United States
Environmental Protection
Agency
MARKAL Scenario
Analyses of Technology
Options for the Electric Sector
The Impact on Air Quality
-------
EPA/600/R-06/114
September 2006
MARKAL Scenario Analyses of Technology Options for the
Electric Sector: The Impact on Air Quality
By
Timothy L. Johnson, Joseph F. DeCarolis
Carol L. Shay, Daniel H. Loughlin, Cynthia L. Gage, and Samudra Vijay
National Risk Management Research Laboratory
Air Pollution Prevention and Control Division
Research Triangle Park, North Carolina 27711
United States Environmental Protection Agency
Office of Research and Development
Washington DC 20460
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and Development
funded and managed the research described here. It has been subjected to the Agency's review
and has been approved for publication as an EPA document. Mention of trade names, products,
or services does not convey, and should not be interpreted as conveying, official EPA approval,
endorsement or recommendation.
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the
Nation's land, air, and water resources. Under a mandate of national environmental laws, the
Agency strives to formulate and implement actions leading to a compatible balance between
human activities and the ability of natural systems to support and nurture life. To meet this
mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage our
ecological resources wisely, understand how pollutants affect our health, and prevent or reduce
environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for
investigation of technological and management approaches for preventing and reducing risks
from pollution that threaten human health and the environment. The focus of the Laboratory's
research program is on methods and their cost-effectiveness for prevention and control of
pollution to air, land, water, and subsurface resources; protection of water quality in public water
systems; remediation of contaminated sites, sediments and ground water; prevention and control
of indoor air pollution; and restoration of ecosystems. NRMRL collaborates with both public and
private sector partners to foster technologies that reduce the cost of compliance and to anticipate
emerging problems. NRMRL's research provides solutions to environmental problems by:
developing and promoting technologies that protect and improve the environment; advancing
scientific and engineering information to support regulatory and policy decisions; and providing
the technical support and information transfer to ensure implementation of environmental
regulations and strategies at the national, state, and community levels.
This publication has been produced as part of the Laboratory's strategic long-term research plan.
It is published and made available by EPA's Office of Research and Development to assist the
user community and to link researchers with their clients.
Sally Gutierrez, Director
National Risk Management Research Laboratory
in
-------
Table of Contents
List of Figures v
List of Tables vii
Acronyms and Abbreviations viii
Executive Summary x
Section 1: Electricity Generation and Air Quality 1
1.1 Electricity Production and Demand 1
1.2 Electricity and Air Quality 4
1.3 Technology Assessment, Scenario Analysis, and the Use of MARKAL 5
1.4 Overview of the Electric Sector Technology Assessment 7
Section 2: EPA National Model Baseline 8
2.1 Technology Characterization 8
2.2 Model Constraints 10
2.3 Baseline Results 15
Section 3: Sensitivity Analysis of the EPANM Database 21
3.1 Background- Sensitivity Analysis 21
3.2 Methodology 22
3.3 Selecting Inputs and Outputs 23
3.4 Global Sensitivity Results - Correlation Analysis 25
3.5 Global Sensitivity Results -Normalized Linear Regression 29
3.6 Parametric Sensitivity Analysis 38
3.7 Summary of Observations from Sensitivity Analysis 45
Section 4: The Future Role of Nuclear Energy in the U.S 47
4.1 Introduction 47
4.2 Nuclear Technology Representation in MARKAL 47
4.3 MARKAL Analysis 51
4.4 Conclusions 60
Section 5: The Air Quality Implications of Carbon Capture and Sequestration in U.S. Electric
Markets 62
