temperature bins, dry-cooled generating units are likely
to experience a greater reduction in eciency per 1

C increase in ambient temperature when
temperatures are above average (i.e., above the 90th percentile) compared to when temperatures
are average (i.e., between the 50th and 75th percentiles). The residual standard error values
remain consistent when regressing all temperatures together versus regressing temperature bins
(Figure A.4), so the accuracy of the regression models across temperature bins are relatively
similar. No distinguishable dierence can be seen between coal steam and natural gas combined-
cycle generating units that utilize dry cooling (Figure A.3). However, a visible dierence can
be seen when analyzing dry-cooled generating units across climate zones. In regions that are
hot and arid, units with dry cooling are more likely to experience greater eciency losses due
to increasing ambient temperature (Figure 2.4). At temperatures above the 90th percentile,
the median change in eciency for dry-cooled generating units located in Hot-Dry regions is -
0.09%/

C. We expected dry-cooled generating units located in cold regions to be less negatively
impacted by temperature increases, but our results show that these units are likely to experience
similar decreases in eciency (Figure 2.4) as units located in Hot-Dry regions. Upon closer
inspection, many EGUs in regions classied as Cold experience warmer temperatures of over
30

C in the summer months (e.g., counties in Nevada and Utah), which can skew the results
dierentiated by climate zone when compared to counties that are typically more mild year-round.
22
Figure 2.3: Regression results for generating unit eciency change per 1

C increase in temper-
ature (=T ) are plotted, characterized by cooling system type. (Outliers are not included
in box plots.) Dry-cooled systems experience the greatest decreases in eciency due to rising
temperatures. For generating units with recirculating cooling towers, wet-bulb temperature was
used in the regression (T = T
wb
). For all other cooling types and for natural gas combustion
generators, dry bulb temperature was used in the regression (T =T
db
).
23
Figure 2.4: Regression results for generating unit eciency change per 1

C increase in tempera-
ture (=T ) are plotted, characterized by cooling system type and climate zone. (Outliers are
not included in box plots.) The most extreme decreases in eciency can be seen for dry-cooled
generating units located in Hot-Dry regions. For generating units with recirculating cooling tow-
ers, wet-bulb temperature was used in the regression (T =T
wb
). For all other cooling types and
for natural gas combustion generators, dry bulb temperature was used in the regression (T =T
db
).
24
The impact of rising ambient temperature on the eciency of wet-cooled electricity generating
untis are still inconclusive and inconsistent after regressing within temperature bins. For 225
generating units, a 1

C increase in ambient temperature leads to increases in eciency, even at
temperatures above the 90th percentile or between the 75th and 90th percentiles. Of the 225
generating units, 66 are coal steam units with once-through cooling (no ponds), 45 are natural
gas combustion units, 40 are coal steam units with recirculating tower cooling, and 25 are natural
gas combined cycle with recirculating tower cooling.
Similar to dry cooling systems, many generating units utilizing recirculating cooling towers
experience greater eciency losses when temperatures are above average (i.e., above the 90th
percentile), but the results also span a wider range at higher temperatures. No signicant trend
can be found for recirculating tower-cooled generating units when further broken down by fuel and
prime mover (Figure A.3) or climate zone (Figure 2.4). For generating units with recirculating
systems that have cooling ponds, the trend of more signicant reductions in eciency at higher
than average temperatures cannot be seen.
No trends in eciency changes due to ambient air temperature increase can be found for
generating units with once-through cooling systems (with or without cooling ponds). The median
change in eciency per 1

C increase in ambient temperature at higher than average temperatures
(i.e., above the 90th percentile) does not vary much from the change in eciency at average
temperatures. Additionally, unlike generating units with recirculating cooling towers, the widest
range of results for once-through cooled generating units do not occur at the highest temperature
bin (above the 90th percentile). Instead, for once-through cooled generating units, the widest
range of eciency responses occur at ambient temperatures between the 50th and 75th percentiles.
Results for natural gas combustion turbines are mainly inconclusive and inconsistent. Distin-
guishing the regression across temperature bins does not have signicant impacts on the results.
Furthermore, there are no visible trends across climate zones (Figure 2.4), nameplate capacity
(Figure A.5), or installation year (Figure A.6).
25
2.4 Discussion
2.4.1 General implications
We observe discernible relationships between ambient temperature and power plant eciency for
dry-cooled generating units. Generating units utilizing dry cooling show a trend in decreasing
eciency with increasing ambient temperature, especially when temperatures are above average
(i.e., above the 90th percentile). We also found that local climate plays a large role in determining
the vulnerability of dry-cooled systems to temperature increases. The power plant that shows the
most signicant decrease in eciency is the Chuck Lenzie Generating Station in Nevada, where its
two dry-cooled natural gas combined-cycle generating units experience a decrease of approximately
-0.2%/

C (at temperatures above the 90th percentile). With increasing installations of dry cooling
technologies, many of which are in hot and arid regions, it is important to consider how these
systems will be impacted by future climate change scenarios.
There are no other empirical analyses focusing on the impacts of ambient air temperature on
the eciency of dry-cooled generating units, so we can only compare our results of theoretical
models. For natural gas combined-cycle units with dry cooling, at temperatures above the 90th
percentile, the median change in eciency at temperatures above the 90th percentile is -0.07%/

C
with a range of -0.20%/

C to +0.05%/

C. Maulbetsch and DiFilippo found that at temperatures
greater than 15

C, natural gas combined-cycle units that are dry-cooled experience a 0.7% de-
crease in capacity per 1

C increase in temperature [10]. While we cannot directly compare change
in eciency to change in capacity, our results are similar in magnitude in some cases and up to
an order of magnitude less in others, depending on the generating unit, than the value obtained
by Maulbetsch and DiFilippo [10].
The results obtained for wet-cooled generating units show that ambient temperature alone are
not sucient to predict the eciency of electricity generating units. While we expected increasing
26
ambient temperatures to result in decreasing generator eciency, our results are mainly incon-
clusive. Additionally, at higher temperature bins (i.e., above the 90th percentile), the change in
eciency for generating units with recirculating cooling systems (with and without cooling ponds)
span a wider range. Generating units with once-through cooling (with and without cooling ponds)
do not indicate a strong or discernible relationship between generator eciency and ambient air
temperature.
Results for generating units utilizing wet cooling systems, while similar to previous empirical
work, dier from thermodynamic and integrated models. Henry and Pratson found the impact
of wet-bulb temperature on the eciency of power plants with recirculating cooling towers to be
approximately -0.06%/

C to +0.04%/

C [19], which re
ects the range of values obtained from our
regression at temperatures below the 50th percentile (-0.07%/

C to +0.08%/

C). However, for all
other temperature bins, the resulting change in eciency for generating units with recirculating
cooling towers span a wider range of values. At temperatures above the 90th percentile, the
change in eciency ranges from -0.33%/

C to +0.21%/

C (Figure 2.3). For natural gas combined-
cycle units with recirculating cooling towers, at temperatures above the 90th percentile, the
median eciency change per 1

C increase in ambient temperature is -0.06%/

C, with a range
of -0.29%/

C to +0.16%/

C (Figure A.3). Our results are a magnitude of order greater than
the results obtained by Gonzalez-Diaz, who found that the eciency of one recirculating tower-
cooled natural gas combined-cycle generating unit decreases from 50.95% to 48.01% when the
temperature increases from 15

C to 45

C (equivalent to -0.098%/

C). Arrieta and Lora found
that between temperatures 0-35

C, the net power of a 600 MW recirculating-cooled natural gas
combined cycle units decreases by 75 MW [35]. The only other study to our knowledge that has
studied the impact of ambient air temperature on the eciency of once-through-cooled generating
units is by Henry and Pratson, who estimated the impact of temperature on the eciency of once-
through cooled power plants to be approximately -0.02%/

C to +0.05%/

C [19]. The results
obtained from our analysis fall within a wider range, especially at higher temperatures (i.e.,
27
above the 90th percentile), where once-through cooled (no ponds) generating units can experience
a change in eciency from -0.16%/

C to +0.13%/

C. Our results have a greater range than the
previous empirical analysis conducted by Henry and Pratson likely because of the greater sample
size used and therefore a wider range of possible responses from the generating units.
Inconsistencies in responses in eciency to rising temperatures for wet-cooled generating units
indicate that there are likely other variables which will in
uence the eciency, especially at
abnormally high temperatures. While recirculating cooling towers are tightly linked with the
evaporative process, the eciency of the generating unit is still dependent on stream
ow variability
as well [15]. The eciencies of power plants with once-through cooling or recirculating cooling
with a pond/reservoir are heavily dependent on the temperature of cooling water, since higher
intake water temperatures are less eective in removing heat in thermoelectric plants [36]. As
mentioned previously in the Results section, a large number of generating units whose eciencies
do not have strong relationships with ambient temperature (p-values greater than 0.05) are coal
and natural gas steam generating units installed before 1980 (and primarily have once-through
cooling systems). The cooling systems of power plants are often not as
exible in operation
as their boilers and generators (i.e., while a generator can operate at partial load, its cooling
system might still operate at full load) [37], and further, dry-bulb air temperature does not scale
linearly with water temperature, which will matter more in terms of moderating power plant
eciency. Thus, particularly for older once-through cooled units with low capacity factors, the
relationships between cooling water usage, generator eciency, and ambient climatic conditions
might be skewed [38, 36].
Previous studies utilizing theoretical models estimated that the eciencies of gas turbines
decrease by approximately 0.08-0.1% per 1

C increase in ambient air temperature [8, 9], but the
median eciency change for natural gas combustion EGUs from our results is an order of mag-
nitude smaller at approximately -0.01%/

C. Generally, increasing ambient temperatures should
lead to decreases in the eciency of natural gas combustion generating units. We expected to
28
see natural gas combustion generating units in Hot-Dry regions to experience more signicant
decreases in eciency as temperature increases when compared to other climate zones. However,
our analysis across climate zones are inconclusive, which we believe to be due to inlet air cooling
technologies that reduce the temperature of the ambient air before it enters the compressor. Inlet
cooling technologies are often installed at natural gas turbines located in hot and arid regions to
prevent decreases in eciency from temperature rises [39, 36].
2.4.2 Data limitations
Increased data availability opens up possibilities for applications of machine learning and articial
intelligence in developing energy models [40]. However, many issues persist in mergining climate,
water, and energy data datasets for integrated analyses [41]. Some data challenges include missing
data, varying spatio-temporal resolutions between datasets, heterogeneity in the data, and non-
uniformity in data collection standards [42]. Furthermore, applying machine learning techniques
to energy, water, and climate data can prove to be dicult in the presence of incomplete datasets
and outliers [42]. Many of these issues also impeded this investigation, particularly in regards
to matching data across disparate datasets (e.g., due to erroneous or missing EGU identication
numbers) and identifying outliers in the data.
It is possible to incorporate water temperature, water elevation, and stream
ow data into a
quantitative analysis, but attempting to utilize historical water data raises complications. Based
on water stations provided the United States Geological Survey (USGS) and the Environmental
Protection Agency, signicant numbers of hourly readings of water temperature did not become
available until recent years. Additionally, only a few water stations provide data at short enough
distances upstream to power plants to be of value. Water temperature readings are also further
complicated by other parameters such as stream
ow and the depth at which the sensor is located.
The EIA Form 923 dataset provides self-reported values for water intake temperature at power
plants [43], but these values are at the monthly level and their validity needs to be assessed.
29
Studies that analyzed reported water withdrawal values from the same EIA Form 923 dataset
found many inconsistent and unrealistic values [44, 45, 46].
Despite the exclusion of water availability and water temperature impacts in this analysis, we
gained many insights on the state of electricity grid operations data. One of the biggest diculties
in working with both the EPA AMPD and EIA Form 860 datasets is the lack of consistencies in
how generators and boilers are named, which makes it dicult to cross reference data between
the databases. These inconsistencies have likely been major factors in preventing other large
statistical analyses using these data. Almost all generator matching between the two datasets
was done through pattern detection in R. However, there were a dozen or so units that had to be
matched manually due to the fact that there was no discernible pattern between the two datasets
to match the units. Instead, generating unit installation dates provided by the EIA and AMPD
were used. We were able to match over a thousand units, a larger sample size than previous
studies utilizing the AMPD and EIA datasets by several orders of magnitude. We acknowledge
that there are likely a few units that were not matched properly, but we believe this to be a very
small percentage.
The reporting of data was also inconsistent across the years analyzed. For example, many
EGUs had unit codes and cooling system types that were missing in the earlier releases of the EIA
Form 860 datasets, which made it dicult to properly match natural gas combined cycle EGUs,
as we were concerned with a single cooling type for the unit (not the generator). We gathered data
for all 10 years in instances with missing data in earlier years, we assigned generating unit with its
unit code in the 2017 EIA data sets. Because the reporting instructions for the AMPD database
diered for ealier years [23], it is possible that power plant operators are not reporting data
consistantly across the span of the analysis (2008-2017). We tried to accomodate for this by doing
most of our characterization of EGUs by year. Another issue with the EPA AMPD database
is that steam generation is not reported for all natural gas combined-cycle EGUs. We believe
many previous studies have been utilizing the AMPD reported load values without thoroughly
30
considering or analyzing the quality of the data. While aggregating electricity generation to
the plant level (as many previous studies have done) is sucient in many contexts, plant-level
aggregation can be a problem when analyzing dierences in eciency across technologies since
many power plants use multiple fuels, prime movers, and cooling systems.
We matched each EGU to the nearest NOAA weather station (for each year in the study).
Because we did not implement a lter for maximum distance, there might be discrepancies between
the temperature reported at a weather station and the temperature at the actual EGU site.
Furthermore, heat sources at power plants can raise the local temperature at the EGU site itself
[31].
2.4.3 Implications for future work
Empirical analyses of electricity generation and climatic data can provide insights into how power
systems have responded historically to changes in climate conditions, but insucient available data
make it dicult to accurately and extensively capture the impacts of climate conditions on power
systems. The use of thermodynamic and/or integrated models, in conjunction with empirical data,
can help (1) address quality and quantity issues in historical data and (2) improve understanding
of the relationship between climatic variables and generator operations in the following ways:
Climate: The inclusion of cooling water characteristics will provide a more robust under-
standing of the vulnerability of wet-cooled generating units to climate variability. Because
current observational data available on stream
ow, water quality, and water elevation are
insucient for a large scale analysis, modeling of cooling water characteristics could ll in
existing data gaps.
Operation: While sucient data on operational variables such as capacity factor and ramp-
ing exist, the use of technologies such as inlet cooling will change how generating units
perform under climate variability. Furthermore, the use of emissions control equipments
31
can impact the net generation (and therefore eciency) of an electricity generating unit.
While the EIA 860 form has information regarding the pollution controls installed at U.S.
power plants [20], the actual operations of emission control systems are not available.
Regulation: While some permitting data are available, no information provides insight into
which hours power plants modify operations to comply with regulations. For example, the
Clean Water Act (CWA) section 316(a) regulates variations in surface water temperature due
to thermal euent from power plants and requires power plants to curtail operations when
either the discharge water temperature or the temperature dierence between the intake
and the discharge is too high [47]. As climate change projections predict decreasing water
availability and increasing water temperatures in many regions, accounting for curtailments
related to regulated intake water temperatures and/or discharge water temperatures will be
important in assessing the resiliency of the electricity grid [15].
2.5 Conclusions
As climate change is expected to increase temperatures in many areas across the country, the
performance and reliability of the electricity grid will decrease. A regression model was developed
and applied to a large set of EGUs to quantify the impact of ambient air temperature on generator
eciency over a 10 year period of time. The impacts were analyzed across varying fuel types,
prime movers, cooling system types and climate regions. This study is the rst to statistically
analyze the impacts of climate and operational variables on generator eciency with such a large
sample size. While previous studies focused on dozens of generators at most, this analysis included
over one thousand electricity generating units (618 unique power plants).
Results indicate that while air temperature alone is insucient to capture the relationship
between generators using wet cooling technologies, dry-cooled generating units can experience
a decrease in eciency of up to 0.2% per 1

