Predicting the Impact of Climate Change on U.S. Power Grids and Its Wider Implications on National Security.

Full text

Page 1

Predicting the Impact of Climate Change on U.S. Power Grids and Its
Wider Implications on National Security
Pak Chung Wong, L. Ruby Leung, Ning Lu, Mia Paget, James Correia Jr., Wei Jiang, Patrick Mackey,
Z. Todd Taylor, YuLong Xie, Jianhua Xu, Steve Unwin, Antonio Sanfilippo

Pacific Northwest National Laboratory
902 Battelle Boulevard, P.O. Box 999, Richland, WA 99352, USA
{pak.wong | ruby.leung | ning.lu | mia.paget | jim.correia | wei.jiang | patrick.mackey | todd.taylor | yulong.xie | jianhua.xu |
steve.unwin | antonio.sanfilippo}@pnl.gov



Abstract
We discuss our technosocial analytics research and devel-
opment on predicting and assessing the impact of climate
change on U.S. power-grids and the wider implications for
national security. The ongoing efforts extend cutting-edge
modeling theories derived from climate, energy, social
sciences, and national security domains to form a unified
system coupled with an interactive visual interface for tech-
nosocial analysis. The goal of the system is to create viable
future scenarios that address both technical and social fac-
tors involved in the model domains. These scenarios enable
policymakers to formulate a coherent, unified strategy to-
wards building a safe and secure society. The paper gives an
executive summary of our preliminary efforts in the past
year and provides a glimpse of our work planned for the
second year of a multi-year project1
being conducted at the
Pacific Northwest National Laboratory.
The paper presents our ongoing technosocial analytics re-
search and development (R&D) on predicting the impact of
climate change on U.S. power grids and its wider implica-
tions on national security. An example of climate change is
an increased atmospheric temperature, which in turn in-
creases electricity consumption. The increased temperature
also affects precipitation, which changes the natural hydro-
logical process and thus hydroelectric generation; it also
influences wind electricity generation. Together, these de-
mands could adversely affect the U.S. power grids and
cause a widespread outage. If such an outage persisted, it
would impair the ability of our entire critical infrastructure
to perform and could potentially cripple our society. Our
work is to investigate the potential impact and implications
Introduction

1 Technosocial Predictive Analytics Initiative is a Laboratory-Directed
Research and Development project at the Pacific Northwest National La-
boratory.

Copyright © 200 , Association for the Advancement of Artificial Intelli-
gence (www.aaai.org). All rights reserved.

9
of these changes to the society in the next 50 years from
both a technical and social perspective.
The interdisciplinary R&D effort extends the latest
modeling theories and practices derived from atmospheric
physics, electrical engineering, building engineering, social
sciences, economics, and public policy to form a tightly
coupled technosocial predictive analytics system. A major
challenge in our work is the granularity differences, in
terms of both data and methodology, among the domain
models. One solution is to provide a highly interactive vis-
ual analytics layer on top of the domain components to fa-
cilitate the integration of evidence and arguments required
by and generated from the different models. The integrated
system creates viable future scenarios that address both
technical and social factors involved in all model domains.
These scenarios enable policymakers and stakeholders to
formulate a coherent, unified strategy towards building a
safe and secure society.
The paper describes the background and motivation of
our work and summarizes our R&D efforts carried out in
the past year. Preliminary results on predicting the impact
of climate change on both social and technical aspects of
our society in the next 50 years are discussed. More ad-
vanced features and parameters are suggested at the end of
the paper.
Background and Motivation
To communicate the significance of our work, we first de-
scribe the background of the problems and our motivation
to pursuing them.
We live in a society that is vitally dependent on a net-
work infrastructure of natural, man-made, and human re-
sources to function—from food to water supplies, from
electric power to other fuel sources, and from communica-
tion and transportation to medical and emergency services.
While these resources are seamlessly integrated into the
fabric of our society, electric power has the highest “net-
work reachability”—and all the other network resources
depend on it to operate. Losing electric power inevitably
impairs the ability of the other resources to perform, which
148