5.1 CCS Technology Background 63
5.2 Implementation of CCS in the U.S. EPA National MARKAL Model 68
5.3 CCS Results and Analysis 70
5.4 Conclusions 80
Section 6: Nuclear Power, Carbon Capture and Sequestration, and Wind Generation: The
Competition Among Technology Alternatives 81
Section 7: Limitations, Future Model Development, and Analysis Directions 89
Section 8: References 92
IV
-------
List of Figures
Figure 1.1: Energy Intensity and Growth Rate Trends for U.S. GDP and Electricity Use 2
Figure 1.2: Fuel Sources for U.S. Electricity Generation 3
Figure 1.3: U.S. Distribution of Electricity Usage in 2004 4
Figure 2.1: Generic Electric Sector RES 8
Figure 2.2: Baseline Electric Generation 16
Figure 2.3: Electricity Use by Demand Sector 16
Figure 2.4: Current Capacity and Capacity Additions 17
Figure 2.5: Electric Generation by Fuel Type 18
Figure 2.6: SO2 Retrofits 19
Figure 2.7: NOX Retrofits 19
Figure 2.8: Carbon Emissions in the Electric Sector 20
Figure 2.9: NOx and SC>2 Emissions in the Electric Sector 20
Figure 3.1: Factors affecting use of coal in electricity generation 30
Figure 3.2: Factors affecting use of natural gas in electricity generation 31
Figure 3.3: Factors affecting use of oil in electricity generation 32
Figure 3.4: Factors affecting CO2 emissions from electricity generation 33
Figure 3.5: Factors affecting use of NOx controls on coal-fired power plants 34
Figure 3.6: Factors affecting system-wide CO2 emissions 35
Figure 3.7: Factors affecting system-wide SO2 emissions 36
Figure 3.8: Factors affecting system-wide NOx emissions 37
Figure 3.9: Factors affecting fossil fuel imports 38
Figure 3.10: Impact of natural gas cost on electricity generation by fuel type 39
Figure 3.11: Impact of oil cost on electricity generation by fuel type 40
Figure 3.12: Impact of nuclear capacity on electricity generation by fuel type 41
Figure 3.13: Impact of the new electricity generation technology hurdle rate on electricity
generation by fuel type 42
Figure 3.14: Changes in CO2 emissions from the electric sector in response to changes in inputs
43
Figure 3.15: Changes in system-wide CO2 emissions in response to changes in inputs 44
Figure 3.16: Changes in net imports in response to changes in inputs 45
Figure 4.1: Schematic diagram of the light water reactor fuel cycle in the EPANMD 49
Figure 4.2: Schematic diagram of the high temperature gas-cooled reactor fuel cycle in the
EPANMD 50
Figure 4.3: Electricity generation in a baseline run of EPANMD with both conventional and
advanced nuclear technologies available 52
Figure 4.4: Change in the 2030 nuclear share of total electricity production as coal price is
increased parametrically 53
Figure 4.5: Change in the 2030 nuclear share of total electricity production as natural gas price is
increased parametrically 53
Figure 4.6: System-wide and electric sector CO2 emissions relative to baseline scenario 55
Figure 4.7: Electricity generation by technology from non-BAU carbon trajectories and both
conventional and advanced nuclear technologies are available 57
Figure 4.8: Conventional and advanced nuclear capacity over time, compared with the maximum
capacity allowed by the growth rate constraints 58
-------
Figure 4.9: SO2 and NOx emissions relative to the baseline scenario under the low carbon
scenarios presented in Figure 4.7 59
Figure 4.10: Electricity generation by generation source with a carbon trajectory reflecting 80
percent of 1995 levels from 2015 to 2030 60
Figure 5.1: Schematic of the three generic routes to carbon capture and sequestration 65
Figure 5.2: Electricity generation per-period by technology class for the baseline CCS scenario
72
Figure 5.3: Electric sector CO2, SO2, and NOX and economy-wide CO2 emissions as a function
of time for the baseline CCS scenario 73
Figure 5.4: Electricity generation per-period by technology class for the 80 percent emissions
scenari 75
Figure 5.5: Electricity generation per-period by technology class for the 50 percent emissions
scenario 75
Figure 5.6: Electric sector CO2, SO2, and NOx and economy-wide CO2 emissions as a function
of time 76