C increase in ambient temperature, especially in
32
areas that are hot and arid. As the number of dry-cooled power plants will likely continue to
grow into the future, the potential impact of the vulnerability of these power plants to climate
variability and climate change can be signicant. Results regarding the eciency losses of natural
gas combustion turbines, which do not utilize water-cooled or air-cooled condensers for electricity
generation, were also inconclusive. The use of inlet cooling technologies, which are commonly
installed at natural gas combustion units in hot regions to improve generator performance during
high temperature, likely distorted the relationship between generator eciency and temperature
in this analysis.
We also believe the insights acquired on the state of power plant operations data are useful
for the scientic community. We suggest that authors who plan to use the EPA CEMS data do
so with caution, particularly for natural gas combined-cycle units because both steam and gas
generation are not always reported. Although we present a method for identifying whether or
not generation from the entire natural gas combined-cycle unit is reported, there is no perfect
solution. For the EIA dataset, we recommend checking for inconsistencies between recent and
older years when attempting to do multi-year analyses. These issues were componded by the fact
that there are many inconsistencies in how generators are named in the EPA and EIA datasets.
Despite these data challenges, there is a great deal of merit to these datasets when meticulous
care is taken in the interpretation and analyses processes.
33
Chapter 3
Spatially allocating lifecycle water use for US coal-red
electricity across producers, generators, and consumers
The content included in this Chapter is published in: M. Meng, E. Grubert, R.A.M. Peer, and
K.T. Sanders. (2020). \Spatially allocating life cycle water use for US coal-red electricity across
producers, generators, and consumers." Energy Technology, 2020, 1901497.
3.1 Introduction
In order to meet demand, energy is often transported from one region to another. There can
be large spatial decouplings across the energy supply chain, as energy production, processing,
conversion, and consumption can occur in disparate locations and are dicult, if not impossible,
to track with existing data. Consequently, a common challenge in quantifying the environmental
damages imposed by energy consumption is spatially resolving energy supply chains to better
provide regional accountability for environmental damages. For example, the regions where energy
is produced are often far away from the regions where energy is ultimately consumed. This
challenge is exemplied in electricity systems, where the additional
ow of secondary energy in
the form of electricity is nearly impossible to track due to the interconnected nature of transmission
and distribution infrastructure [48].
34
Many studies analyzing the life cycle environmental impacts of energy systems have focused on
quantifying greenhouse gases, such as carbon dioxide [49, 50, 51, 52, 53]. Because carbon dioxide
is homogeneously mixed in the troposphere, the location of emissions is relatively unimportant
in understanding their consequences for global climate change [54]. Conversely, other notable
environmental impacts of energy systems, such as water consumption and air-quality pollutant
emissions, are most signicant on local and regional spatial scales [44]. Thus, the location of these
impacts is much more important to understanding environmental and public health consequences
than in the case of greenhouse gas emissions. Despite the challenge and importance of tracking
these impacts, current methodologies are insucient for characterizing the link between energy
demand and energy impacts across regions.
A large body of work has consolidated ndings from peer-reviewed literature and other sources
(e.g., government reports) to develop water intensities for the entire life cycle of electricity gen-
eration fuels and technologies [55, 56, 57, 58, 59, 60, 61, 62]. Grubert and Sanders derived water
intensities and detailed water footprints for the US energy system in 2014 across multiple fuel
cycles [63]. Other studies have quantied the water footprint of primary energy extraction, such
as coal mining [64] or hydraulic fracturing [65, 66, 67], fuel transportation [62, 68, 69], and power
plant infrastructure [56, 70]. Multiple studies have developed water intensities for electricity gen-
eration using a variety of sources. Macknick et al. synthesized primary literature estimates of
water use by electricity generating technologies to derive technology-specic water intensities [71].
The United States Geological Survey developed water intensities using theoretical heat budget
models [3]. Peer and Sanders utilized nationally reported water usage data by power plant op-
erators to develop fuel, prime mover, and cooling technology-specic water intensities [72]. Most
water intensities for electricity generation in the literature are technology specic, derived irre-
spective of geographic location or regional context (though see Peer et al. [73], which assesses
electricity generation water intensities within US eGrid regions).
35
The concept of virtual water was introduced to represent the
ow or transfer of water embedded
in the import and export of commodities [74]. The body of work in the virtual trade of water
has been largely dominated by water embedded in the transfer of agricultural goods [75, 76], but
there has been growing interest in the virtual
ow of water embedded in energy trade. A number
of studies have quantied the water embedded in economic trade, including the trade of primary
coal energy, spanning across global [77, 78], national [79, 80], and regional [81, 82] geographic
scales. A number of studies have assessed the virtual water embedded in the trade of electricity,
usually focused on electricity grids within China [83, 84, 85, 86, 87], the US [88, 89], or Europe
[90]. These studies of virtual water
ows for electricity transfers, however, generally only account
for the water used for electricity generation and do not consider upstream water used for fuel
extraction. One recent exception includes an analysis by Zhu et al. that assesses the life cycle
water impacts of electricity consumption, accounting for primary energy water use, electricity
generation water use, and the transport of electricity [69]. The authors derived life cycle water
footprints of coal-based electricity generation within Chinas provinces and virtual water
ows
across the provinces through electricity transmission [69]. However, the study did not consider
water use dierences across coal rank, and no similar study has been conducted in the context of
the US.
This manuscript focuses on improving methods to quantify water use across electricity supply
chains, from primary energy production through electricity transmission. We propose a frame-
work to spatially resolve this life cycle water use (withdrawal and consumption) and apply the
framework to a case-study of coal-red electricity consumption in the US. We contextualize the
water embedded in US coal energy by coal rank, mining practices, regional dierences, generator
technology, and electricity transport [71, 91, 92]. We note the various scales and intensities at
which water is used for each stage of the coal energy life cycle, highlighting the spatial decoupling
between raw energy producers, electricity generators, and electricity consumers.
36
Analyzing the case study of US coal-red power generation is interesting as processes associ-
ated with producing, processing, and converting coal into electricity at power plants can impose a
variety of associated socio-environmental impacts, including greenhouse gas emissions, air pollu-
tant emissions, and water withdrawals and consumption (water being the focus of this analysis).
From a water resource accounting perspective, coal is an interesting case study in the US because
almost all of the coal produced (93%) is consumed to produce electric power [93] and the spa-
tial decouplings between primary energy production regions, electricity production (i.e., coal-red
power plants) and electricity consumption are large. Further, despite a decline over recent years,
coal-red power plants still represent a major fraction of power production in the US (approxi-
mately 27% of total US electricity generation in 2018 [94]) and consequently a major fraction of
responsibility for the environmental impacts of electricity generation.
Coal production occurs in multiple mining regions across the US, with variations in rank and
grade. Prior to production, many coal mines must be dewatered, which is a major contributor
to water intensity. The water intensity of mining coal is highly region-specic and is not linearly
related to fuel production volumes [63]. Some coal is washed after it is extracted from mines
and before it reaches power plants, to remove impurities to improve coal quality (i.e., to increase
combustion eciency or reduce emissions potential). The most common coal ranks in the US
are bituminous (BIT), subbituminous (SUB), and lignite (LIG). Waste coal (WC) is a useable
byproduct of coal processing operations, often comprised of a mixture of coal, soil, and rock.
Rened coal (RC), which made up approximately 20% of net coal generation in 2017 [43], is
usually feedstock coal combined with proprietary additives to reduce moisture and pollutants.
Coal rening most commonly happens at the power plant [95].
After coal is mined and prepared, it is transported to power plants across the US. Water is
generally used as a cooling
uid at thermal power plants to condense steam exiting the mechani-
cal turbine and maximize the pressure dierential across the turbine for power production. Each
37
power plants water withdrawal and consumption intensities are heavily dependent on character-
istics such as fuel type, prime mover, and cooling technology. Thermoelectric power plant cooling
is a major source of water withdrawals in the country, having accounted for 41% of US water
withdrawals in 2015 [96]. Most of the water consumed in the life cycle of coal-red electricity
consumption is dedicated to power plant cooling [63].
After electricity is generated, it is not necessarily consumed locally. The operation of power
plants in the US electric system is managed by balancing authorities, which can import or export
electricity from each other to balance supply and demand within their region. Once electricity
generated at a power plant is pushed to the grid, electric energy can travel to any demand
center connected through the transmission and distribution network. It is impossible to trace
the exact source of electric energy that is consumed at any particular grid connected location.
In other words, power generated in one region can be consumed in a completely dierent region,
particularly in cases when a balancing authority does not meet demand with electricity produced
within its own boundaries.
The goal of this work is to spatially resolve water use in the coal energy system. We use
comprehensive water consumption and withdrawal intensities for each stage of the coal fuel cycle
mining, preparation, and power plant cooling to quantify the water used for each process. One
of the novelties of this work is a detailed accounting of the mining region source of nearly all
coal burned to produce electric power in the US. We also provide a methodology to estimate the
amount of coal-red electricity transferred between balancing authorities. We seek to answer the
following research questions: (1) Where is water consumed for coal-red electricity consumption
at each life cycle stage? (2) How does water consumption vary between where the coal energy is
produced versus where the coal energy is consumed?
38
3.2 Methods and Materials
Our analysis links water use by coal-related life cycle processes from producer (region of coal
extraction) to consumer (region of electricity consumption). Although denitions of water with-
drawal and water consumption vary across the literature, here we dene withdrawal as the total
volume of water removed from a source (river, reservoir, ocean, etc.) and consumption as the
subset of withdrawn water that is not returned to the source. We treat each mining province and
balancing authority as nodes in a network, and the
ow of energy is the transfer of coal energy
(in units of gigajoules (GJ) when referring to primary energy and megawatt-hours (MWh) when
referring to electric energy) from node to node. We assigned region-specic water intensities as-
sociated with each of three life cycle stages to each coal energy
ow. The three life cycle water
uses considered are mining, processing, and power plant cooling. The three spatial specications
are: the mining province where coal is produced, the balancing authority where the electricity is
generated, and the balancing authority where the electricity is consumed.
The study focuses on coal-red electricity consumption in 2017, re
ecting the most recent
vetted dataset at the time that the analysis was conducted [95, 20, 43]. A schematic summarizing
the methodology and datasets utilized is provided (Figure 3.1).
3.2.1 Water for coal energy produced in mining provinces
We used 2017 coal fuel consumption at power plants data provided by the US Energy Information
Administration (EIA) Form 923 [43]. Information on power plant ID, coal rank, and annual
fuel consumption (in units of GJ) was utilized. The EIA Form 923 fuel consumption dataset
includes a separate record of fuel consumption for each coal type (bituminous, subbituminous,
lignite, waste coal, and rened coal). Purchase receipt data from the EIA Form 923 dataset were
used to determine the coal mining province sourcing coal to each power plant. While we focused
our energy
ow calculations on the year 2017, we used coal purchase data from the years 2013
39
Water for coal energy
production
Water for coal-fired electricity
generation
Water for coal-fired electricity
consumption
Flow of water
mining province to
balancing authority
balancing authority to
balancing authority
Determine water withdrawal
and consumption for mining
and preparing coal
Determine water
withdrawal and
consumption for coal-
fired electricity
generation
Determine water withdrawal
and consumption for net coal-
fired electricity consumption
Coal purchased (by rank)
Calculate fraction of coal
energy purchased from
each province (by rank)
If refined coal, assume
2017 distribution of coal
rank and province
Aggregate coal energy
consumption by rank and
mining province
Apply rank-specific water
intensities for each
mining province
1
spatial scale: power plant coal mining province balancing authority
Coal-fired generation
Calculate
balancing
authority-specific
water intensities
Apply technology-specific
water intensities
2
Aggregate net coal-fired
generation and water use
Generation,
separated by fuel
source
Fraction of net
coal-fired
generation
Monthly electricity
exchanges across
balancing
authorities
Aggregate to yearly
exchanges
Calculate coal-fired
electricity exchanges
Apply balancing
authority-specific water
intensities to electricity
imports and exports
Figure 3.1: Methodology schematic. Notes and Data Sources:
1
Water rates by Grubert and
Sanders [63];
2
Water rates by Peer and Sanders [72]; For balancing authorities where the exported
coal-red electricity was greater than the net coal-red electricity generated, we calculated the
ratio of exports to other balancing authorities and multiplied the net coal-red generation by
each ratio. This set the net coal-red electricity generation as the maximum coal-red electricity
exported.
to 2017 to derive the average contribution of each coal mining region to coal burned at power
plants over the ve year span (considering that coal is not necessarily consumed immediately and
can be stored onsite). Each purchase record in the coal purchase dataset included the following
information: plant ID, coal type (bituminous, subbituminous, lignite, or waste coal), coal mine
state, coal mine county, quantity purchased, and average heat content of coal purchased. Coal
purchases, given in units of short tons, were multiplied with the average heat content of that
specic coal to estimate the amount of fuel purchased in units of GJ.
40
Mining provinces were assigned to each coal purchase record, according to regions dened by
Grubert and Sanders [72]. For coal mines outside of Kentucky, each mining province was assigned
based on the coal mine state. Since eastern Kentucky is considered part of the Appalachia/Eastern
mining region, and western Kentucky is part of the Interior mining region, coal mines in Kentucky
were assigned mining provinces based on county. The EIA Form 7a contains information on coal
mines, mine state, mine county, and the corresponding mining region [43]. The EIA Form 7a uses
the region conventions of Appalachia Central for eastern Kentucky and Illinois Basin for western
Kentucky. Thus, the Appalachia Central and Illinois Basin naming conventions in the EIA Form
7a were changed to Appalachia/Eastern and Interior regions, respectively. We used the 2008 to
2017 releases of the EIA Form 7a dataset to create a list that matches Kentuckys counties to
mining provinces. The list of these matches was then combined with coal purchase records from
mines located in Kentucky. While most purchase records were assigned provinces using state or
county information, approximately 92 (out of 407) power plants generated coal-red electricity but
had either (a) no associated coal purchase records detailing the nature of these purchases between
2013 to 2017 or (b) purchase records that did not provide any information on the coal source. For
these plants with unknown coal sources, the mining provinces were determined through manual
internet searches (e.g., news articles, power plant operator reports, facility webpage) that detailed
the coal type and regional source. If no information on a power plant's coal source were found,
the power plant's mining province was assigned as \Unknown". In total, 54 coal-red power
plants in 2017 (representing approximately 7% of coal generation in 2017) were labeled as having
an Unknown coal source. Thus, 93% of coal-red electricity in 2017 was sourced to a mining
province.
While separate fuel consumption records are available for all coal ranks in the EIA 923 dataset,
there are no purchase receipts available for rened coal. Thus, we do not know precisely which coal
ranks (and how much coal within those ranks) were rened. To resolve the transition from mining
province to power plant in the case of rened coal consumption, we assumed the coal purchase
41
receipts of non-rened coal ranks were representative of the rened coal consumed. In the cases
of power plants that consumed rened coal, we calculated the ratios of coal energy purchased
from mining provinces (not separating by coal type). We multiplied each power plants rened
coal energy consumption in 2017 with the province-specic ratios to obtain the
ow of coal energy
from mining province to power plant. For all other coal types (bituminous, subbituminous, lignite,
and waste coal), we calculated the ratios of coal energy purchased from mining provinces, for each
coal type, for each power plant. We then multiplied each power plants coal energy consumption
with the corresponding coal type- and mining province-specic ratios.
The water withdrawn and consumed for mining coal and preparing coal were determined using
water intensities developed by Grubert and Sanders [63]. At the mining level, we assume that
dewatering is a consumptive withdrawal, as water is removed from its source and not returned
(consistent with the denition of consumption used by Grubert and Sanders [63]). The water
intensities for coal mining and coal preparation are provided in Tables B.1 and B.2, respectively.
The assumptions made for calculating the water withdrawals and consumption at each stage for
each mining region are further explained in the Supporting Information. Water intensities (in
units of m
3
per GJ) were multiplied with the coal energy consumed at the mine mouth (in units
of GJ) to obtain the volume of water (in units of m
3
) withdrawn and consumed to mine and
prepare coal in all mining provinces.
3.2.2 Water for coal-red electricity generated in balancing authorities
We used the EIA Form 923 dataset to obtain net electricity generation for coal power plants in
2017. We also used the 2017 release of the EIA Form 860 dataset to match each power plant
to a balancing authority [20]. After matching each power plant to a balancing authority, we
aggregated all coal-red electricity generation (in units of MWh) to the balancing authority level.
The balancing authority names and codes (as used by the EIA) are provided in the Supporting
Information (Table B.3).
42
The water withdrawal and consumption for cooling coal-red power plants were determined
using water intensities developed by Peer and Sanders [72]. Each coal power plant was assigned
the median water intensity from Peer and Sanders study based on the power plants prime mover
and cooling technology [72]. We multiplied the plant-level water intensities (which were created
using 2014 electricity data) with 2017 electricity generation to obtain 2017 cooling water use for
each power plant. We then aggregated the 2017 cooling water use and 2017 coal-red electricity
generation to the balancing authority level to obtain balancing authority-specic water withdrawal
intensities and water consumption intensities. The water intensities for all balancing authorities
are provided in the Supporting Information (Table B.4).
3.2.3 Water for coal-red electricity consumed in balancing authorities
We estimated
ows of electricity between balancing authorities, from where electricity generation
occurred to where electricity consumption occurred. We used the EIAs Open Data Application
Programming Interface (API) to obtain monthly balancing authority-to-balancing authority elec-
tricity interchanges [97]. The interchange data were aggregated to the yearly level to remain
consistent with the primary energy
ows previously established.
While the data provide electricity imports and exports between balancing authorities, they do
not provide information on the exact source (i.e., power plant) of electricity generation. Therefore,
there is no way to know with certainty what fractions of fuels were burned to produce electricity
exported from a balancing authority. Instead, we assumed that the fuel sources embedded in
the electricity exported out of a balancing authority is likely to re
ect the proportions of fuels
used for power production across each respective balancing authority's generation portfolio. We
used net generation (for all fuel types) data from the EIA Form 923 dataset to calculate the
percent contribution from coal-red sources for each balancing authority [43]. We then multiplied
all electricity exported from a balancing authority with the percentage of coal-red electricity
generation within the respective balancing authority. By doing so, we were able to estimate the
43