Page 2

could cripple society if a widespread outage persisted for
prolonged periods.
The August 2003 blackout in the Northeast and Great
Lakes provides a glimpse of the struggle for survival
brought by a widespread power loss. In a matter of hours,
the lives of 50 million people in these areas were signifi-
cantly affected by overwhelmed or disrupted infrastructure
services.
Even though power to most of the affected areas was
restored within a day, the official estimation of the outage-
related financial losses was $6 billion USD. Similar black-
outs happened in the western grid in July and August of
1996. The high summer temperature played a critical role
in pushing the electric power systems to the point of failure
when people turned to fans and air conditioning to beat the
heat.
Soon after the 2003 blackout, the vulnerability of our na-
tion’s electric power grids was the focus of attention of
policy makers and their scientific advisors. A scientific
panel (United States Congress, Office of Technology As-
sessment 1990) testified before the U.S. Congress and
painted a grim scenario of the consequences of an orga-
nized terrorist attack on our electric power systems:
“…with the power out even a day or two, both food
and water supplies would soon fail...Work, jobs, em-
ployment, business and production would be stopped.
Our economy would take a major hit. All in all our ci-
ties would not be very nice places to be... Haves and
have-nots would get involved. It would not be a very
safe place to be either. Martial law would likely fol-
low….”
While the congressional panel focused mainly on the ter-
rorist attack scenario, the climate change impact on the
power grids would only act as a “threat multiplier” that
would greatly increase the likelihood of a widespread ex-
tended blackout and further intensify the scope of damage
to our society. In a highly influential and widely publicized
2004 article in the journal Science, Meehl and Tebaldi
(Meehl, G.A. and Tebaldi C. 2004) predicted that “more
intense, more frequent, and longer lasting heat waves” will
arrive soon in the 21st century. To remind people about the
threat of an extended heat wave, the article cited the 2003
Paris heat wave when the death toll reached nearly 15,000.
Meehl and Tebaldi’s prediction, if correct, would almost
guarantee extended periods of blackouts in our future. Our
underlying motivation is to attempt to take a view into the
probable future of our society and predict likely scenarios
that may very well come into being in our life time.
Related Work
From a literature review standpoint, our work represents a
unique opportunity to “connect the dots” of several do-
main-specific modeling efforts that constitute our underly-
ing predictive analytics problem. We have yet to find simi-
lar work that shares a comparable degree of ambition and
complexity as our problem. On the other hand, a number of
recent investigations have examined problems similar to
portions of our work, but with different approaches and
emphases.
The State of California has recently sponsored an inves-
tigation on the impact of global climate change on building
energy usage (Xu, P. et al. 2007). The study, which focuses
geographically on the state of California, pays great atten-
tion to the building energy usage and does not consider the
impact of electric power stability and social dynamics.
The National Infrastructure Simulation and Analysis
Center (NISAC) (National Infrastructure Simulation and
Analysis Center 2008) at Sandia National Laboratory has a
suite of modeling tools to simulate different national secu-
rity problems from bioterrorism to natural disasters. Multi-
agent technology for adaptation in dynamic environments
is applied extensively to play out possible scenarios, which
is very different from our approach that relies heavily on
scientific evidence and consensus.
Finally, researchers at the Pacific Northwest National
Laboratory are investigating the vulnerability of food secu-
rity and energy infrastructures to climate change and ter-
rorism (Vulnerability of Food Security and Energy Infra-
structures to Climate Change and Terrorism 2008). The
study, which focuses on the areas of India, Pakistan, and
Bangladesh, models the “broad domains that are crucial to
the understanding of global issues involved in climate
change and human security.” This project shares a com-
mon theme with our research that demonstrates the link be-
tween climate change, security, and social development.
Technosocial Predictive Analytics
This section presents an overview of our multidisciplinary
technosocial predictive and analytics project, highlights re-
search and development results to date, and discusses new
tasks planned for the next phase of this research.
Overview
Our interdisciplinary project involves four major compo-
nents that address problems arise in the 1) climate, 2) so-
cial, 3) building and power grids, and 4) security and infra-
structure analytics domains. Figure 1 shows an overview of
the system with various modeling features highlighted in
individual model components. While the climate compo-
nent accepts input mainly from external sources, the other
three accept input from each other as well as from external
sources. On top of these components is a thin visual analyt-
ics layer that facilitates the integration of evidence and ar-
guments required by and generated from the model com-
ponents.
Preliminary Results
The R&D activities of this project are ordered and coordi-
nated gradually and sequentially. The lead components will
provide just enough groundwork for the next component to
take off before returning to the refinement stage to enrich
the component models.
149