Figure 5.7: Share of electricity generation for conventional gas and all CCS technologies under
three gas-price scenarios 79
Figure 6.1: Electricity generation per period by technology class for the baseline CCS scenario
82
Figure 6.2: Electricity generation per-period by technology class for the 50 percent emissions
scenario 83
Figure 6.3: Electric sector SO2, and NOx emissions as a function of time for the 50 percent
emissions scenario 83
Figure 6.4: Fraction of 2030 electric power generation by technology class for five scenarios.. 84
Figure 6.5: Fraction of 2030 electric power generation by technology class for four scenarios
with the relaxed wind growth constraints 86
Figure 6.6: Electricity generation per-period by technology class for the 50 percent emissions
scenario 87
VI
-------
List of Tables
Table 1.1: Percent of Total Emissions from U.S. Electric Generation Technologies 4
Table 2.1: Existing Electric Generation Technologies 9
Table 2.2: Electricity Generation Technologies and Associated Cost Parameters 11
Table 2.3: Lifetime and Availability Parameters for Electricity Generation Technologies 12
Table 2.4: Electricity Transmission and Distribution Costs 13
Table 2.5: Total Allowed Generating Capacity of Renewable Technologies in GW 13
Table 2.6 Electric Sector NOx and SC>2 Emission Constraints in the EPANMD 13
Table 2.7: Emission Factors for CO2 and SO2 and NOX in the EPANMD 14
Table 2.8: Emission Control Retrofit Data 15
Table 3.1: Ranges for inputs used in the Monte Carlo simulation 24
Table 3.2: Correlation coefficients between model inputs and electric sector outputs 25
Table 3.3: Correlation coefficients between model inputs and system-wide outputs 27
Table 3.4: Correlation coefficients between electric sector and all outputs 28
Table 3.5: Correlation coefficients between system outputs and all outputs 29
Table 4.1: Cost and performance estimates for LWRs used in EPANMD 48
Table 4.2: Cost and performance estimates used in EPANMD for HTGRs 51
Table 4.3: Annual maximum growth rates by model time period and incremental capacity
allowable over the growth rate 51
Table 4.4: GW of advanced nuclear added by model time period 60
Table 5.1: Estimated cost and performance ranges from the literature for new CCS
technologies as summarized by the IPCC 66
Table 5.2: CCS technology parameters as implemented in the EPANMD 68
Table 5.3: Baseline CCS scenario electric power generation by technology class and time .... 71
Table 5.4: Baseline CCS scenario new capacity investment by technology class 72
Table 5.5: Share of electricity generation in 2030 by technology for the initial CCS scenario
and four parametric analyses of differences in CCS technology costs relative to baseline
values 77
Table 5.6: Share of electricity generation in 2030 by technology for the initial CCS scenario
and four parametric analyses of differences in CCS retrofit technology costs and efficiencies
relative to baseline values 78
vn
-------
Term
Acronyms and Abbreviations
Definition
AEO Annual Energy Outlook
AER Annual Energy Review
ANOVA analysis of variance
AQA air quality assessment
BAU business as usual
BTU British thermal units
CAA Clean Air Act
CAIR Clean Air Interstate Rule
CCS carbon capture and sequestration
CCSP Climate Change Science Program
CO carbon monoxide
CO2 carbon dioxide
Conv Nuc scenario of relaxed wind constraints and nuclear limited to conventional
DOE Department of Energy
EIA Energy Information Administration
ELC electric
EOR enhanced oil recovery
EPA Environmental Protection Agency
EPA9R EPA 9 region MARKAL database
EPANMD EPA national model MARKAL database
EPRI Electric Power Research Institute
FGD flue gas desulfurization
GDP gross domestic product
GHG greenhouse gas
GJ gigajoules
Gt giagtonnes
GT-MHR gas turbine-module helium reactor
GW gigawatts
H2 hydrogen
H2S hydrogen sulfide
HTGR high temperature gas-cooled reactor
IE A International Energy Agency
IGCC integrated gasification combine cycle
IHM initial heavy metal, i.e. enriched uranium
IPCC International Panel on Climate Change
ISA-W Integrated Systems Analysis Workgroup
kg kilogram
kt kilotonnes
kW kilowatts
kWh kilowatt hours
LNB low NOX burner