ows of coal-red electricity traded between balancing authorities. We multiplied the interchanged
coal-red electricity with the balancing authority-level water intensities (Table B.4) to calculate
the
ow of cooling water.
3.3 Results and Discussion
Table 3.1 displays water consumption and water withdrawals for coal-red electricity consumption
in the US in 2017. Cooling power plants made up the highest percentage of water consumed
and water withdrawn (80% and 99%, respectively) for the coal fuel cycle. Our estimate that coal
mining consumed 410
8
m
3
of water in 2017 is approximately 33% less than Grubert and Sanders
estimate for 2014 (610
8
m
3
) [63]. Similarly, our estimates for water consumed and withdrawn for
coal preparation in 2017 are both approximately 40% less than Grubert and Sanders estimates for
2014. This dierence is likely driven by the decrease in coal production in the US between 2014
and 2017, where coal production nationwide decreased by 23% [93]. Additionally, states with more
water-intensive coal mines, such as Kentucky, contributed a smaller fraction of total national coal
production in 2017 versus 2014 [93]. Meanwhile, states in the Northern Great Plains province,
such as Wyoming, that consume less water for mining had increased contribution to nationwide
coal production [93]. Our ndings for the quantities of water consumed and withdrawn for cooling
coal-red power plants in 2017 are approximately 22-28% less than the estimates by Peer et al.
for 2014, which is reasonable given coal consumption for electric power production in the US
decreased by 22% between 2014 and 2017 [93].
Given the large variation in anticipated consumptive water intensity associated with dierent
coal basins, we note the importance of accounting for the variability in coal supply in estimating
water consumption. Although approximately 49% of all coal fuel consumed for electricity gener-
ation came from the Northern Great Plains mining province in 2017, the region contributed to
less than 2% of national US water consumption and withdrawal for coal mining and preparation
44
Table 3.1: Water withdrawn and water consumed at each life cycle stage for coal-red electricity
consumption in the US in 2017. Each percentage represents the percent contribution from each
process stage to the total amount of water consumed or water withdrawn.
Process
Water Consumed Water Withdrawn
m
3
Percentage m
3
Percentage
Mining 4.0E+08 19% 4.0E+08 0.4%
Preparation 3.6E+07 2% 4.4E+08 0.5%
Cooling Power Plants 1.7E+09 80% 9.7E+10 99%
Total 2.2E+09 100% 9.8E+10 100%
(Figure 3.2). The Northern Great Plains mining province has the lowest water consumption and
withdrawal per unit of coal energy produced (Table S2). Instead, coal from the Interior province
was responsible for the greatest amount of water consumed and withdrawn for coal mining and
preparation (Figure 3.2). (Unlike the mining process, where all water withdrawn is assumed to
equal water consumed, not all water withdrawn for the coal preparation process is consumed.)
Fuel Consumed
(13 × 10
9
GJ)
Water Consumed
(440 × 10
6
m
3
)
Water Withdrawn
(850 × 10
6
m
3
)
0%
25%
50%
75%
100%
0%
25%
50%
75%
100%
0%
25%
50%
75%
100%
Percent contributed by
each mining province
Mining Province
Appalachia/Eastern
Gulf Coast
Interior
Northern Great Plains
Rocky Mountain Region
Imported
Unknown
Figure 3.2: (Left) Primary coal fuel consumed for coal-red electricity consumption in 2017.
(Middle) Water consumed for mining and preparing coal. (Right) Water withdrawn for mining
and preparing coal.
45
Figure 3.3 displays the
ow of coal energy from mining province where the coal is produced
to the balancing authority where the coal is burned. Figure 3.3 also includes the associated
water consumption for mining, water withdrawal for preparation, and water consumption for
preparation. The largest transfer of coal energy between mining province and balancing authority
in 2017 was from the Northern Great Plains to the Midcontinent Independent Transmission System
Operator (MISO). However, the greatest
ow of water consumed for mining was for coal from
the Interior province to the MISO balancing authority. The dierence between energy and water
consumption is again due to the fact that the Powder River Basin, the dominant Northern Great
Plains district, has essentially been fully dewatered already [63]. A lot of the coal consumed at
power plants come from mines that are located locally or regionally (e.g., power plants in Illinois
tend to burn coal from the Interior province and power plants in Minnesota mostly burn coal
from the Northern Great Plains). However, power plants located in states that do not produce
coal have to import coal from faraway locations. For example, power plants in Florida burn coal
from mines in the Interior and Appalachia/Eastern provinces.
At the power plant cooling stage, we nd that water consumption is closely tied to the amount
of electricity generated, but water withdrawals are driven by cooling system technology. While
water consumption intensities do not vary greatly between cooling technologies, water withdrawal
intensities do vary signicantly between once-through cooling systems and recirculating cooling
towers. Figure 3.4 displays the coal-red electricity generated within each balancing authority in
2017, as well as the water consumed and water withdrawn for cooling coal power plants in each
balancing authority. In each plot, the bars are ordered from greatest to least amount of coal-red
electricity generated. The Southwest Power Pool (SWPP) balancing authority generated the third
highest amount of coal-red electricity in 2017 (120 TWh) and consumed the third highest amount
of water for cooling coal power plants as well. The Electric Reliability Council of Texas (ERCO)
balancing authority, despite generating 8.3% (10 TWh) of coal-red electricity less than SWPP
in 2017, withdrew a higher amount of water for cooling power plants. This is because Texas has
46
Unknown
Rocky Mountain
Region
Northern Great
Plains
Interior
Gulf Coast
Appalachia/
Eastern
WALC
WACM
TVA
TEPC
TEC
SWPP
SRP
SPA
SOCO
SEC
SCEG
SC
PSCO
PNM
PJM
PGE
PACE
OVEC
NYIS
NWMT
NEVP
MISO
LGEE
LDWP
JEA
ISNE
IPCO
GVL
FPL
FPC
FMPP
ERCO
EEI
DUK
CPLE
CISO
BPAT
AZPS
AECI
AEC
0
5
10
Mining Region Balancing Authority
10
9
GJ
Coal Energy Consumed for Electricity
(a)
Unknown
Rocky Mountain
Region
Northern Great
Plains
Interior
Gulf Coast
Appalachia/
Eastern
WALC
WACM
TVA
TEPC
TEC
SWPP
SRP
SPA
SOCO
SEC
SCEG
SC
PSCO
PNM
PJM
PGE
PACE
OVEC
NYIS
NWMT
NEVP
MISO
LGEE
LDWP
JEA
ISNE
IPCO
GVL
FPL
FPC
FMPP
ERCO
EEI
DUK
CPLE
CISO
BPAT
AZPS
AECI
AEC
0
100
200
300
400
Mining Region Balancing Authority
Million Cubic Meters
Water Consumed for Mining Coal
(b)
Unknown
Rocky Mountain
Region
Northern Great
Plains
Interior
Gulf Coast
Appalachia/
Eastern
WALC
WACM
TVA
TEPC
TEC
SWPP
SRP
SPA
SOCO
SEC
SCEG
SC
PSCO
PNM
PJM
PGE
PACE
OVEC
NYIS
NWMT
NEVP
MISO
LGEE
LDWP
JEA
ISNE
IPCO
GVL
FPL
FPC
FMPP
ERCO
EEI
DUK
CPLE
CISO
BPAT
AZPS
AECI
AEC
0
10
20
30
Mining Region Balancing Authority
Million Cubic Meters
Water Consumed for Preparing Coal
(c)
Unknown
Rocky Mountain
Region
Northern Great
Plains
Interior
Gulf Coast
Appalachia/
Eastern
WALC
WACM
TVA
TEPC
TEC
SWPP
SRP
SPA
SOCO
SEC
SCEG
SC
PSCO
PNM
PJM
PGE
PACE
OVEC
NYIS
NWMT
NEVP
MISO
LGEE
LDWP
JEA
ISNE
IPCO
GVL
FPL
FPC
FMPP
ERCO
EEI
DUK
CPLE
CISO
BPAT
AZPS
AECI
AEC
0
100
200
300
400
Mining Region Balancing Authority
Million Cubic Meters
Water Withdrawn for Preparing Coal
(d)
Figure 3.3: (a) Flow of coal energy consumed to generate electricity in 2017, from mining province
to balancing authority. (b) Flows of water consumed for mining the coal used to generate elec-
tricity in 2017, from mining province to balancing authority. (c) Flows of water consumed to
prepare coal that was consumed to generate electricity in 2017, from mining province to balanc-
ing authority. (d) Flows of water withdrawals dedicated to preparing coal that was consumed to
generate electricity in 2017.
47
a higher fraction of coal-red generation that is cooled with once-through systems than SWPP,
resulting in a higher water withdrawal intensity in the ERCO balancing authority (Table B.4).
0
100
200
300
MISO
PJM
SWPP
ERCO
SOCO
PACE
TVA
WACM
LGEE
DUK
SRP
AECI
NWMT
OVEC
PNM
AZPS
PSCO
SC
CPLE
FPC
TEPC
SCEG
LDWP
SEC
FMPP
JEA
BPAT
TEC
EEI
NEVP
SPA
PGE
ISNE
WALC
HICC
AEC
NYIS
ASCC
GVL
CISO
FPL
IPCO
Net Coal Electricity
Generated
(TWh)
0
100
200
300
400
MISO
PJM
SWPP
ERCO
SOCO
PACE
TVA
WACM
LGEE
DUK
SRP
AECI
NWMT
OVEC
PNM
AZPS
PSCO
SC
CPLE
FPC
TEPC
SCEG
LDWP
SEC
FMPP
JEA
BPAT
TEC
EEI
NEVP
SPA
PGE
ISNE
WALC
HICC
AEC
NYIS
ASCC
GVL
CISO
FPL
IPCO
Cooling Water Consumed
for Coal Power Plants
(Million Cubic Meters)
0
10,000
20,000
30,000
MISO
PJM
SWPP
ERCO
SOCO
PACE
TVA
WACM
LGEE
DUK
SRP
AECI
NWMT
OVEC
PNM
AZPS
PSCO
SC
CPLE
FPC
TEPC
SCEG
LDWP
SEC
FMPP
JEA
BPAT
TEC
EEI
NEVP
SPA
PGE
ISNE
WALC
HICC
AEC
NYIS
ASCC
GVL
CISO
FPL
IPCO
Cooling Water Withdrawn
for Coal Power Plants
(Million Cubic Meters)
Figure 3.4: (Top) Annual coal-red electricity generated in each balancing authority in 2017.
(Middle) Amount of water consumed to cool coal-red power plants in 2017. (Bottom) Amount
of water withdrawn to cool coal-red power plants in 2017. For each plot, the bars are ordered
by greatest to least coal-red electricity generation.
The interchange of electricity across the US grid means coal-red electricity could be imported
into balancing authorities that generated little-to-no coal-red electricity. For example, the Cal-
ifornia Independent System Operator (CISO) balancing authority generated very little coal-red
electricity (the coal generation in CISO in 2017 came from a co-generation plant that burned coal,
but was not primarily a coal-red power plant), however, electricity was imported from balanc-
ing authorities with signicant coal-red electricity, such as the Arizona Public Service Company
(AZPS) and Los Angeles Department of Water and Power (LDWP) (Figure 3.5). CISO and
MISO had the highest imports of coal-red electricity in 2017, each having imported 13 TWh of
coal-red electricity (Figure B.1) that amounts to 98% of CISOs coal-red electricity consumption
48
and 4% of MISOs coal-red electricity consumption. The greatest amount of virtual cooling water
transferred between balancing authorities in 2017 aligned with the greatest amount of electric-
ity transferred, which was from LDWP to CISO (6.6 TWh). (LDWP, which provides electricity
solely to the Los Angeles metropolitan area, owns a 1.9 GW coal power plant in Utah that con-
tributed approximately 38% of LDWPs electricity generation in 2017). The largest
ow of virtual
cooling water withdrawals was associated with the 6 TWh of coal-red electricity transferred
from the Ohio Valley Electric Corporation (OVEC) balancing authority to the Pennsylvania-New
Jersey-Maryland Interconnection (PJM) balancing authority. The OVEC balancing authority is
comprised of two power plants, each of which is coal-red and greater than 1 GW capacity. Both
of OVECs power plants use once-through cooling systems, resulting in a high water withdrawal
intensity for the balancing authority.
AEC 0
AECI
0
AVA 0
AZPS
0
9
BANC 0
BPAT
0
9
CISO
0
9
CPLE
0
CPLW 0
DUK
0
EEI
0
EPE
0
ERCO
0
FMPP
0
FPC
0
FPL
0
GCPD
0
GVL
0
GWA
0
HGMA
0
IID
0
IPCO
0
ISNE
0
JEA
0
LDWP
0
9
LGEE
0
MISO
0
9
18
NEVP
0
NSB
0
NWMT
0
NYIS 0
OVEC
0
PACE
0
9
PACW 0
PGE 0
PJM
0
9
18
PNM
0
PSCO
0
PSEI 0
SC
0
SCEG
0
SCL 0
SEC
0
SEPA
0
SOCO
0
SPA
0
SRP
0
SWPP
0
TAL
0
TEC
0
TEPC
0
TPWR
0
TVA
0
9
WACM
0
9
WALC
0
WAUW
0
YAD
0
BCHA
0
CFE
0
CEN
0
MHEB
0
AESO
0
SPC
0
(a)
AEC 0
AECI
0
600
AVA 0 AZPS
0
600
1200
BANC
0
BPAT
0
600
CISO
0
CPLE
0
CPLW
0
DOPD 0
DUK
0 EEI
0
600
EPE
0
ERCO 0
FMPP 0
FPC 0
FPL 0
GCPD
0
GRMA
0
GVL
0
IID
0
IPCO 0
ISNE
0
JEA
0
LDWP
0
LGEE
0 MISO
0
600
1200
NEVP
0
NSB
0
NWMT
0
NYIS 0
OVEC
0
600
PACE
0
PACW
0
PGE
0
PJM
0 600
1200 1800
PNM
0
PSCO
0
PSEI
0
SC
0
SCEG
0
SCL 0
SEC 0
SEPA 0
SOCO
0
SPA
0
SRP
0
SWPP
0
TAL
0
TEC
0
TEPC
0
TIDC
0
TPWR
0
TVA
0 600 1200
WACM
0 600
WALC
0
WAUW
0
YAD
0
BCHA
0
GRID
0
CFE
0
CEN
0
MHEB
0
AESO
0
SPC
0
(b)
Figure 3.5: (a) Exchange of coal-red electricity between balancing authorities in 2017 in units of
TWh. (b) Exchange of virtual water withdrawals associated with cooling coal-red power plants
between balancing authorities in 2017 in units of million cubic meters. The color of each chord
represents the balancing authority where electricity is being exported from.
49
3.3.1 Uncertainties and Limitations
One of the main uncertainties in our accounting of water consequences related to coal mining and
preparation comes from the nature of coal stocking; since coal stocks can be stored onsite for
long durations of time, we cannot know with complete certainty how much coal is actually from
which region. Another uncertainty is in our accounting of rened coal consumption. The use of
rened coal for electricity generation increased by over 2,000% (from 52 thousand short tons to
150 million short tons) between 2014 and 2018 [43]. In 2014, rened coal represented 0.01% of
coal consumption for electricity generation. In 2017, that number increased to 19%. Thus, rened
coal has become a signicant portion of coal consumption nationally. We did not consider how
the water associated with rening coal would impact our results.
Another uncertainty is in the quantication of how much of the electricity transferred between
balancing authorities came from a coal-red power plant. In this analysis, we approximated the
amount of coal-red electricity transferred using the percentage of coal-red electricity generation
within the exporting balancing authority. One of the shortcomings of this approach is that we
cannot know how much electricity from a particular power plant is transported across the grid
at any point in time. However, given available data, using an approximation is the only way
to comprehensively account for fuel source of electricity transferred. Another shortcoming of
our approximation method is it cannot account for the coal types or the mining sources of the
coal burned within the exporting balancing authority. The water intensities derived for coal-red
electricity generation within each balancing authority are based on the prime movers and cooling
technologies used by the coal power plants in the respective balancing authority. Although the
cooling water intensities do not account for variations in coal types and coal mining source, cooling
water usage is more dependent on cooling technology (once-through versus recirculating cooling
system) than the coal burned [36].
50
3.3.2 Implications
The ndings for this work are based on fuel consumption, electricity generation, and electricity
consumption in 2017. Shifts in the power sectors fuel mix and cooling technologies will in
uence
the water used for electricity consumption across the country in the future. Competitive market
prices for natural gas and renewable portfolio standards in many states will likely lead to a
continued decrease in coal consumption in the US [36]. The electricity sectors trend towards
using recirculating cooling over once-through cooling will impact the water withdrawal intensity
of electricity generation [45]. At the policy level, power plant cooling is aected by the Clean
Water Act (CWA) 316(a), which regulates the temperature at which water can be discharged
back into the environment after being used to cool thermal power plants [47]. Implementation
of the CWA 316(a), which varies between states, can in
uence how power plants operate (e.g.,
curtailments or shutdowns during high water temperatures) as well as the cooling systems utilized
by the power plants. For example, Californias regulations on ocean water withdrawals has led to
a phase out of once-through ocean cooling systems in favor of recirculating towers or dry-cooling
systems [98].
One implication of our study is that water footprint accounting at the producer level versus
the consumer level could lead to dierent results. In Figure B.2, we show that accounting for the
water use for coal-red electricity generated within each balancing authority is slightly dierent
than accounting for water use for coal-red electricity consumed within each balancing authority.
Because of the interconnected grid, accounting for water use for coal-red electricity consumption
crosses many balancing authority (and state) boundaries. For example, a region with no coal
generation could have virtual water withdrawals and consumption associated with its electricity
consumption due to imported coal-sourced electricity. Similarly, a region that exports a signicant
amount of its coal-red electricity would have a lower net water use if accounting for electricity
51
consumption (i.e., total water use for coal-red generation less the virtual water for coal-red
exports) than if accounting for electricity production.
3.4 Conclusions
The trade of energy (primary and secondary) complicates the interactions between energy produc-
ers and energy consumers. In the case of electricity consumption, primary energy is transported
to the point of electricity generation, and the generated electricity is transported to the nal
consumer. While previous studies have either quantied (a) the life cycle water footprint of elec-
tricity generation or (b) the virtual water associated with cooling thermoelectric power plants
embedded in electricity transfers, few studies have combined the two to create a spatially-resolved
assessment of the life cycle water use embedded in energy, including traded energy. We assessed
the life cycle water use of coal-based electricity consumption in the US. We focused our analysis to
include the production of coal, preparation of coal, burning of coal for electricity, and the trans-
port of electricity. Region-specic water intensities were applied to coal mining processes, and
technology-specic water intensities are applied to the electricity generation process to determine
the amount of water consumed and withdrawn.
With a focus on US coal consumption in 2017, we nd that although most of the coal consumed
for electricity came from the Northern Great Plains, the largest percentage of the water use
impacts for mining and preparing coal came from coal mined in the Interior province. Our results
also indicate that due to the interconnected nature of the electric grid, regions that do not generate
coal-red electricity can withdraw and consume cooling water virtually through the import of coal-
sourced electricity from other regions. The attribution of water consumption for energy can vary
depending on where the impacts are incurred and where the energy is actually consumed.
52
Chapter 4
Integrating water, energy, and climate modeling to assess
vulnerabilities to the US Southwest power grid
4.1 Introduction
The US electricity sector is heavily dependent on water resources. Hydroelectric generation, which
makes up 6% of total US electricity generation, requires water for spinning turbines [4]. Ther-
moelectric generation, which contributes to 82% of total national electricity generation, requires
large quantities of water for cooling [4]. Electricity generation can hence be constrained by wa-
ter and climate variables such as: water availability, intake water temperature, discharge water
temperature, and air temperature.
Reductions in water availability decrease the generation capacity of the electricity sector.
Changes in reservoir storage can reduce the magnitude of hydropower potential [99, 100]. De-
creasing stream
ows can cause thermal power plants, which use water withdrawn from a river
or pond to condense steam exiting turbines, to curtail or shut down operations [101]. In a once-
through (open loop) cooling system, a large quantity of water is removed from a source, used to
condense the steam exiting turbines, then returned [31, 32]. Once-through cooling systems with-
draw large quantities of water, but minimal amounts are actually consumed. In a recirculating
(closed-loop) cooling system the withdrawn water is reused after transferring heat to the air by
53
way of evaporation rather than being discharged back to the source [31, 32]. Recirculating cooling
systems withdraw much less water than once-through systems, but most of the water is consumed.
Intake water temperature can have great implications on the eciencies of thermoelectric
generators. In a thermoelectric generator, fuel is burned to create steam that spins the turbine
to generate electricity. After the steam leaves the turbine, it is cooled to its liquid state by
transferring heat to a cooling water system. The eciency of a thermoelectric generator is driven
by the temperature dierential between the steam and the cooling water temperature. As coolant
(water) temperature increases, the generator eciency decreases as well [102, 103]. After cooling
water is utilized, the fraction that is not evaporated is discharged back to the water source
at a higher temperature, which can pose risks to aquatic ecosystems [104]. As a result, when
the temperature of discharged cooling water, or the temperature dierential between discharged
euent and ambient water becomes too high, power plants are subject to operational curtailment
or shut-down according to the thermal discharge limits set by the Clean Water Act (CWA)
section 316(a) [47]. Ambient air temperature can also in
uence thermoelectric generation, albeit
to a lesser extent than water variables.
The extent and intensity of climate change impacts on energy systems have been widely re-
searched in recent literature. Simulated studies suggest that the US electricity grid is vulnerable
to climate change due to the combined eects of decreased water availability, higher water tem-
peratures, and regulatory enforcement. Van Vliet et al. used a hydrology-electricity modelling
framework to show that during the period spanning 2031-2060, approximately 4.4-16% and 6.3-
19% of thermoelectric generation in the US and Europe, respectively, could be lost to rising
temperatures and intensifying drought [11]. They also found that by the 2050s, thermal power
plants across the globe could experience capacity reductions of 7-12% [12]. Sathaye et al. utilized
temperature projections from global climate models to estimate changes in natural gas power
plant capacity in California and found that 1.1-4.6% of peak capacity could be lost by the end
of the century [13]. Bartos and Chester combined climate, hydrology, and thermodynamic power
54
plant models and estimated that 1-3% of summer generating capacity would be decreased by
mid-century in the Western US, with reductions up to 7-9% when analyzed within the context
of a ten-year drought scenario [105]. Cook et al. used a combination of regression, climate, and
thermodynamic modelling to nd that a 1