Page 3

We have completed the climate analytics work on tem-
perature modeling and finished most of the building energy
simulations, which are required for the power grid simula-
tion. On the social study front, we have focused our inves-
tigation on the impacts of demographic and technological
changes to our study. A preliminary visual analytics sys-
tem prototype has been developed to guide analysis among
the domain components and present results.
We have so far made progress on all four domain com-
ponents. While the linkages among the components have
not been fully established after the first year of the work,
individual results have already shed some light on possible
issues in the future.
Climate Analytics In the past year, we downloaded cli-
mate model outputs archived at the Program for Climate
Model Diagnostics and Intercomparison (PCMDI) (Pro-
gram for Climate Model Diagnosis and Intercomparison
2008) for the Intergovernmental Panel on Climate Change
(IPCC) (Intergovernmental Panel on Climate Change
2008). We obtained hourly simulation outputs from 23
global climate models that used the IPCC Special Report
on Emission Scenarios (SRES) A1B emission scenario. We
focused on two 10-year periods around 2000 and 2050 for
comparison of future and current climate. The simulated
surface temperatures were bias corrected based on the ob-
served climatology for Calgary (AB, Canada), Vancouver
(BC, Canada), Portland (OR), Billings, Salt Lake City,
Sacramento, Los Angeles, San Francisco, Boulder, and
Phoenix in the United States and Canada.
The bias-corrected hourly temperature data (a total of
260 time series) for the above ten cities have been used in
developing the building, energy, and power grid analytics
described below. The global climate simulations show an
increase in the mean daily maximum temperature of 2-3oC,
but larger increase in extreme daily maximum temperature
by 4-6oC, especially during late summer and fall. Analysis
is being performed on potential changes in extreme tem-
peratures and heat waves. The bias corrected surface tem-
peratures for the ten cities are stored in a relational data-
base system, which can be interactively queried using our
visual analytics tool when it is ready for deployment.
Building, Energy, and Power Grid Analytics We use the
DOE-2 (DOE-2 2008) program to simulate building elec-
tric energy end-use. DOE-2 uses a description of the build-
ing layout, construction, usage, conditioning systems
(lighting, HVAC, etc.) and utility rates provided by the us-
er, along with weather data, to perform an hourly simula-
tion of the building and to estimate utility bills. The com-
mercial building prototypes are from the Database for
Energy Efficient Resources (DEER) (DOE-2 Weather Pro-
cessor). The DOE-2 model uses building prototype and
measure characterization information by building type,
vintage, and climate zone in its estimation of measure sav-
ings.
The temperature (T) sensitivities of the building total
energy consumption (Edaily) as well as the peak hourly load
(Ppeak) are derived from the DOE-2 simulation results using
the temperature profile for a typical meteorology year
(TMY). Then piece-wise curve fitting technique is used to
derive the T-Edaily and T-Ppeak curves for each building type
based on day types. There are three day types: Weekday,
Weekends, and Holiday. Note that all the Sundays and
public holidays are included in the Holiday category. An
example is shown in Figure 22
The bias-corrected IPCC modeling results of years
1991-2000 represent the temperature profile for period
Now. For each day in a year within this period, there are
260 Tmax and Tmin. The 260 data points are bias corrected
using TMY data. The bias-corrected IPCC modeling re-
sults of year 2045-2054 represent the temperature profiles
for the period Future. Building level energy consumptions
are then calculated using these temperature profiles as in-
puts.
The following assumptions are made to produce the
baseline result for future building energy consumption:
Assume that the 260 Tmax and Tmin predicted for a specific
day in a year by the 26 IPCC climate models for a ten-year
span are treated equally, i.e,, they have equally likely
chance of occurrence. For the base Future case, no new
technology and no new policy are implemented. Every-
thing stays the same for the building simulated over the
next 50-year period.