LHV lower heating value
LWR light water nuclear reactor
Vlll
-------
MARKAL MARKet ALlocation energy-systems computer model
MOX mixed oxide reactor
M million (mega)
MtCO2 megatonnesofCO2
MWd megawatt days
MWe megawatt electricity
N2 nitrogen
NETL National Energy Technology Laboratory
NGA natural gas
NGA MU scenario with natural gas price mark up
NGCC natural gas combined cycle
NGL natural gas liquids
NOx nitrogen oxides
2x Nuc Inv scenario with nuclear investment costs doubled
62 oxygen
O&M operation and maintenance
ORD Office of Research and Development
PBMR pebble bed modular reactor
PC pulverized coal
PJ petajoules
PM particulate matter
PMio PM with aerodynamic diameter 10 |j,m or less
PM2.5 fine PM with aerodynamic diameter of 2.5 |j,m or less
PU plutonium
PUREX plutonium and uranium extraction
PV photovoltaics
R2 sum of squared residuals
RES reference energy system
SCR selective catalytic reduction
SNCR selective non-catalytic reduction
SC>2 sulfur dioxide
TAG technology assessment guide
Th thorium
TIHM tons of initial heavy metal
U uranium
VOC volatile organic carbon
yr year
IX
-------
Executive Summary
The U.S. EPA contributes to the U.S. Climate Change Science Program (CCSP) by working to
develop an understanding of the potential human health, ecosystem, and socioeconomic impacts
of global change. A central part of this work builds on traditional EPA expertise by examining
the connection between climate and air quality. Climate variability will likely result in changes
in regional meteorology. These meteorological changes, in turn, may affect air pollution levels
by altering atmospheric chemical fate and transport processes as well as biogenic and
anthropogenic emissions. To characterize these changes, EPA's Office of Research and
Development (ORD) is working on a multiyear Global Change Air Quality Assessment (AQA)
to explore the potential consequences of global change on criteria pollutant emissions through
midcentury. The research has advanced through a series of projects geared towards building the
ability to analyze the relationship between global change and air quality and involves research
teams with expertise in meteorological, emissions, and air quality modeling, as well as
technology assessment. The research also takes into account demographic, economic, and
technological changes that would be expected to occur independent of global change.
EPA's Integrated Systems Analysis Workgroup (ISA-W) contributes to the AQA by crafting
scenarios of plausible technology change and assessing how this evolution might impact air
pollutant emissions. These activities focus on the two economic sectors that impact air quality
the most: transportation and energy (other sectors are considered to capture important system
effects). Combined, the sectors account for roughly two-thirds of the pollutants that impact air
quality and are areas where significant technological changes are expected to occur over the next
several decades. EPA researchers will use the results of ISA-W s technology assessments and
corresponding emission growth rates from the scenario analyses to calculate the future emission
profiles needed as input to the EPA's air quality models. In 2004, ISA-W produced a report
which outlined its approach for technology assessments and provided initial results for the
transportation sector. The present technology assessment provides a companion piece focused
on electricity generation.
The future of air quality will be controlled by two primary factors: pollutant emissions
(influenced by technology use and adoption) and ambient temperature (influenced by climate
change). This report presents a series of scenario-driven technology assessments focused the first
of these drivers by examining different "classes" of power generation technologies. The
assessment investigates the barriers and potential limits to the market penetration through 2030
of both conventional and alternative technologies, and examines the air emission impacts from
their use. Information from this assessment will aid the selection of scenarios for inclusion in
ORD's 2010 AQA synthesis report.