C rise in cooling water temperature can lead to a
0.15%-0.5% decrease in power output [14]. Liu et al. assessed the impact of climate change and
thermal discharge regulations on thermoelectric generators in the US using a regional earth sys-
tem model and a thermoelectric power generation model. The study found that by the 2060s,
climate change alone could reduce generation capacity by 2-3%, but environmental regulations
could actually reduce capacity up to 12% if power plant operators are forced to curtail operation
when water discharge temperatures exceed legal limits [15].
Existing modeling studies of climate change impacts on the electricity sector typically include
one or more of the following impacts on electricity generation: thermoelectric capacity loss due to
water constraints and ambient temperature
unctuations, hydropower capacity reductions due to
water unavailabiltiy, and regulatory constraints on plant operations. Most simulated studies look
at projected changes to operational capacity of the electricity grid, but no study has investigated
how high-resolution grid dispatch is aected by ambient temperature changes, water availabil-
ity constraints, and regulations. Additionally, most modeling studies utilize large hydrological
models to predict future hydroelectric capacity, but few have utilized basin operations models to
simulate local hydroelectricity generation at high temporal resolution. In this study, we develop a
framework integrating downscaled data from climate and hydrological models, as well as a reser-
voir operations model, as physical constraints in an electricity dispatch model. This framework
allows us to evaluate the impacts of changing climatic conditions and water usages on electricity
generation in a major provisioning basin. We are also able to assess how changes in the local
electricity generation impact the larger electricity grid system.
55
4.2 Methods and Materials
4.2.1 Study Site
The San Juan River (SJR) basin was selected to investigate the in
uences of variability in climate
parameters and water resources on local to regional electricity generation. The SJR basin is
located in the Four Corners region of the southwestern US (Figure 4.1). This basin is of particular
interest as it is characterized by three distinct scales of operations: a small local utility that services
most of the basin's electricity needs, two local power plants that supply electricity to utilities in
basins external to the SJR, and the Western Electric Interconnection with which the basin's plants
interact.
The SJR basin is also of interest as it is a critical tributary to the Colorado River and provides
interbasin transfers to the Rio Grande [106]. The SJR basin is a headwater system that
ows into
the Colorado River Basin (CRB) at Lake Powell. The SJR basin covers approximately 65,000
km
2
, contributes 15% of the stream
ow to the Upper CRB at Lees Ferry, and extends over
10% of the entire CRB. Monthly average temperatures range from -2.5