2 Contact authors for color version of the graphics.
.
Figure 1: An overview of the model system.
Figure 2: The temperature sensitivity of the total energy con-
sumption of a typical secondary school building.
150

Page 4

n
ave
n
ave
f
ave
E
ave
E
EE
k
?
?
Heating loads are supplied by natural gas in the DOE-2
model and in the unit of btu. At this stage, we only study
the influence of the cooling loads because it is the cooling
load that puts stress on the power system when average
temperatures in each region increase.
The percentage changes are calculated by:
where En
Year 1991 – 2000 and Ef
sumption of the Year 2045 – 2054 for all the 260 model
produced monthly energy consumptions.
So far we have made the following observations using
the results from the Portland and Phoenix areas:
? The total building energy consumption for Portland
shows a very small increase because of the decrease of
the winter load and the mild increase of the summer load
for an average year. For example, for each building type,
under a baseline Tmax and Tmin, which are 40ºF and 30ºF
for January in Portland, the monthly consumption may
decrease by 5% in 2045-2055, for which year, the Tmax
and Tmin are 45ºF and 35ºF.
? For extreme cases, however, we will see a huge increase
in the building load consumption. This suggests that for
a typical hot year, the energy consumption can skyrock-
et, causing a power shortage in a wider area than now.
? The residential buildings see more increases because a/c
load consists of a large percent of total building loads.
? Phoenix sees more cooling load increase than Portland in
general, but the trends are different for the two areas as
shown in Figure 3. Portland mainly sees load increase in
summer months, while Phoenix sees an energy increase
in winter, spring and autumn months, which suggests the
hot days are spreading out to these “cooler” seasons.
? During winter, energy consumption may drop due to the
drop of heating load. Note that the major heating load is
not included in the total loads because in many areas, the
major heating load is supplied by gas. In the future, we
will present the climate impacts on the heating load.
Social Analytics The above study shows that the direct so-
cial impact is the penetration of a/c load. In residential
buildings, to maintain a comfortable living environment,
people who can afford to will install a/c in their house-
holds, and those who cannot afford will either move to
areas with mild temperature changes, such as the Northern
coastal cities, or seek help from government. This will re-
sult in a demographical change in population and create a
social crisis from low-income homes.
For commercial building owners, a significant increase
in load may saturate the a/c system and overload the elec-
trical circuits. Much higher energy bills may incur if real
time pricing is to be implemented in the future.
To mitigate the adverse impact of the load increases re-
sulting from climate change, technology should advance in
a direction that helps to either directly reduce the load in-
ave is the monthly average consumption of the
ave is the monthly average con-
crease by making appliances or electrical equipment more
efficient or manage the load to shift the peak load to off-
peak periods. Three technologies are studied: increase the
a/c system efficiency, increase the lighting efficiency, and
change the thermostat setpoints to manage peak loads. The
results are shown in Figures 4 and 5.
Security Analytics The electricity infrastructure is a phys-
ical network where nodes represent transmission substa-
tions, power substations, and distribution buses, and edges
represent transmission lines. We use network analysis
(Newman, M.E.J. 2008), which has shed light on the prop-
erties of a variety of networks from the internet backbone
(Faloutsos et al. 1999), transportation networks (Banavar et
Figure 3: The trend of energy consumption increase in Port-
land and Phoenix area.
Phoenix-Assembly/Grocery Average Monthly Energy Consumption
Percentage Increase (2045-2054 over 1990-1999)
0
2
4
6
8
10
12
14
16
18
20
123456
Month
789101112
Percentage Increase
Assembly
Grocery
Portland-Assembly/Grocery Average Monthly Energy Consumption
Percentage Increase (2045-2054 over 1990-1999)
0
2
4
6
8
10
12
14
123456
Month
789 1011 12
Percentage Increase
Assembly
Grocery
Figure 4: Predicted changes of energy consumption by buildings
in Phoenix between 1991-2000 (black) and 2045-2054(pink).
151