The report first provides a general overview of EPA's national MARKAL database and energy
systems model (EPANMD) and presents results for the business as usual (BAU) baseline
scenario. Under baseline assumptions, total electricity use increases 1.3% annually from 13,378
PJ in 2000 to 19,622 PJ in 2030. Annual growth in electricity demand varies between 1.0% in
the residential sector, to 2.1% in the commercial and 1.5% in the industrial sectors of the U.S.
economy. A total of 293 GW of new electric generation capacity is added between 2000 and
2030 to meet this growth. More than 76% of the new capacity is natural gas technologies, with
-------
61% being natural gas combined cycle and 15% being natural gas combustion turbines. New
conventional coal-fired power plants are not added until 2020, though a small amount of
integrated gasification combined cycle generation comes on-line in 2015. Renewables add 34
GW of capacity, with 61% coming from wind power generation, 15% from biomass combined
cycle, and 14% from geothermal.
Coal-fired power plants continue to generate the most electricity in the BAU scenario. As the
emissions constraints in the electric generation sector tighten over time, much of the existing
capacity is retrofitted with NOx and SO2 controls and 20 GW of new capacity is built. New coal
capacity includes FGD for SO2 and SCR for NOx and therefore does not need retrofit
technologies for emissions controls. Nuclear power capacity increases slightly, but this is due to
capacity upgrades at existing facilities; no new nuclear plants are built in the business-as-usual
case. Overall, coal electric generation grows 0.5% annually, natural gas grows 3.8%, and
renewables grow 2.7%.
The model constrains NOX and SO2 starting in 2010 and 2015, respectively, to conform with
DOE projections of the impacts from the Clean Air Act regulations. These constraints are
binding over the applicable time periods and emissions do not drop below the upper limits. CO2
emissions in the electric sector grow 1% annually from 689,296 kt in 2000 to 927,302 kt in 2030.
The EPANMD representation of coal plant retrofits is simplified in that it averages emissions
over the whole country and ignores regional issues related to air quality. As a result, the retrofit
coal capacity is smaller than it actually is (or will be), and when retrofits are selected they tend to
be lower cost technologies with lower removal efficiencies. Retrofits for SO2 control are
currently installed on 4.8% of existing coal fired power plants. Tightening of emissions
constraints leads to FGD units being installed on over 14% of plants. Low NOx Burners (LNB)
for NOx control are currently on 80% of existing plants. Over time, SNCR units are added to the
existing LNB units for increased emissions reductions. In later years, SCR units play a larger
role (SCR would be favored from the start if the EPANMD captured emissions trading).
Sensitivity and uncertainty analyses show that much of the EPANMD electric sector's behavior
appears to be influenced by whether specific technologies and fuels meet base or peak load
electricity demands. The predominant base load technologies are coal-fired power plants and
nuclear power plants. Coal is the most competitive, and has the largest market share. Natural
gas- and oil-fueled technologies are used to meet peak electricity demands. Both fuels experience
some cross-sector interactions with the transportation sector because natural gas and refined oil
products can be used within vehicles. While there was some evidence of cross-sector
interactions, these were largely of secondary importance to in-sector technology and fuel
competition.
The electric sector emissions constraints have interesting effects on the generation technology
mix. When natural gas becomes more expensive and natural gas technology utilization
decreases, for instance, use of coal also decreases. This behavior, which may at first seem
counter-intuitive, is explained by the system's response to the electric sector NOx constraint.
Since coal technologies had higher NOx emissions than natural gas technologies, and since NOx
constraints on electricity generation were binding, the model opted to replace natural gas with
XI
-------
oil. Oil combustion also leads to greater NOx emissions than that from natural gas, thought it is
less than coal. Oil therefore displaces some coal. This fuel-switching and related technology
change had implications on CO2 emissions as well.
The electric sector NOx constraint also affects the model's response to increased nuclear
capacity. By introducing electricity generation capacity that does not have NOx emissions, coal-
fired power plants are less constrained in meeting the electric sector NOx constraint. In response,
the fraction of coal-fired plants projected to make use of NOX controls, such as SCR, decreases.