C in January to 22.7

C in July (2001-2010) [107], while precipitation follows a bi-modal pattern, with the highest
precipitation falling during the summer months when short-duration, intense rainfall events occur
owing to remnant monsoonal weather patterns. During winter, precipitation falls largely as snow
in the upper headwaters of the SJR, contributing the second highest volume of precipitation to
the basin. Spring melt occurs historically during May and June in the SJR basin.
The Navajo Nation, the largest tribal nation in the US and one of the most signicant water
users in the SJR basin, entered into a settlement agreement with the US government and State of
New Mexico regarding the tribes water rights in the basin in 2005, which was signed into federal
law in 2009. The Navajo Nations allocation of water in the basin is currently being utilized
(current use) largely for agriculture in the Navajo Indian Irrigation Project (NIIP). Pursuant to
the settlement, water use by the Navajo Nation will increase (future use) with the expansion of the
56
Figure 4.1: The San Juan River basin in the Four Corners region [106]. Major features are
labelled, including the Colorado River and Rio Grande basins, Lake Powell, the Navajo Reservoir,
Heron Reservoir, and the major tributaries to the San Juan. Power plants (coal = black, natural
gas = red, hydropower = blue) in the region are shown. Arrows indicate the major exports of
energy and water external to the basin.
57
NIIP and completion of the Navajo-Gallup Pipeline Project, which is currently under construction
and is expected to come online in 2024.
The second largest non-tribal water use in the SJR basin is for cooling two coal-red power
plants in the basin. The Four Corners Generating Station is a coal-red power plant located
near Farmington and run by the Arizona Public Service Company (AZPS), to largely supply the
electricity needs of the Phoenix metropolitan area. Four Corners utilizes a recirculating cooling
system with a cooling pond. The San Juan Generating Station is a coal-red plant operated by
the Public Service Company of New Mexico (PNM), and provides approximately one-third of the
power supplied to the state of New Mexico. The San Juan Generating Station uses a recirculating
cooling system with induced draft cooling towers. There are three smaller (each with nameplate
capacities of less than 65 MW) natural gas, combined-cycle power plants in the basin. The SRJ
basin also includes ve run-of-the-river hydropower facilities that make up a combined nameplate
capacity of 53.2 MW. All power plants are listed in Table 4.1. The power plants in the SJR basin
are operated as part of the Western Electricity Coordinating Council (WECC).
Table 4.1: List of power plants in San Juan River Basin.
Plant
Nameplate
Capacity (MW)
Fuel Prime
Mover
Cooling
System
Balancing
Authority
Four Corners 1636.2 CL ST RC AZPS
San Juan 1848.0 CL ST RI PNM
Animas 24.6 NG CC RF WACM
Bluview 64.0 NG CC RF WACM
Milagro 60.8 NG CC NA WALC
Navajo Dam 30.0 HYD HYD NA WACM
Tacoma 4.6 HYD HYD NA WACM
McPhee 1.3 HYD HYD NA WACM
Towaoc 11.5 HYD HYD NA WACM
Vallecito 5.8 HYD HYD NA WACM
Fuel: CL (coal), NG (natural gas), HYD (hydroelectric)
Prime Mover: ST (steam), CC (combined cycle), HYD (hydroelectric)
Cooling System: RC (recirculating with cooling pond(s) or canal(s)), RI (recirculating
with induced draft cooling tower(s)), RF (recirculating with forced draft cooling tower(s))
58
4.2.2 Reservoir Operations
Navajo Reservoir is operated to meet irrigation demands, municipal and industrial demands, as
well as to satisfy downstream environmental commitments. Releases are made to satisfy the
minimum downstream target base
ow of 14.2 m
3
/s. The model then calculates the amount
left over each year in the reservoir above an end-of-year target of 1.84 km (6,053 feet). That
volume is shaped into a spring
ush as recommended by the San Juan River Basin Recovery
Implementation Program. In years that projected supply cannot meet all of the project objectives
without reducing the reservoir to an elevation at or below 1.83 km (5,990 feet), it triggers a set
of rules (Shortage Sharing Agreement) that proportionally split the shortage volume between the
downstream release, the diversion for the NIIP, and Rio Grande Basin deliveries.
The SJR basin is discretized into fourteen reaches dened by gages located along the reaches.
Losses in the reaches between gages are used to represent collective losses due to diversions,
evapotranspiration, and groundwater loss, as well as gains from minor tributary in
ows and
return
ows. The losses and gains between each reach are based on historical statistics and time
of year.
4.2.3 Modeling system
Climate impacts on tributary in
ows are simulated using the Variable Inltration Capacity (VIC)
surface hydrology model. In turn, VIC-calculated
ows inform the river and reservoir routing
model (RiverWare
TM
). A more detailed description of the climate and water model is provided by
Bennet et al. 2019 [106]. Out
ows from RiverWare
TM
are used to constrain hydropower capacity
in the basin, and water levels are used to calculate reductions in annual water withdrawals for
thermal power plants.
59
4.2.3.1 Model simulations of the climate and water systems
Two dierent water use scenarios, a current-use case and a higher-use case that assumed a full
buildout with full utilization of existing water rights (including tribal reserved water rights result-
ing in about a 20% increase in use), were considered. We also considered climate-induced changes
to vegetation.
Base: Historical climate conditions with current water uses
Future Climate (FC): Future climate conditions with current water use and without vege-
tation change
Full Water Use (WU): Future climate with full utilization of existing water rights (once full
build-out of NIIP and the Navajo-Gallup Pipeline are completed) and without vegetation
change
Vegetation and Full Water Use (DCU): Future climate and vegetation change with full
utilization of water rights
The Base scenario is re
ective of historical 2010 conditions. For the three remaining scenarios,
the years 2044, 2064, and 2084 were chosen to re
ect a range of climatic conditions (no drought,
moderate drought, and severe drought, respectively).
To model potential future climate scenarios and impacts to water supplies, we utilized the
HadGEM2-ES365 Earth System Model (ESM) simulations in the Coupled Model Intercomparison
Project, Phase 5 CMIP5 [108]. The ESM simulations were downscaled using the Multivariate
Adaptive Constructed Analogues (MACA) technique [109]. A single emission pathway was utilized
in this work, Representative Concentration Pathway (RCP) 8.5, which tracks current emissions
when land use changes are included [110].
To assess how changes in climate will impact stream
ow and reservoir evaporation, the Variable
Inltration Capacity (VIC) hydrologic model was used. VIC is a semi-distributed, empirically
60
based hydrological model with physical components that simulate full energy and water balances
for grid cells within a river basin [111] VIC version 4.2 was implemented in full energy balance mode
at a 1/16th degree grid scale resolution (7 x 7 km) with an hourly timestep. To simulate the timing
of stream
ow, runo and base
ow were routed from the grid cells via an oine routing model
described by Lohmann, Nolte-Holube and Raschke 1996 [112]. Lake evaporation was estimated
using VICs lake module.
RiverWare
TM
was used to simulate the operations of water management infrastructure and
water deliveries throughout the basin. RiverWare
TM
is an object-oriented river and reservoir
modeling tool that can be set up to simulate the operations of physical and structural features
(e.g., dams, reservoirs, rivers, tunnels, diversions, water users, power generators) in a specic
basin according to established management policies. Here we use the San Juan River Basin
Daily Operations Model (created in RiverWare
TM
) which was developed by local water managers
to assist with short-term, day-to-day operations. The model simulates the entire SJR basin,
including its upper basin reservoirs (Lemon, Vallecito and Nighthorse).
Water availability, or shortages, to the SJR basin were determined using the San Juan-Chama
Project model, a model based the Upper Rio Grande Water Operational Model (URGWOM).
Computed deliveries consider annual water allotments made from the San Juan River system,
available supply at Heron Reservoir and water allotments made to the downstream San Juan-
Chama Project contractors. Shortages to water users in the SJR basin were analyzed at the
beginning of each calendar year, starting in 1976, when Heron reservoir was initially lled. Addi-
tional details for both RiverWare
TM
models are included in Bennet et al. 2019 [106]. A complete
list of the scenarios analyzed and the corresponding annual water withdrawal limit for power
plants (due to water shortages) is provided in Table 4.2.
61
Table 4.2: Scenarios studied in analysis, along with the water otake limits enacted in each scenario. The "FC", "WU" and "DCU" denotes
the Future Climate, Full Water Use, and Vegetation and Full Water Use scenarios, respectively.
Scenario Future
Climate
Full Water
Rights Use
Climate-Induced
Vegetation
Disturbance
Study
Year
Four Corners Annual
Water Withdrawal Limit
(Million Gallons)
San Juan Annual Water
Withdrawal Limit
(Million Gallons)
Base % % % 2010 No limit No limit
2044-FC " % % 2044 No limit No limit
2064-FC " % % 2064 No limit No limit
2084-FC " % % 2084 7,161 10,600
2044-WU " " % 2044 No limit No limit
2064-WU " " % 2064 6,715 9,941
2084-WU " " % 2084 2,022 2,994
2044-DCU " " " 2044 No limit No limit
2064-DCU " " " 2064 6,423 9,508
2084-DCU " " " 2084 1,426 2,111
62
4.2.3.2 Model simulations of electric system
Impacts on electricity grid operations were simulated using a 2010 electric grid infrastructure
of the WECC interconnection in Energy Exemplar's PLEXOS Integrated Energy Model [113].
PLEXOS is a commercial production cost model that models the unit commitment and dispatch
of generators in the electric power system. PLEXOS solves 365 daily steps of 24, 1-hour intervals.
For every hour of the year the optimization algorithm in PLEXOS seeks to minimize the overall
cost of operating the system. This 2010 representative database was developed by modifying the
Transmission Expansion Planning Policy Committee (TEPPC) 2024 Common Case to represent
the 2010 grid conguration, while incorporating changes from a prior study of the western US
[114]. The TEPPC database has commissioning and retirement dates for most generators; missing
data from the TEPPC database was supplemented using US Energy Information Administrations
Form EIA 860 dataset [20]. The modeled 2010 generation
eets capacity was compared to histor-
ical capacity from the EIA and outage rates of both coal and nuclear generators were increased
so available generating capacity matched historical data. All post-2010 changes and generator
retirements were removed from the TEPPC 2024 database to revert to a 2010 representation.
The water withdrawal rate for the Four Corners power plant was calculated by dividing water
diversion amounts reported by the power plant's operator for July 2015 by the plant's net gener-
ation in July 2015, as reported by the EIA Form 923 [43]. The water consumption rate for the
Four Corners plant is based on monthly reported evaporation rate by the plant operator. Thus,
dierent consumption rates were used for the Four Corners generators in each month. The San
Juan power plant is a zero liquid discharge plant, so all water withdrawn for cooling is consumed.
The San Juan power plant's water consumption (and withdrawal) rate for each month was pro-
vided by the plant's operators. The reported monthly water rates were treated as the average
water rate.
63
Water shortage information from each scenario (as described in Section 4.2.3.1) were imple-
mented into PLEXOS to constrain (a) the amount of water withdrawn by the Four Corners and
San Juan power plants and (b) hydroelectric generation output. Water withdrawal and water
consumption rates were incorporated into PLEXOS as emission rates for each of the thermal
power plants' generators. Because the water withdrawal rates were treated as emission rates in
the PLEXOS model, we constrained water withdrawal by placing a \cap" on water withdrawals.
The aects of ambient temperature
unctuations on thermoelectric generator eciency were
also incorporated into the PLEXOS model. General Electric's GateCycle is a heat and mass
balance modeling tool used to analyze the design, simulation and performance of power plant
equipment. We utilized GateCycle to model changes in eciencies for thermoelectric generator
congurations (fuel type, nameplate capacity, and cooling type) at various ambient temperatures.
Using the temperature-eciency relationships from GateCycle and daily temperatures at the SRJ
basin from the climate models, we incorporated daily scalar values into PLEXOS that altered the
heat rates of the thermoelectric generators in the SRJ basin for each day of the year.
We used the daily out
ows at the hydropower facilities from the RiverWare
TM
as inputs into
the PLEXOS model. Power curves provided by the hydropower facility operators were used to
determine maximum capacity based on daily out
ows. The daily maximum capacity values were
then implemented into PLEXOS as constraints for the hydropower plants in the SRJ basin.
4.3 Results
4.3.1 San Juan River Basin thermoelectric generation
Most of the basin's generation comes from the Four Corners and San Juan power plants, so the
total generation remains relatively constant across scenarios when there are no limits placed on
annual water withdrawal at the coal plants. In the scenarios where there are no water withdrawal
constraints, the Four Corners and San Juan power plants do not experience signicant changes
64
in generation (Figure 4.2). However, when water withdrawal constraints are placed on the Four
Corners and San Juan power plants, the generators ramp up and shut down, while decreasing
their operating capacity. In the scenarios with moderate water withdrawal limits, the generating
capacity from the coal power plants do not change signicantly, even during the ramping up
and shutting down. However, for the extreme scenarios, the Four Corners and San Juan power
plants decrease their operating capacity signicantly (in addition to the ramping). Operations
at the smaller natural gas combined-cycle plants are not expected to suer similar limitations as
their water is serviced through municipal water systems which are expected to address shortages
through reduced outdoor irrigation. Note that climate and vegetation impacts are relatively
muted on the basin as the Navajo Reservoir is often able to smooth out variability associated
with climate alone. However, it is only when a consistent increase in water use is coupled with
increased climate variability when water shortages are induced [106].
The cost of generation includes the start and shutdown cost, the fuel cost, and the variable
operations and maintenance (VO&M) cost. Along with regular cost of generation, we also analyzed
the normalized cost, which was calculated by dividing each power plant's dollar amount per cost
type by the respective power plant's annual generation. The cost follows the trend of generation
closely, since fuel cost makes up the majority of total cost. The change in generation cost from the
Base scenario ranges from 0.06% to -82% (Figure 4.3). For both coal plants, the normalized costs
remain mostly constant across scenarios, except for the visible changes in startup and shutdown
costs (Figure 4.3). The change in normalized startup and shutdown costs (compared to the Base
scenario) range from a decrease of 0.08% to an increase of 308% for the Four Corners facility
and a decrease of 0.11% to an increase of 195% for the San Juan power plant. The increase in
normalized start and shutdown cost is a result of the coal generators undergoing more ramping
in the scenarios with water withdrawal constraints. Because fuel cost makes up the majority
of generation cost, the absolute fuel cost for both the Four Corners and San Juan power plants
65
Figure 4.2: Daily generation for thermoelectric and hydroelectric power plants in the San Juan
River Basin. For scenarios without water withdrawal constraints, the power plants do not adjust
their operation signicantly and operate at their nameplate capacity for most of the year. However,
when there are limits placed on annual water constraints, power plants ramp up and down much
more often, in order to meet the water constraint scenarios.
66
Figure 4.3: (top) Annual generation of power plants in San Juan River Basin for all scenarios.
(middle) Annual normalized generation cost (bar graph, left axis) and absolute generation cost
(point plot, right axis) for Four Corners power plant. (bottom) Annual normalized generation
cost (bar graph, left axis) and absolute generation cost (point plot, right axis) for San Juan power
plant.
67
decrease in the water-constrained scenarios. However, the normalized fuel cost is higher in water-
constrained scenarios due to decreased generator eciency during ramping. At low capacity,
generators are often much less ecient than if operating at their maximum or optimal capacity.
Because the generators are ramping up and down much more and operating at reduced capacity,
the units are operating less eciently for longer periods of time, requiring more fuel to generate
the same amount of electricity.
4.3.2 San Juan River Basin hydroelectric generation
The hydropower facilities experience reductions in generation across all scenarios. In the 2044
scenarios, there is only a slight decrease in hydroelectric generation, but in all other scenarios
hydroelectricity generation is signicantly less than the Base scenario (Figure 4.3). Unlike the coal
power plants in the SJR basin, which
unctuate generation day-to-day, the hydropower facilities
experience large reductions in generation over the span of multiple days at a time, especially in
the summer months.
4.3.3 WECC-wide
In addition to analyzing the changes in generation in the SJR basin, the electricity simulation
model allows us to assess how generation reductions in the basin impacts the rest of the WECC
region. The decreases in generation from the San Juan and Four Corners power plants resulted
in other generators in WECC having to increase their output to meet electricity demand in the
region. While the total generation across WECC in each scenario did not change (since demand is
the same across all scenarios), the amount of electricity generated per fuel type did vary. Most of
the lost coal generation is accommodated by mostly natural gas combined-cycle systems (Figure
4.4).
The change in generation across WECC also result in increases in generation costs (Figure
4.4). In a unit commitment and dispatch model, such as the one utilized for our electricity grid
68
simulation, the model optimizes the grid to generate at the lowest cost possible. As such, coal
power plants, such as the San Juan and Four Corners power plants, are considered cheaper to
operate and therefore dispatched in the Base scenario. However, when the operations of the SRJ
basin coal plants are constrained by the water withdrawal limits, other generators that cost more
to operate have to be dispatched to meet electricity demand. Additionally, because the price
of natural gas is greater than coal in our electricity model assumption for 2010, the increase in
natural gas generation led to a net increase in fuel costs in scenarios where the lost coal generation
in the SRJ basin had to be replaced with natural gas generation elsewhere (Figure 4.4).
Analyzing results at the balancing authority level, the most visible changes are the decreases in
coal generation in AZPS and PNM, where the Four Corners and San Juan generating stations are
located, respectively (Figure 4.5) (For a plot containing all balancing authorities, refer to Figure
C.3.) Some of the supplementary generation came from within AZPS and PNM, but most of the
lost generation is oset by generators in other balancing authorities. The balancing authorities
that experience increases in electricity generation also see increases in generation cost (Figure
C.4).
4.4 Discussion
4.4.1 Impacts of climate change and water constraints on basin-level
electricity generation
In this study, we modeled how electricity generation in a local river basin is impacted by climate
change, water use rights, and water constraints due to regulation enforcement. In scenarios where
only future climate conditions are considered (not future water use or water restrictions), we do
not see signicant changes in generation from the San Juan and Four Corners power plants. In
scenarios where water withdrawal constraints are placed on the power plants due to (a) increased
water use in the basin and/or (b) climate-induced water reductions, the coal power plants have
69
Figure 4.4: (top) Dierence in annual generation in WECC for all scenarios, compared to Base
scenario. Each color represents a fuel type. (bottom) Dierence in generation cost in WECC for
all scenarios, compared to the Base scenario. Each color represents a dierent cost type.
70
Figure 4.5: Annual generation in WECC balancing authorities for all scenarios. The top row is the
annual generation in the base scenario, and all subsequent rows show the dierence in generation
from the base scenario.
71
to change their generating patterns in order to comply with the Shortage Sharing Agreement.
Specically, the San Juan and Four Corners power plants ramp generation up and down in order
to reduce total annual generation and meet water withdrawal limits. This result indicates that
policy constraints, including at the local level, can have great in
uence on power plant operations.
Thus, careful consideration of local water use regulations are important when modeling climate
change impacts on the electricity grid.
The hydroelectric power plants experience decreases in generation when there are less out
ow
in the basin during the summer months. Because the hydroelectric facilities in the SJR basin are
all run-of-the-river, they are largely dependent on water availability (out
ow) and not so much
on operation constraints like the thermoelectric generators. Whereas the coal power plants in the
basin have to consider their annual water withdrawal and adjust their generation accordingly, the
output of the hydroelectric power plants are re
ective of the out
ow in the basin.
4.4.2 Impacts of local electricity generation changes on grid-wide elec-
tricity generation
In the scenarios where the San Juan and Four Corners power plants experience reductions in
generation due to water constraints, other generators throughout the WECC region have to be
dispatched in order to meet electricity demand. It is also important to consider the environmental
impacts of the shifted generation throughout WECC. While there is an overall decrease in CO
2
emissions by burning natural gas rather than coal, the spatial and temporal context of CO
2
does
not matter as much as when analyzing criterial pollutants and water resources. When considering
air pollutants such as NO
X
(Figure C.6) and SO
2
(Figure C.7), the location of the increased
pollution is critical in assessing the level of impact on nearby communities and the environment.
Additionally, the time of increased generation matters when considering possible formation of
smog, which would be dependent on sunlight exposure. Finally, our results for the SJR basin
emphasize the importance of water as a local resource and the complexities of managing water
72
for dierent users. Such that thermoelectric generation impacts water use in the SJR basin,
the increased generation in other balancing authorities would impact water resources in those
respective areas. The level of impact of that increased generation, however, would depend on
water availability and water usage in the respective areas.
4.4.3 Limitations
The framework developed did not include a river temperature model, which prevented us from
being able to analyze the impacts of water intake temperature constraints on power plant oper-
ations. Furthermore, we do not consider water discharge regulations or limits, which can have
signicant in
uence on generator dispatch, especially for generators with once-through cooling
systems [15, 115]. Cooling water temperatures at power plants will be incorporated in future
developments of the framework.
4.5 Conclusions
Water, energy, and climate systems are heavily dependent on one another, yet modeling all three
and their interdependencies is very complex. The various levels of scale { from global climate
change, to basin-level water management, to regional electricity grid operations { makes it dif-
cult to understand how all these systems interact with one another. We develop a framework
integrating data from climate, hydrological, and reservoir operations as constraints in an elec-
tricity market simulation tool. We assessed local changes in electricity generation due to climate
change and water constraints, as well as the cascading impacts on the rest of the western electricity
grid.
Results indicate that under most of the scenarios, when there are only changes in tempera-
tures and no constraints on water withdrawal, the SJR basin's thermoelectric generators operate
similarly to how they usually do. However, when water withdrawal limits are placed, the coal
73
power plants reduce their capacity and undergo generation ramping. Reduced generation from
the SJR basin causes other power plants in the western US to increase their generation in order
to meet grid-wide electricity demand. Most of the increased generation is provided by natural
gas combined-cycle systems. The replacement of coal generation by natural gas generation means
there is a net decrease in GHG emissions and air pollutants in WECC, but the level of impact of
these environmental changes will depend on where the increase generation occur.
This study illustrates the importance of integrated modeling in capturing the interlinkages
between water, energy, and climate systems. With climate change expected to increase tempera-
tures and reduce water availability in many regions, using detailed climate and basin models will
allow for the assessment of how variability and changes in climate conditions will in
uence the
eciency of generators. The use of a high-resolution electricity market simulation model allows
for the capture of technical and economic factors that impact the electricity grid. The framework
developed can be modied to incorporate dierent models and applied to other regions. Studying
other locations will also allow for the analysis of dierent policy scenarios, including water-related
regulations that are specic to certain localities.
74
Chapter 5
Conclusion
The primary objective of the research described in this dissertation was to create holistic frame-
works for modeling how energy systems operate and impact the environment within the contexts
of historic climate conditions and future climatic change. The US energy system imparts a num-
ber of environmental externalities, including water consumption, across many lifecycle stages:
primary fuel extraction, thermoelectric power plant cooling, and hydroelectricity generation. Cli-
matic
unctuations and water constraints can lead to vulnerabilities and reduced eciencies in the
energy system. Modeling the interactions between water, energy, and climate systems requires
the careful consideration of temporal and spatial scales. This body of work addresses several
knowledge gaps in the literature:
1. There is no empirical analysis investigating the relationship between ambient temperature
and power plant performance for a statistically signicant number of power plants. My work