Page 5

al. 2000), to social networks (Wasserman and Faust 1994),
to identify structurally important nodes in an electricity
network. This information may be useful in determining
where the limited resource is allocated and thus prevent
undesired events (like blackouts).
Anticipating the results of the power grid simulation re-
sults using the Positive Sequence Load Flow Software
(PSLF) (Positive Sequence Load Flow Software 2008) si-
mulation from the power grid analytical component, we
use four network measures, which are betweenness central-
ity, closeness, degree centrality and eigenvector, to identify
potential vulnerabilities of the Western Electricity Coordi-
nating Council (WECC) (Western Electricity Coordinating
Council 2008) infrastructure. These vulnerabilities are
ranked and the results will be correlated to the outcomes of
the PSLF simulation to better understand and predict the
future evolution of the North American power grid. Figure
6 shows the top-ranked locations that are ranked highly in
different centrality measures. For examples, from a net-
work security standpoint, breaking up of buses (network
nodes) that carry high betweenness ratings (red in Figure
6) can lead to serious islanding problems, which, if not
handled properly, may quickly result in a power blackout.)
Visual Analytics To address the problem of the “black
box” effect in many modeling processes, we include a thin
visual analytics layer to encapsulate the major domain
processes into one predictive system. The customized visu-
al analytics layer, which provides a high degree of anima-
tion and interactive modeling scenarios, allows the mod-
elers to analytically explore inputs, assumptions, and algo-
rithms of individual domain theories and their cascading
impacts contributing to the scenarios. The system, which is
currently under development, will support visualization of
climate scenario, energy consumption and generation, crit-
ical infrastructure implications, and environmental changes
alongside the corresponding geographic information.
Ongoing and Future Work
The climate analytics study will continue to investigate the
potential impact of different temperature scenarios on the
WECC grid. We will study the influence of solar activities
and humidity on California building energy consumption.
The correlation of temperature in some coastal areas is not
as strong as the correlation in inland areas. Factors other
than temperature will also be included.
The building, energy, and power analytics study will
finish up the building energy model work for all ten se-
lected cities, run the PSLF regional level simulation to
study the system level impact, work with the social analyt-
ics colleagues to investigate and integrate the results of the
social analytics studies, and identify the potential impact to
the power grid infrastructure. We will also investigate ad-
verse impact of the extreme events on power system secu-
rity and reliability.
The social analytics study will continue to insert new
social factors into our predictive model. Major topics to
consider include lifestyle changes, economic sectors, mar-
ket behaviors, and policy changes. The results will be re-
dispatched by the model to study their impacts on the pow-
er grids.
The infrastructure analytics study will investigate mul-
tiple critical infrastructures related to the power grid infra-
structure. Network graph theories will be applied exten-
sively to investigate the dependencies among the affected
infrastructures. The outcomes will be correlated with the
PSLF simulation results to discover hidden security vulne-
Figure 5. Predicted changes of building energy consumption
for grocery store between 1991-2000 and 2045-2054: 1) No
technology improvement, 2) set point change, 3) more efficient
lighting, 4) and more efficient AC system.
Figure 6. Buses (network nodes) that are ranked highly in
different centrality measures are shown in red (betweenness),
green (closeness), blue (degree), and orange (eigenvector).
152