addressed this gap by investigating how real-world generation units respond to temperature
changes, as the operational and climatic variables aecting the performance of power plants
are typically too complex to capture in purely theoretical models.
2. There can be large spatial decouplings across the energy supply chain, as energy production,
processing, conversion, and consumption can occur in disparate locations that make it di-
cult to quantify the environmental damages imposed by energy consumption. My research
75
spatially resolved the water consumed and water withdrawn for coal mining, coal prepara-
tion, and power plant cooling from (a) where the coal is mined to where the coal is burned
for power production and (b) where the electricity is generated to where the electricity is
consumed.
3. The impacts of climate-induced changes in temperature and water resources on the power
system operations have not been investigated in high resolution. My research addressed this
gap by developing a modeling framework to identify vulnerabilities in a local and regional
power grid due to projected changes in climate conditions and regional water supply and
demand.
Chapter 2 developed regression models to estimate changes in the eciencies of over one
thousand coal and natural gas generators as functions of ambient air temperature and operational
variables across dierent fuel types, prime movers, cooling systems, and climate zones. The
eciencies of generators with dry cooling, particularly those in hot and dry climates, demonstrated
the greatest sensitivity to increases in ambient temperature. Results for generators utilizing wet
cooling systems were largely inconclusive, most likely because other factors, such as cooling water
temperature, are better predictors of generator eciency. Out results also indicated that natural
gas combustion generators in hot and arid regions likely utilize inlet air cooling technologies to
reduce the temperature of ambient air before it enters the compressor, thereby mitigating eciency
losses. This study was the rst to statistically analyze the impacts of ambient temperature
and operational variables on generator eciency with such a large sample size. The analytical
framework developed oers generalized methods for cleaning, processing, and merging federally
available electricity generation and climate datasets to increase their value in future studies.
Chapter 3 spatially resolved the water consumed and water withdrawn for coal mining, coal
preparation, and coal power plant cooling from (a) where the coal is mined to where the coal
is burned to produce power and (b) where the electricity is generated to where the electricity is
76
consumed. Our results indicated that the attribution of water consumption for energy can vary
depending on where the impacts are incurred and where the energy is actually consumed. Coal
is mined at only a number of regions in the US, but some coal traveled great distances before
reaching a power plant. Additionally, due to the interconnected nature of the electric grid, regions
that did not generate coal-red electricity withdrew and consumed cooling water virtually through
the import of coal-red electricity from other regions.
Chapter 4 developed a multi-model framework to assess the impacts of basin-level climate
change and water constraints on electricity generation. Climate, hydrological, and reservoir man-
agement models were applied to the San Juan River basin in the southwestern US to project
changes in regional climatic conditions and water resources. An electricity production cost model
developed in PLEXOS was used to identify vulnerabilities in local electricity generation and in the
western US electricity grid. Results showed in severe drought scenarios, hydroelectric power plants
saw decreases in generation from decreased out
ow, and coal power plants had to decrease their
generation to meet water withdrawal constraints. Reduced generation in the basin caused other
generators in the western US to increase their generation to meet electricity demand. The ndings
of this work emphasized the importance of the consideration of local water use and regulations
when studying the vulnerabilities of power production to climate change.
The frameworks and methodologies developed in this dissertation oer insight in how to study
the complex relationships in the water-energy-climate nexus. The data-driven methodologies pro-
vide insight into how to best utilize and synergize federally reported energy data sets. Federal
datasets are often utilized by the scientic community, but there exists many issues when linking
multiple electricity and climate datasets, including across various lifecycle stages. Chapters 2 and
3 provide recommendations on how to best utilize and combine various energy and environmen-
tal datasets. The integrated modeling framework developed in Chapter 4 resolve many of the
spatial and temporal issues that come with studying global climate change, regional electricity
grid operations, and local water resources. Furthermore, the framework can be applied to other
77
case studies and assess how climatic, water, and regulatory constraints impact local and regional
electricity operations elsewhere.
78
Reference List
[1] U.S. Environmental Protection Agency, \Sources of Greenhouse Gas Emissions," 2019.
[2] U.S. Environmental Protection Agency, \Air Pollutant Emissions Trends Data."
[3] T. H. Diehl and M. A. Harris, \Withdrawal and consumption of water by thermoelectric
power plants in the United States, 2010: U.S. Geological Survey Scientic Investigations
Report," tech. rep., 2014.
[4] U.S. Energy Information Administration, \Electric Power Annual 2015," U.S. Energy In-
fromation Administration, no. November, p. 231, 2016.
[5] A. S. Stillwell and M. E. Webber, \Evaluation of power generation operations in response to
changes in surface water reservoir storage," Environmental Research Letters, vol. 8, no. 2,
p. 025014, 2013.
[6] H. H. Erdem and S. H. Sevilgen, \Case study: Eect of ambient temperature on the elec-
tricity production and fuel consumption of a simple cycle gas turbine in Turkey," Applied
Thermal Engineering, vol. 26, no. 2-3, pp. 320{326, 2006.
[7] U.S. Energy Information Administration, Monthly Energy Review - September 2018,
vol. 0035. 2018.
[8] B. G. Jabboury and M. A. Darwish, \Performance of gas turbine co-generation power de-
salting plants under varying operating conditions in Kuwait," Heat Recovery Systems and
CHP, vol. 10, no. 3, pp. 243{253, 1990.
[9] A. De Sa and S. Al Zubaidy, \Gas turbine performance at varying ambient temperature,"
Applied Thermal Engineering, vol. 31, no. 14-15, pp. 2735{2739, 2011.
[10] J. S. Maulbetsch and M. N. DiFilippo, \Cost and value of water use at combined-cycle
power plants," Tech. Rep. CEC-500-2006-034, California Energy Commission, 2006.
[11] M. T. H. van Vliet, J. R. Yearsley, F. Ludwig, S. V ogele, D. P. Lettenmaier, and P. Kabat,
\Vulnerability of US and European electricity supply to climate change," Nature Climate
Change, vol. 2, no. 9, pp. 676{681, 2012.
[12] M. T. H. van Vliet, D. Wiberg, S. Leduc, and K. Riahi, \Power-generation system vulnera-
bility and adaptation to changes in climate and water resources," Nature Climate Change,
vol. 6, no. January, pp. 375{380, 2016.
[13] J. A. Sathaye, L. L. Dale, P. H. Larsen, G. A. Fitts, K. Koy, and S. M. Lewis, \Estimating
impacts of warming temperatures on California' s electricity system," Global Environmental
Change, vol. 23, pp. 499{511, 2013.
79
[14] M. A. Cook, C. W. King, F. T. Davidson, and M. E. Webber, \Assessing the impacts of
droughts and heat waves at thermoelectric power plants in the United States using integrated
regression, thermodynamic, and climate models," Energy Reports, vol. 1, pp. 193{203, 2015.
[15] L. Liu, M. Hejazi, H. Li, B. Forman, and X. Zhang, \Vulnerability of US thermoelectric
power generation to climate change when incorporating state-level environmental regula-
tions," Nature Energy, vol. 2, no. July, p. 17109, 2017.
[16] A. Miara, J. E. Macknick, C. J. V or osmarty, V. C. Tidwell, R. Newmark, and B. Fekete,
\Climate and water resource change impacts and adaptation potential for US power supply,"
Nature Climate Change, vol. 7, no. 11, pp. 793{798, 2017.
[17] J. Colman, The Eect of Ambient Air and Water Temperature on Power Plant Eciency.
PhD thesis, Duke University, 2013.
[18] Y. Rousseau, Impact of Climate Change on Thermal Power Plants University of Iceland.
PhD thesis, University of Iceland, 2013.
[19] C. L. Henry and L. F. Pratson, \Eects of environmental temperature change on the e-
ciency of coal- and natural gas-red power plants," Environmental Science and Technology,
vol. 50, no. 17, pp. 9764{9772, 2016.
[20] U.S. Energy Information Administration, \Form EIA-860 detailed data with previous form
data."
[21] U.S. Environmental Protection Agency, \Air Markets Program Data."
[22] U.S. Environmental Protection Agency, \Documentation for EPA's Power Sector Modeling
Platform v6: Using the Integrated Planning Model," Tech. Rep. November, U.S. Environ-
mental Protection Agency, Washington, D.C. 20460, 2018.
[23] U.S. Environmental Protection Agency, \40 CFR Part 75 Emissions Monitoring Policy
Manual," tech. rep., Washington, D.C., 2013.
[24] U.S. Environmental Protection Agency, \ECMPS Reporting Instructions: Emissions," 2018.
[25] U.S. Energy Information Administration, \Average Tested Heat Rates by Prime Mover and
Energy Source, 2007 - 2016."
[26] M. Hahsler, M. Piekenbrock, S. Arya, and D. Mount, \Package dbscan': Density Based
Clustering of Applications with Noise (DBSCAN) and Related Algorithms," 2018.
[27] National Oceanic and Atmospheric Administration, \Local Climatological Data (LCD),"
2019.
[28] M. C. Baechler, T. L. Gilbride, P. C. Cole, M. G. Hefty, and K. Ruiz, \High-Performance
Home Technologies: Guide to Determining Climate Regions by County," Tech. Rep. 3, 2015.
[29] K. Rayaprolu, Boilers for Power and Process. Boca Raton, FL: CRC Press/Taylor &
Francis, 2009.
[30] R. Kehlhofer, F. Hannemann, B. Rukes, and F. Stirnimann, Combined-Cycle Gas & Steam
Turbine Power Plants. Lilburn, GA: Fairmont Press, 3 ed., 2009.
[31] J. C. Hensley, ed., Cooling Tower Fundamentals. Overland Park, Kansas: SPX Cooling
Technologies, Inc., 2 ed., 2012.
80
[32] \Energy-Water Nexus: Improvements to Federal Water Use Data Would Increase Under-
standing of Trends in Power Plant Water Use," Tech. Rep. October, U.S. Government
Accountability Oce, 2009.
[33] \Comparison of Alternate Cooling Technologies for California Power Plants: Economic, En-
vironmental and Other Tradeos," tech. rep., EPRI, Palo Alto, CA, and California Energy
Commission, Sacramento, CA, 2002.
[34] M. M. Rahman, T. K. Ibrahim, and A. N. Abdalla, \Thermodynamic performance analysis
of gas-turbine power-plant," Physical Sciences, vol. 6, no. 14, pp. 3539{3550, 2011.
[35] F. R. Arrieta and E. E. Lora, \In
uence of ambient temperature on combined-cycle power-
plant performance," Applied Energy, vol. 80, no. 3, pp. 261{272, 2005.
[36] K. T. Sanders, \Critical Review: Uncharted Waters? The Future of the Electricity- Water
Nexus," Environmental Science & Technology, vol. 49, pp. 51{66, 2015.
[37] V. Tidwell, C. Shaneyfelt, K. Cauthen, G. Klise, F. Fields, Z. Clement, and D. Bauer,
\Implications of Power Plant Idling and Cycling on Water Use Intensity," Environmental
Science and Technology, vol. 53, no. 8, pp. 4657{4666, 2019.
[38] Z. Clement, F. Fields, D. Bauer, V. Tidwell, C. R. Shaneyfelt, and G. Klise, \Eects of
Cooling System Operations on Withdrawal for Thermoelectric Power," in Proceedings of
the ASME 2017 Power Conference, (Charlotte, NC), 2017.
[39] M. Chaker and C. B. Meher-Homji, \Evaporative Cooling of Gas Turbine Engines: Climatic
Analysis and Application in High Humidity Regions," pp. 761{773, 2009.
[40] F. G. Li, C. Bataille, S. Pye, and A. O'Sullivan, \Prospects for energy economy modelling
with big data: hype, eliminating blind spots, or revolutionising the state of the art?," Applied
Energy, vol. 239, no. February, pp. 991{1002, 2019.
[41] Z. Khan, P. Linares, M. Rutten, S. Parkinson, N. Johnson, and J. Garc a-gonz alez, \Spatial
and temporal synchronization of water and energy systems: Towards a single integrated
optimization model for long-term resource planning," Applied Energy, 2017.
[42] S. Mohammed, A. Zaidi, V. Chandola, M. R. Allen, J. Sanyal, R. N. Stewart, B. L. Bhaduri,
A. Ryan, S. Mohammed, A. Zaidi, V. Chandola, M. R. Allen, J. Sanyal, R. N. Stewart, B. L.
Bhaduri, and R. A. Mcmanamay, \Machine learning for energy-water nexus: challenges and
opportunities," Big Earth Data, vol. 2, no. 3, pp. 228{267, 2018.
[43] U.S. Energy Information Administration, \Form EIA-923 Detailed Data."
[44] R. A. Peer, J. B. Garrison, C. P. Timms, and K. T. Sanders, \Spatially and Temporally
Resolved Analysis of Environmental Trade-Os in Electricity Generation," Environmental
Science & Technology, vol. 50, no. 8, pp. 4537{4545, 2016.
[45] R. A. Peer and K. T. Sanders, \The water consequences of a transitioning US power sector,"
Applied Energy, vol. 210, pp. 613{622, 2017.
[46] M. A. Harris and T. H. Diehl, \A Comparison of Three Federal Datasets for Thermoelec-
tric Water Withdrawals in the United States for 2010," Journal of the American Water
Resources Association, vol. 53, no. 5, pp. 1062{1080, 2017.
[47] U.S. Environmental Protection Agency, \Implementation of Clean Water Act Section 316(a)
Thermal Variances in NPDES Permits (Review of Existing Requirements)," 2008.
81
[48] C. L. Weber, P. Jaramillo, J. Marriott, and C. Samaras, \Life cycle assessment and grid elec-
tricity: What do we know and what can we know?," Environmental Science and Technology,
vol. 44, no. 6, pp. 1895{1901, 2010.
[49] M. Whitaker, G. A. Heath, P. O'Donoughue, and M. Vorum, \Life Cycle Greenhouse Gas
Emissions of Coal-Fired Electricity Generation," Journal of Industrial Ecology, vol. 16,
pp. S53{S72, 2012.
[50] D. Weisser, \A guide to life-cycle greenhouse gas (GHG) emissions from electric supply
technologies," Energy, vol. 32, no. 9, pp. 1543{1559, 2007.
[51] R. Turconi, A. Boldrin, and T. Astrup, \Life cycle assessment (LCA) of electricity generation
technologies: Overview, comparability and limitations," Renewable and Sustainable Energy
Reviews, vol. 28, pp. 555{565, 2013.
[52] B. K. Sovacool, \Valuing the greenhouse gas emissions from nuclear power: A critical sur-
vey," Energy Policy, vol. 36, no. 8, pp. 2950{2963, 2008.
[53] D. Nugent and B. K. Sovacool, \Assessing the lifecycle greenhouse gas emissions from solar
PV and wind energy: A critical meta-survey," Energy Policy, vol. 65, pp. 229{244, 2014.
[54] V. Ramaswamy, O. Boucher, J. Haigh, D. Hauglustaine, J. Haywood, G. Myhre, T. Naka-
jima, G. Shi, and S. Solomon, \Radiative forcing of climate change," in Climate Change
2001: The Scientic Basis. Contribution of Working Group I to the Third Assessment Re-
port of the Intergovernmental Panel on Climate Change (J. Houghton, Y. Ding, D. Griggs,
M. Noguer, P. van der Linden, X. Dai, K. Maskell, and C. Johnson, eds.), p. 881, Cambridge,
United Kingdom and New York, NY, USA: Cambridge University Press, 2001.
[55] P. Gleick, \Water and Energy," Annual Review of Energy and the Environment, vol. 19,
no. 1, pp. 267{299, 1994.
[56] H. Inhaber, \Water Use in Renewable and Conventional Electricity Production," Energy
Sources, vol. 26, no. 3, pp. 309{322, 2004.
[57] V. Fthenakis and H. C. Kim, \Life-cycle uses of water in U.S. electricity generation," Re-
newable and Sustainable Energy Reviews, vol. 14, no. 7, pp. 2039{2048, 2010.
[58] B. E. Mielke, L. D. Anadon, and V. Narayanamurti, \Water Consumption of Energy Re-
source Extraction , Processing , and Conversion," no. October, pp. 1{48, 2010.
[59] J. Meldrum, S. Nettles-Anderson, G. Heath, and J. Macknick, \Life cycle water use for
photovoltaic electricity generation: A review and harmonization of literature estimates,"
2014 IEEE 40th Photovoltaic Specialist Conference, PVSC 2014, pp. 1458{1460, 2014.
[60] J. Gao, P. Zhao, H. Zhang, G. Mao, and Y. Wang, \Operational water withdrawal and
consumption factors for electricity generation technology in China-A literature review,"
Sustainability (Switzerland), vol. 10, no. 4, 2018.
[61] B. Ali and A. Kumar, \Development of life cycle water-demand coecients for coal-based
power generation technologies," Energy Conversion and Management, vol. 90, pp. 247{260,
2015.
[62] Y. Chang, R. Huang, R. J. Ries, and E. Masanet, \Life-cycle comparison of greenhouse gas
emissions and water consumption for coal and shale gas red power generation in China,"
Energy, vol. 86, pp. 335{343, 2015.
82
[63] E. Grubert and K. T. Sanders, \Water Use in the United States Energy System: A Na-
tional Assessment and Unit Process Inventory of Water Consumption and Withdrawals,"
Environmental Science and Technology, vol. 52, no. 11, pp. 6695{6703, 2018.
[64] J. K. Lovelace, \Methods for Estimating Water Withdrawals for Mining in the United
States, 2005," tech. rep., U.S. Geological Survey, Reston, Virginia, 2009.
[65] Y. Kuwayama, S. Olmstead, and A. Krupnick, \Water Quality and Quantity Impacts of
Hydraulic Fracturing," Current Sustainable and Renewable Energy Reports, vol. 2, no. 1,
pp. 17{24, 2015.
[66] I. J. Laurenzi and G. R. Jersey, \Life cycle greenhouse gas emissions and freshwater con-
sumption of Marcellus shale gas.," Environmental sciencef&g technology, vol. 47, no. 9,
pp. 4896{4903, 2013.
[67] A. Kondash and A. Vengosh, \Water Footprint of Hydraulic Fracturing," Environmental
Science & Technology Letters, p. 150918090943001, 2015.
[68] J. Meldrum, S. Nettles-Anderson, G. Heath, and J. Macknick, \Life cycle water use for
electricity generation: a review and harmonization of literature estimates," Environmental
Research Letters, vol. 8, no. 1, p. 015031, 2013.
[69] Y. Zhu, S. Jiang, Y. Zhao, H. Li, G. He, and L. Li, \Life-cycle-based water footprint
assessment of coal-red power generation in China," Journal of Cleaner Production, vol. 254,
p. 120098, 2020.
[70] X. D. Wu, X. Ji, C. Li, X. H. Xia, and G. Q. Chen, \Water footprint of thermal power in
China: Implications from the high amount of industrial water use by plant infrastructure
of coal-red generation system," Energy Policy, vol. 132, no. May, pp. 452{461, 2019.
[71] J. Macknick, R. Newmark, G. Heath, and K. C. Hallett, \Operational water consumption
and withdrawal factors for electricity generating technologies: a review of existing litera-
ture," Environmental Research Letters, vol. 7, no. 4, p. 45802, 2012.
[72] R. A. Peer and K. T. Sanders, \Characterizing cooling water source and usage patterns
across US thermoelectric power plants: A comprehensive assessment of self-reported cooling
water data," Environmental Research Letters, vol. 11, no. 12, 2016.
[73] R. A. M. Peer, E. Grubert, and K. T. Sanders, \A regional assessment of the water embedded
in the US electricity system," Environmental Research Letters, vol. 14, no. 8, p. 084014, 2019.
[74] J. A. Allan, \Virtual water: a strategic resource," Groundwater, vol. 36, no. 4, pp. 545{546,
1998.
[75] M. Konar, C. Dalin, S. Suweis, N. Hanasaki, A. Rinaldo, and I. Rodriguez-Iturbe, \Water
for food: The global virtual water trade network," Water Resources Research, vol. 47, no. 5,
pp. 1{17, 2011.
[76] C. Dalin, M. Konar, N. Hanasaki, A. Rinaldo, and I. Rodriguez-Iturbe, \Evolution of the
global virtual water trade network," Proceedings of the National Academy of Sciences of the
United States of America, vol. 109, no. 16, pp. 5989{5994, 2012.
[77] P. D'Odorico, J. Carr, C. Dalin, J. Dell'Angelo, M. Konar, F. Laio, L. Ridol, L. Rosa,
S. Suweis, S. Tamea, and M. Tuninetti, \Global virtual water trade and the hydrological
cycle: Patterns, drivers, and socio-environmental impacts," Environmental Research Letters,
vol. 14, no. 5, 2019.
83
[78] X. D. Wu, J. L. Guo, C. H. Li, L. Shao, M. Y. Han, and G. Q. Chen, \Global socio-hydrology:
An overview of virtual water use by the world economy from source of exploitation to sink
of nal consumption," Journal of Hydrology, vol. 573, no. March, pp. 794{810, 2019.
[79] R. R. Rushforth and B. L. Ruddell, \A spatially detailed blue water footprint of the United
States economy," Hydrology and Earth System Sciences, vol. 22, pp. 3007{3032, 2017.
[80] Y. Liu, B. Chen, W. Wei, L. Shao, Z. Li, W. Jiang, and G. Chen, \Global water use
associated with energy supply, demand and international trade of China," Applied Energy,
vol. 257, no. October 2019, p. 113992, 2020.
[81] S. Mubako, S. Lahiri, and C. Lant, \Input-output analysis of virtual water transfers: Case
study of California and Illinois," Ecological Economics, vol. 93, pp. 230{238, 2013.
[82] A. Mayer, S. Mubako, and B. L. Ruddell, \Developing the greatest Blue Economy: Water
productivity, fresh water depletion, and virtual water trade in the Great Lakes basin,"
Earth's Future, vol. 4, no. 6, pp. 282{297, 2016.
[83] X. Zhu, R. Guo, B. Chen, J. Zhang, T. Hayat, and A. Alsaedi, \Embodiment of virtual
water of power generation in the electric power system in China," Applied Energy, vol. 151,
pp. 345{354, 2015.
[84] C. Zhang, L. Zhong, S. Liang, K. T. Sanders, J. Wang, and M. Xu, \Virtual scarce water
embodied in inter-provincial electricity transmission in China," Applied Energy, vol. 187,
pp. 438{448, 2017.
[85] X. Liao, X. Zhao, J. W. Hall, and D. Guan, \Categorising virtual water transfers through
China's electric power sector," Applied Energy, vol. 226, no. October 2017, pp. 252{260,
2018.
[86] C. Wang, R. Wang, E. Hertwich, Y. Liu, and F. Tong, \Water scarcity risks mitigated
or aggravated by the inter-regional electricity transmission across China," Applied Energy,
vol. 238, no. August 2018, pp. 413{422, 2019.
[87] B. Cai, W. Zhang, K. Hubacek, K. Feng, Z. L. Li, Y. Liu, and Y. Liu, \Drivers of virtual
water
ows on regional water scarcity in China," Journal of Cleaner Production, vol. 207,
pp. 1112{1122, 2019.
[88] B. L. Ruddell, E. A. Adams, R. R. Rushforth, and V. C. Tidwell, \Embedded resource
accounting for coupled natural-human systems: An application to water resource impacts
of the western U.S. electrical energy trade," Water Resources Research, pp. 1{17, 2014.
[89] C. M. Chini, L. A. Djehdian, W. N. Lubega, and A. S. Stillwell, \Virtual water transfers of
the US electric grid," Nature Energy, vol. 3, no. 12, pp. 1115{1123, 2018.
[90] C. M. Chini and A. S. Stillwell, \The changing virtual water trade network of the European
electric grid," Applied Energy, vol. 260, no. October 2019, p. 114151, 2020.
[91] K. T. Sanders, M. F. Blackhurst, C. W. King, and M. E. Webber, \The Impact of Water Use
Fees on Water Used for Cooling Thermoelectric Power Generators," Environmental Science
& Technology, p. accepted with revisions, 2014.
[92] C. W. King, A. S. Stillwell, K. M. Twomey, and M. E. Webber, \Coherence between water
and energy policies," Natural Resources Journal, vol. 53, no. 1, pp. 117{215, 2013.
[93] U.S. Energy Information Administration, \Annual Coal Report 2018," tech. rep., U.S. En-
ergy Information Administration, Washington, D.C., 2019.
84
[94] U.S. Energy Information Administration, \Electricity Data Browser: Net generation," 2019.
[95] U.S. Energy Information Administration, \Rened coal has made up nearly one-fth of
coal-red power generation so far in 2017," 2017.
[96] C. A. Dieter, M. A. Maupin, R. R. Caldwell, M. A. Harris, T. I. Ivahnenko, J. K. Lovelace,
N. L. Barber, and K. S. Linsey, \Estimated use of water in the United States in 2015: U.S.
Geological Survey Circular 1441," tech. rep., 2018.
[97] U.S. Energy Information Administration, \Open Data API."
[98] California Water Boards, \Ocean Standards - CWAx316(b) Regulation: Cooling Water
Intake Structures Once-Through Cooling," 2020.
[99] P. Mukheibir, \Potential consequences of projected climate change impacts on hydroelec-
tricity generation," Climatic Change, vol. 121, no. 1, pp. 67{78, 2013.
[100] S. C. Parkinson and N. Djilali, \Robust response to hydro-climatic change in electricity
generation planning," Climatic Change, vol. 130, no. 4, pp. 475{489, 2015.
[101] J. McCall and J. Macknick, \Water-Related Power Plant Curtailments: An Overview of
Incidents and Contributing Factors," Tech. Rep. December, National Renewable Energy
Laboratory, Golden, CO, 2016.
[102] K. Linnerud, T. K. Mideksa, and G. S. Eskeland, \The Impact of Climate Change on
Nuclear Power Supply," The Energy Journal, vol. 32, no. 1, pp. 149{168, 2011.
[103] A. Durmayaz and O. S. Sogut, \In
uence of cooling water temperature on the eciency of a
pressurized-water reactor nuclear-power plant," International Journal of Energy Research,
vol. 30, no. 10, pp. 799{810, 2006.
[104] S. Fleischli and B. Hayat, \Power Plant Cooling and Associated Impacts," NRDC Issue
Brief, no. April, 2014.
[105] M. D. Bartos and M. V. Chester, \Impacts of climate change on electric power supply in
the Western United States," Nature Climate Change, no. May, pp. 1{5, 2015.
[106] K. E. Bennett, V. C. Tidwell, D. Llewellyn, S. Behery, L. Barrett, M. Stansbury, and R. S.
Middleton, \Threats to a Colorado river provisioning basin under coupled future climate
and societal scenarios," Environmental Research Communications, vol. 1, no. 9, p. 095001,
2019.
[107] B. Livneh, T. J. Bohn, D. W. Pierce, F. Munoz-arriola, B. Nijssen, R. Vose, D. R. Cayan,
and L. Brekke, \A spatially comprehensive , hydrometeorological data set for Mexico , the
U.S ., and Southern Canada 1950 2013," Scientic Data, vol. 2, pp. 1{12, 2015.
[108] K. E. Taylor, R. J. Stouer, and G. A. Meehl, \An overview of CMIP5 and the experiment
design," Bulletin of the American Meteorological Society, vol. 93, no. 4, pp. 485{498, 2012.
[109] J. T. Abatzoglou and T. J. Brown, \A comparison of statistical downscaling methods suited
for wildre applications," International Journal of Climatology, vol. 32, no. 5, pp. 772{780,
2012.
[110] G. P. Peters, R. M. Andrew, T. Boden, J. G. Canadell, P. Ciais, C. Le Quere, G. Marland,
M. R. Raupach, and C. Wilson, \The challenge to keep global warming below 2 degrees C,"
Nature Climate Change, vol. 3, no. 1, pp. 4{6, 2013.
85
[111] X. Liang, E. F. Wood, and D. P. Lettenmaier, \Surface soil moisture parameterization of
the VIC-2L model: Evaluation and modication," Global and Planetary Change, vol. 13,
no. 1-4, pp. 195{206, 1996.
[112] D. Lohmann, R. Nolte-Holube, and E. Raschke, \A largescale horizontal routing model to
be coupled to land surface parametrization schemes," Tellus, vol. 48A, pp. 708{721, 1996.
[113] Energy Exemplar, \PLEXOS Integrated Energy Model."
[114] G. Brinkman, J. Jorgenson, A. Ehlen, and J. H. Caldwell, \Low Carbon Grid Study: Anal-
ysis of a 50% Emission Reduction in California," tech. rep., National Renewable Energy
Laboratory, 2016.
[115] J. Macknick, E. Zhou, M. O. Connell, G. Brinkman, A. Miara, E. Ibanez, and M. Hummon,
\Water and Climate Impacts on Power System Operations: The Importance of Cooling Sys-
tems and Demand Response Measures," tech. rep., National Renewable Energy Laboratory,
2016.
86
Appendix A
Supplemental Information for Chapter 2: A data-driven
approach to investigating the impacts of air temperature
on the eciencies of coal and natural gas generators
A.1 Methods
A.1.1 Data ltering
Table A.1: Lower and upper limits placed on heat rates, based on the EPA's assumptions for the
Platform model [22].
Unit Type Nameplate Capacity Lower Heat Rate Limit (Btu/KWh) Upper Heat Rate Limit (Btu/KWh)
CL-ST All 8,300 14,500
NG-ST All 8,300 14,500
NG-GT >= 80 MW 8,700 18,700
NG-GT < 80 MW 8,700 36,800
NG-CC All 5,500 15,000
87
A.1.2 Final set of units
Figure A.1: Generating units analyzed for each year are mapped and characterized by distance
to nearest NOAA weather station.
88
Figure A.2: Generating units analyzed for each year are mapped and characterized by fuel, prime
mover, and cooling system type.
89
A.2 Results
Table A.2: Results of change in eciency per 1

C increase in ambient air temperature are
dierentiated by cooling type: NONE (no cooling, for natural gas combustion generators), RT
(recirculating cooling with tower), DRY (dry cooling), RC (recirculating with cooling pond), OC
(once-through with cooling pond), and ON (once-through without cooling pond). The statisti-
cal summaries provided are: Min (minimum), Q25 (25th percentile), Med (median), Q75 (75th
percentile), and Max (maximum).
=T
wb
(%=

C)
Cooling Type Temperature Break Category # of EGUs Min Q25 Med Q75 Max
RT All 491 -0.09% -0.03% -0.01% 0.01% 0.07%
RT T
wb
<= T
wb;50
491 -0.07% -0.02% 0.004% 0.02% 0.08%
RT T
wb;50
< T
wb
<= T
wb;75
491 -0.15% -0.06% -0.02% 0.01% 0.10%
RT T
wb;75
< T
wb
<= T
wb;90
491 -0.23% -0.09% -0.04% 0.001% 0.14%
RT T
wb
> T
wb;90
491 -0.33% -0.12% -0.06% 0.01% 0.21%
=T
db
(%=

C)
Min Q25 Med Q75 Max
DRY All 37 -0.05% -0.02% -0.01% 0.001% 0.03%
DRY T
db
<= T
db;50
37 -0.05% -0.01% 0.003% 0.02% 0.07%
DRY T
db;50
< T
db
<= T
db;75
37 -0.08% -0.03% -0.01% 0.01% 0.05%
DRY T
db;75
< T
db
<= T
db;90
37 -0.15% -0.07% -0.03% -0.01% 0.05%
DRY T
db
> T
db;90
37 -0.20% -0.11% -0.07% -0.03% 0.03%
RC All 102 -0.06% -0.03% -0.01% 0.002% 0.04%
RC T
db
<= T
db;50
102 -0.06% -0.01% 0.01% 0.02% 0.07%
RC T
db;50
< T
db
<= T
db;75
102 -0.13% -0.06% -0.03% 0.002% 0.06%
RC T
db;75
< T
db
<= T
db;90
102 -0.11% -0.05% -0.02% 0.003% 0.06%
RC T
db
> T
db;90
102 -0.16% -0.06% -0.02% 0.01% 0.12%
OC All 51 -0.07% -0.03% -0.01% 0.005% 0.05%
OC T
db
<= T
db;50
51 -0.04% -0.01% 0.0002% 0.01% 0.05%
OC T
db;50
< T
db
<= T
db;75
51 -0.19% -0.08% -0.02% 0.03% 0.10%
OC T
db;75
< T
db
<= T
db;90
51 -0.16% -0.06% -0.01% 0.02% 0.12%
OC T
db
> T
db;90
51 -0.14% -0.05% -0.01% 0.02% 0.12%
ON All 383 -0.09% -0.04% -0.01% 0.01% 0.07%
ON T
db
<= T
db;50
383 -0.08% -0.02% -0.003% 0.02% 0.08%
ON T
db;50
< T
db
<= T
db;75
383 -0.19% -0.08% -0.03% 0.001% 0.12%
ON T
db;75
< T
db
<= T
db;90
383 -0.16% -0.06% -0.02% 0.02% 0.12%
ON T
db
> T
db;90
383 -0.16% -0.05% -0.01% 0.02% 0.13%
NONE (GT) All 204 -0.06% -0.02% -0.01% 0.01% 0.04%
NONE (GT) T
db
<= T
db;50
204 -0.06% -0.02% -0.003% 0.01% 0.04%
NONE (GT) T
db;50
< T
db
<= T
db;75
204 -0.08% -0.03% -0.01% 0.01% 0.07%
NONE (GT) T
db;75
< T
db
<= T
db;90
204 -0.10% -0.03% -0.01% 0.01% 0.08%
NONE (GT) T
db
> T
db;90
204 -0.09% -0.02% -0.001% 0.02% 0.08%
90
Figure A.3: Regression results for generating unit eciency change per 1

C increase in tempera-
ture (=T ) are plotted, characterized by fuel, prime mover, and cooling system type. (Outliers
are not included in box plots.) For generating units with recirculating cooling towers, wet-bulb
temperature was used in the regression (T =T
wb
). For all other cooling types and for natural gas
combustion generators, dry bulb temperature was used in the regression (T =T
db
).
91
Figure A.4: Residual standard error (RSE) values from regression models are plotted and charac-
terized by cooling type. Note that in this regression model, the modeling is not split by tempera-
ture categories. (Outliers are not included in box plots.) For generating units with recirculating
cooling towers, wet-bulb temperature was used in the regression (T =T
wb
). For all other cooling
types and for natural gas combustion generators, dry bulb temperature was used in the regression
(T =T
db
).
92
Figure A.5: Regression results for generating unit eciency change per 1