Page 6

rabilities and identify necessary security safeguards and/or
corrective actions.
The visual analytics study will continue to work on a
prototype system that links all four models together into a
working analytical tool. Much of the information under-
neath the system front-end will be made available for the
users through interactive data visualization and navigation
means.
Additionally, we will verify the accuracy of our work
and results using our models to predict both the past and
future. The past model results will be compared with his-
torical archives from sources such as U.S. Climate Change
Science Program (CCSP) (Climate Change Science Pro-
gram 2008) and Western Electricity Coordinating Council
(WECC) model validation working group.
Conclusion
The paper discusses our ongoing technosocial analytics re-
search and development on predicting the impact of cli-
mate change on U.S. power grids and its wider implica-
tions on national security. Preliminary results from the past
year provide significant evidence to support our hypothes-
es for individual components; however we will learn more
when we have established all the linking among the com-
ponents. We plan to report more in-depth results and dis-
cussion in the future when all the required linkages among
the domain models, as depicted in Figure 1, are fully estab-
lished.
Acknowledgments
This work has been supported by the Technosocial Predic-
tive Analytics Initiative (TPAI) at the Pacific Northwest
National Laboratory under the Laboratory Directed Re-
search and Development (LDRD) program. The Pacific
Northwest National Laboratory is managed for the U.S.
Department of Energy by Battelle Memorial Institute under
Contract DE-AC05-76RL01830.
References
Banavar, J.R., Colaiori, F., Flammini, A., Maritan, A., and Rinal-
do, A. 2000. “Topology of the Fittest Transportation Network,”
Phys. Rev. Lett., 84(20), 4745-4748.

Climate Change Science
http://www.climatescience.gov.

DOE-2 2008. http://doe2.com.

DOE-2 Weather
http://gundog.lbl.gov/dirsoft/d2whatis.html.

Faloutsos, M., Faloutsos, P., and Faloutsos, C. 1999. “On Power-
Law Relationships of the Internet Topology,” Comput. Com-
munc. Rev., 29(4), 251-262.

Program (CCSP)
2008.
Processor 2008.
Intergovernmental Panel on Climate Change (IPCC) 2008.
http://www.ipcc.ch/.

Meehl, GA and Tebaldi C. 2004. “More Intense, More Frequent,
and Longer Lasting Heat Waves in the 21th Century,” Science,
vol. 306, no. 5686, pp. 994-997, August 2004.

Newman, M.E.J. 2008. “The Mathematics of Network,” The New
Palgrave Encyclopedia of Economics," 2nd edition, Blume, L.E.
and Durlauf, S.N. eds., Palgraf Macmillan, Basingstoke.

National Infrastructure Simulation and Analysis Center (NISAC)
2008.
http://www.sandia.gov/mission/homeland/programs/critical/nisac.
html.

North American Electric Reliability Corporation (NERC) 2008.
http://nerc.com.

Positive Sequence Load Flow Software (PSLF) 2008.
http://www.gepower.com/prod_serv/products/utility_software/en/
ge_pslf/index.htm.

Program for Climate Model Diagnostics and Intercomparison
2008, http://www-pcmdi.llnl.gov/.

Technosocial Predictive Analytics Initiative (TPAI) 2008,
http://predictiveanalytics.pnl.gov/.

United States Congress, Office of Technology Assessment 1990.
“Physical Vulnerability of Electric System to Natural Disasters
and Sabotage,” OTA-E-453, Washington, DC: U.S. Government
Printing Office, June 1990.

Vulnerability of Food Security and Energy Infrastructures to Cli-
mate Change and
http://predictiveanalytics.pnl.gov/projects/food_security_energy_i
nfra.stm
Wasserman, S. and Faust, K. 1994. Social Network Analysis, Me-
thods and Applications, 1st Ed., Cambridge University Press, New
York.

Western Electricity Coordinating Council (WECC) 2008.
http://www.wecc.biz.

Xu, P., Huang, Y.J., Miller, N., and Schlegel, N. 2007. “Effects of
Global Climate Change on Building Energy Consumption and its
Implications on Building Energy Codes and Policy in California,”
Preparing Public Interest Energy Research (PIER) Program
Technical Research Project Reports, California Energy Commis-
sions.
Terrorism 2008.
153