C increase in tem-
perature (=T ) are plotted, characterized by cooling system type and nameplate capacity.
(Outliers are not included in box plots.) For generating units with recirculating cooling towers,
wet-bulb temperature was used in the regression (T = T
wb
). For all other cooling types and for
natural gas combustion generators, dry bulb temperature was used in the regression (T =T
db
).
Figure A.6: Regression results for generating unit eciency change per 1

C increase in temper-
ature (=T ) are plotted, characterized by cooling system type and generating unit operation
year. (Outliers are not included in box plots.) For generating units with recirculating cooling tow-
ers, wet-bulb temperature was used in the regression (T =T
wb
). For all other cooling types and
for natural gas combustion generators, dry bulb temperature was used in the regression (T =T
db
).
93
Appendix B
Supplemental Information for Chapter 3: Spatially
allocating lifecycle water use for US coal-red electricity
across producers, generators, and consumers
B.1 Computational Methods
When possible, coal type-specic and province-specic water intensities developed by Grubert and
Sanders [63] for mining and preparing coal were utilized. However, a few additional assumptions
were made for coal rank and mining province combinations. In the case of waste coal, water
intensities were based on the provinces and the dominant coal rank in that province.
Appalachia/Eastern waste coal = Appalachia/Eastern bituminous coal
Rocky Mountain Region waste coal = Rocky Mountain Region bituminous coal
Northern Great Plains waste coal = Northern Great Plains subbituminous coal
Coal from Unknown provinces were assigned the average of all water intensities across coal
rank and mining provinces. Imported coal was assigned water values of zero, since we drew the
bounds of our analysis within the US, and the water impacts incurred for mining and preparing
imported coal occurred outside of the US.
At the balancing authority level (all balancing authority names and codes are provided in
Table B.3), cooling water consumption and withdrawal intensities were based on fuel, prime
mover, and technology-specic intensities developed by Peer and Sanders [72]. The corresponding
water intensites for each balancing authority are provided in Table B.4.
94
Table B.1: Water withdrawn and water consumed at each life cycle stage for coal-red electricity
consumption in the US in 2017. Each percentage represents the percent contribution from each
process stage to the total amount of water consumed or water withdrawn.
Mining Region Coal Type
Mining Water Consumption and
Withdrawal Intensity (m
3
/GJ)
Appalachia/Eastern BIT, SUB, WC 0.016
Gulf Coast LIG 0.058
Interior BIT 0.14
Northern Great Plains BIT 0.021
Northern Great Plains SUB, LIG, WC 0.0011
Rocky Mountain Region BIT, SUB, WC 0.021
Unknown BIT, SUB, WC 0.03
Imported BIT, SUB 0
Table B.2: Water withdrawn and water consumed at each life cycle stage for coal-red electricity
consumption in the US in 2017. Each percentage represents the percent contribution from each
process stage to the total amount of water consumed or water withdrawn.
Mining Region
Coal Preparation Water
Consumption Intensity (m
3
/GJ)
Coal Preparation Water
Withdrawal Intensity (m
3
/GJ)
Appalachia/Eastern 0.004 0.050
Gulf Coast 0.003 0.037
Interior 0.008 0.095
Northern Great Plains 0.0002 0.002
Rocky Mountain Region 0.004 0.051
Unknown 0.003 0.037
Imported 0 0
95
Table B.3: Balancing authority codes and corresponding names.
Balancing Author-
ity Code
Balancing Authority Name
AEC PowerSouth Energy Cooperative
AECI Associated Electric Cooperative, Inc.
AVA Avista Corporation
AZPS Arizona Public Service Company
BANC Balancing Authority of Northern California
BPAT Bonneville Power Administration
CHPD Public Utility District No. 1 of Chelan County
CISO California Independent System Operator
CPLE Progress Energy Carolinas - EAST
CPLE Duke Energy Progress East
CPLW Duke Energy Progress West
CSTO Constellation Energy Control and Dispatch, LLC
DEAA Arlington Valley, LLC - AVBA
DOPD PUD No. 1 of Douglas County
DUK Duke Energy Carolinas
EEI Electric Energy, Inc.
EPE El Paso Electric Company
ERCO Electric Reliability Council of Texas, Inc.
FMPP Florida Municipal Power Pool
FPC Progress Energy Florida
FPC Duke Energy Florida Inc
FPL Florida Power & Light Company
FPL Florida Power & Light Company
GCPD Public Utility District No. 2 of Grant County, Washington
GRIF Grith Energy, LLC
GRIS Gridforce South
GRMA Gila River Power, LLC
GVL Gainesville Regional Utilities
GWA NaturEner Power Watch, LLC
HECO Hawaiian Electric Co Inc
HGMA New Harquahala Generating Company, LLC - HGBA
HST City of Homestead
IID Imperial Irrigation District
IPCO Idaho Power Company
ISNE ISO New England Inc.
JEA JEA
LDWP Los Angeles Department of Water and Power
LGEE LG&E and KU Services Company as agent for Louisville Gas and Electric
Company and Kentucky Utilities
MISO Midcontinent Independent Transmission System Operator, Inc..
NBSO New Brunswick System Operator
NEVP Nevada Power Company
NSB New Smyrna Beach, Utilities Commission of
NWMT NorthWestern Energy
NYIS New York Independent System Operator
OVEC Ohio Valley Electric Corporation
96
PACE PaciCorp - East
PACW PaciCorp - West
PGE Portland General Electric Company
PJM Pennsylvania-New Jersey-Maryland Interconnection (PJM Interconnection),
LLC
PNM Public Service Company of New Mexico
PS Public Service Company of Colorado
PSCO Public Service Company of Colorado
PSEI Puget Sound Energy
SC South Carolina Public Service Authority
SC South Carolina Electric & Gas Company
SCEG South Carolina Electric & Gas Company
SCL Seattle City Light
SEC Seminole Electric Cooperative
SEPA Southeastern Power Administration
SOCO Southern Company Services, Inc. - Trans
SPA Southwestern Power Administration
SRP Salt River Project
SWPP Southwest Power Pool
TAL City of Tallahassee
TEC Tampa Electric Company
TEPC Tucson Electric Power Company
TIDC Turlock Irrigation District
TPWR City of Tacoma, Department of Public Utilities, Light Division
TVA Tennessee Valley Authority
WACM Western Area Power Administration - Rocky Mountain Region
WALC Western Area Power Administration - Desert Southwest Region
WAUW Western Area Power Administration UGP West
WWA NaturEner Wind Watch, LLC
YAD Alcoa Power Generating, Inc. - Yadkin Division
Table B.4: Water consumption and withdrawal intensities for cooling coal power plants, by bal-
ancing authority adapted from Peer and Sanders [72].
Balancing
Author-
ity
Percentage of
Electricity Generated
from Coal
Cooling Water
Consumption Intensity
(m
3
/MWh)
Cooling Water
Withdrawl Intensity
(m
3
/MWh)
AEC 24% 1.3 79
AECI 75% 0.77 160
ASCC 9% 1.4 160
AVA 0% 0 0
AZPS 64% 1.5 110
BANC 0% 0 0
BPAT 5% 1.4 160
CHPD 0% 0 0
CISO 0% 1.4 160
CPLE 15% 1.4 87
CPLW 0% 0 0
97
CSTO 0% 0 0
DEAA 0% 0 0
DOPD 0% 0 0
DUK 22% 1 120
EEI 99% 0.77 160
EPE 0% 0 0
ERCO 29% 1.3 100
FMPP 40% 1.7 20
FPC 21% 1 120
FPL 0% 1.8 2
GCPD 0% 0 0
GRIF 0% 0 0
GRIS 0% 0 0
GRMA 0% 0 0
GVL 24% 1.8 2
GWA 0% 0 0
HGMA 0% 0 0
HICC 14% 1.8 2
HST 0% 0 0
IID 0% 0 0
IPCO 0% 1.4 160
ISNE 2% 1.6 44
JEA 42% 1.8 26
LDWP 38% 1.8 2
LGEE 86% 1.6 32
MISO 47% 1.4 110
NBSO 0% 0 0
NEVP 6% 1.8 2
NONE 0% 0 0
NSB 0% 0 0
NWMT 73% 1.8 4.2
NYIS 1% 0.95 160
OVEC 100% 1.4 160
PACE 72% 1.7 13
PACW 0% 0 0
PGE 14% 1.4 130
PJM 30% 1.6 61
PNM 68% 1.8 2
PS 0% 0 0
PSCO 33% 1.6 94
PSEI 0% 0 0
SC 57% 1.8 10
SCEG 32% 1.5 52
SCL 0% 0 0
SEC 69% 1.8 2
SEPA 0% 0 0
SOCO 27% 1.6 42
SPA 25% 1.8 2
SRP 27% 1.8 2
SWPP 45% 1.5 91
98
TAL 0% 0 0
TEC 25% 0.77 150
TEPC 84% 1.4 160
TIDC 0% 0 0
TPWR 0% 0 0
TVA 23% 0.91 140
WACM 84% 1.5 48
WALC 9% 1.8 2
WAUW 0% 0 0
WWA 0% 0 0
YAD 0% 0 0
B.2 Additional Results
Electricity transfers between balancing authorities mean the attribution of water consumption and
withdrawal impacts of cooling coal-red power plants between where the electricity is generated
and where the electricity is consumed are not equal. Figure B.1 provides bar plots of the coal-red
electricity generated, cooling water consumed, and cooling water withdrawn for each balancing
authority. Figure B.2 includes maps of the balancing authorities, color-graded by how much water
is consumed (or withdrawn) for the coal-red electricity generated within the balancing authority
versus the coal-red electricity consumed. The electricity delivered is the electricity consumed
in the balancing authority, which is equal to the electricity generated minus the electricity ex-
ported plus the electricity imported into the balancing authority. The cooling water consumed
and withdrawn for the electricity delivered takes into consideration the virtual water transfers
embedded in the transfer of electricity between balancing authorities. Balancing authorities in
the western US, such as CISO, see the largest visible dierence between cooling water consumed
(and withdrawn) for coal-red electricity generated within the balancing authority and coal-red
electricity consumed within the balancing authority.
99
0
100
200
300
AEC
AECI
AVA
AZPS
BANC
BPAT
CHPD
CISO
CPLE
CPLW
DEAA
DOPD
DUK
EEI
EPE
ERCO
FMPP
FPC
FPL
GCPD
GRIF
GRMA
GVL
GWA
HGMA
HST
IID
IPCO
ISNE
JEA
LDWP
LGEE
MISO
NEVP
NSB
NWMT
NYIS
OVEC
PACE
PACW
PGE
PJM
PNM
PSCO
PSEI
SC
SCEG
SCL
SEC
SEPA
SOCO
SPA
SRP
SWPP
TAL
TEC
TEPC
TIDC
TPWR
TVA
WACM
WALC
WAUW
WWA
YAD
Coal−Fired Electricity
Consumed
(TWh)
Imported
Within
Exported
0
100
200
300
400
AEC
AECI
AVA
AZPS
BANC
BPAT
CHPD
CISO
CPLE
CPLW
DEAA
DOPD
DUK
EEI
EPE
ERCO
FMPP
FPC
FPL
GCPD
GRIF
GRMA
GVL
GWA
HGMA
HST
IID
IPCO
ISNE
JEA
LDWP
LGEE
MISO
NEVP
NSB
NWMT
NYIS
OVEC
PACE
PACW
PGE
PJM
PNM
PSCO
PSEI
SC
SCEG
SCL
SEC
SEPA
SOCO
SPA
SRP
SWPP
TAL
TEC
TEPC
TIDC
TPWR
TVA
WACM
WALC
WAUW
WWA
YAD
Cooling Water Consumed
for Coal Power Plants
(Million Cubic Meters)
Imported
Within
Exported
0
10000
20000
30000
AEC
AECI
AVA
AZPS
BANC
BPAT
CHPD
CISO
CPLE
CPLW
DEAA
DOPD
DUK
EEI
EPE
ERCO
FMPP
FPC
FPL
GCPD
GRIF
GRMA
GVL
GWA
HGMA
HST
IID
IPCO
ISNE
JEA
LDWP
LGEE
MISO
NEVP
NSB
NWMT
NYIS
OVEC
PACE
PACW
PGE
PJM
PNM
PSCO
PSEI
SC
SCEG
SCL
SEC
SEPA
SOCO
SPA
SRP
SWPP
TAL
TEC
TEPC
TIDC
TPWR
TVA
WACM
WALC
WAUW
WWA
YAD
Cooling Water Withdrawn
for Coal Power Plants
(Million Cubic Meters)
Imported
Within
Exported
Figure B.1: (top) Coal-red electricity consumed in each balancing authority in 2017. (middle)
Cooling water consumed for coal-red electricity consumed in each balancing authority in 2017.
(bottom) Cooling water withdrawn for coal-red electricity consumed in each balancing authority
in 2017. In all three plots, each bar represents a balancing authority (ordered alphabetically).
The orange color represents the electricity and cooling water associated with coal-red electricity
that was generated and consumed within the same balancing authority. The blue color represents
the electricity and cooling water associated with coal-red electricity that was imported into
the respective balancing authority. The grey color represents the electricity and cooling water
associated with coal-red electricity that was exported out of the balancing authority.
100
Figure B.2: Net generation, cooling water consumed, and cooling water withdrawn for electricity
generated and electricity delivered (taking into consideration electricity imports and exports).
Each region is a balancing authority.
101
Appendix C
Supplemental Information for Chapter 4: Integrating
water, energy, and climate modeling to assess
vulnerabilities to the US Southwest power grid
C.1 San Juan River Basin
Figure C.1: Annual water consumption (top) and water withdrawal (bottom) in San Juan River
Basin.
102
Figure C.2: Annual CO
2
(top), NO
x
(middle) and SO
2
(bottom) emissions in San Juan River
Basin.
103
C.2 WECC
Figure C.3: Annual generation in balancing authorities.
104
Figure C.4: Annual cost in balancing authorities.
105
Figure C.5: Annual CO
2
emissions in balancing authorities.
106
Figure C.6: Annual NO
x
emissions in balancing authorities.
107
Figure C.7: Annual SO
2
emissions in balancing authorities.
108
Figure C.8: Dierence in annual CO
2
, NO
x
, and SO
2
emissions from base scenario in WECC.
109

Abstract (if available)

Abstract The United States (US) energy system is responsible for numerous environmental and climate impacts throughout the various lifecycle stages. The processes of extracting fossil fuels and converting fossil fuels into electric power emit greenhouse gases that lead to the warming of the climate. Additionally, both primary fuel production and electric power production require the withdrawal and consumption of water resources. Changes in water resources and the climate, such as reduced water availability, increased water temperatures, and increased ambient temperatures, can constrain energy production. The complex interdependencies between climate, water, and energy systems are important to understand, especially since climate change is expected to exacerbate the tensions between these systems. This body of work develops empirical and integrated modeling frameworks to understand the operations and environmental impacts of energy systems under climate change and climate variability. The findings emphasize the importance of understanding the spatio-temporal contexts of environmental impacts, particularly water consumption, incurred throughout the different lifecycle stages of energy consumption. There are decouplings between where energy is produced, where energy is consumed, and where environmental impacts occur. Another emphasis in this body of work is that changes in electricity generation in one local area has cascading impacts on other areas in the interconnected grid. As the electricity grid continues to experience various transitions, the vulnerabilities of the electricity grid are going to change as well.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

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

PDF

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

PDF

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

PDF

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

PDF

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

PDF

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

PDF

Energy-efficient shutdown of circuit components and computing systems

PDF

Distribution system reliability analysis for smart grid applications

PDF

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

PDF

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

PDF

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

PDF

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

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

Integration of energy-efficient infrastructures and policies in smart grid

PDF

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

PDF

Defending industrial control systems: an end-to-end approach for managing cyber-physical risk

PDF

Advancing energy recovery from food waste using anaerobic biotechnologies: performance and microbial ecology

PDF

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

PDF

Integrated technologies, blending schemes, and reuse practices to address contaminant and energy challenges in water reclamation

Asset Metadata

Creator Meng, Measrainsey (author)

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

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Engineering (Environmental Engineering)

Publication Date 07/21/2020

Defense Date 06/18/2020

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

Tag electricity grid,energy modeling,life cycle assessment,OAI-PMH Harvest,power systems,water-energy-climate nexus

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sanders, Kelly T. (committee chair), Ban-Weiss, George (committee member), Beshir, Mohammed (committee member)

Creator Email measraim@usc.edu,measrainsey@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-334925

Unique identifier UC11663763

Identifier etd-MengMeasra-8716.pdf (filename),usctheses-c89-334925 (legacy record id)

Legacy Identifier etd-MengMeasra-8716.pdf

Dmrecord 334925

Document Type Dissertation

Rights Meng, Measrainsey

Type texts

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

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

Repository Name University of Southern California Digital Library

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