The same behavior can be expected from increased electricity generation from low- or zero-NOx
renewables, such as solar or hydropower.
The electricity generation investment hurdle rate has an additional impact on future-year energy
sector technologies. Increasing the hurdle rate effectively makes it more difficult for new,
efficient technologies to penetrate the electricity generation market. This has the effect of
increasing the marginal peak electricity price, increasing CO2 emissions from electricity
generation, and decreasing the penetration of renewables. Thus, addressing hesitancy to adopt
new technologies through some approaches for hedging risk may yield a more efficient
electricity generation system.
Imports of fossil fuels are correlated with the use of oil in electricity generation, but inversely
correlated with natural gas use, reflecting the fact that natural gas demand is largely met by
domestic supplies in the model. Cross-sector fuel switching resulting from changes in natural
gas and oil consumption in the transportation sector may also have an impact, but further
analysis is needed to characterize this behavior.
Finally, changes in system-wide CO2 emissions in response to variation in model inputs were
minor, with decreases of less than 3% observed. This output is influenced by the inability of low-
CO2 emitting technologies, and in particular, renewables, to achieve high market penetrations.
When these technologies do penetrate, the potential reductions in CO2 emissions are often offset
by increased use of coal and other fossil fuels, made possible via the room under the NOx limit
created by the renewables.
Moving beyond its detailed sensitivity analysis of BAU results, the assessment focuses on the air
quality impacts of two advanced electric sector technologies: nuclear power and carbon capture
and sequestration (CCS). The analysis first treats these technologies separately in order to
examine their independent impacts on air quality relative to the BAU results; a final series of
scenarios allows competition between these advanced technologies and a rudimentary
representation of wind power.
The nuclear scenario results offer important insights into the role that this family of technologies
can play in the U.S. electricity system over the next three decades. Increased natural gas and
coal prices, for instance, do not have a significant impact on the penetration of new nuclear units.
As a result, nuclear power only provides a modestly effective hedge against high fossil fuel
prices in the electric sector, at least over the fuel price ranges explored here.
xn
-------
Though modest, these nuclear technology penetration levels contribute to meeting the electric
sector's CAA limits for SO2 and NOx. Nuclear generation, however, has a limited impact on
system-wide carbon emissions. When the electric sector follows an arbitrary reduced carbon
emissions trajectory, nuclear capacity plays a more significant role and leads to SO2 and NOX
reductions below CAA constraints.
Although all nuclear technology options are economical, the model prefers advanced high
temperature gas-cooled reactors (HTGRs) to conventional designs such as light water and mixed-
oxide reactors. The issues governing this preference include whether HTGR developers can
meet the model's optimistic nth-of-a-kind cost assumptions and how rapidly units can be
manufactured and deployed over the next few decades.
The assessment also approaches CCS from a combined technology adoption and air quality
perspective. The EPANMD's energy systems framework is important in this analysis as CCS
would compete with measures to reduce the carbon intensity of power production and efforts to
improve end-use efficiency. Driven by an arbitrary electric sector CO2 trajectory, for instance,
the model relies on coal-to-gas fuel-switching as much as it does on CCS. The model adopts
CCS mainly for baseload generation in the form of new IGCC capture capacity. CO2 capture
retrofits are not an economically attractive option as retrofit capacity incurs a significant energy
penalty. Only when this penalty is reduced does retrofit capacity generate a significant share of
the electric sector's output; model results are less sensitive to retrofit technology costs. Like the
nuclear power analysis, moderate levels of CCS adoption do not significantly affect electric
sector criteria pollutant emissions. CCS displaces new (conventional) baseload capacity when
the electric sector follows a non-BAU carbon trajectory, but this adoption does not significantly
impact the operation of existing coal-fired plants, which contribute most of the sector's CAA
constrained SO2 and NOx emissions.
Several key results emerge from an integrated assessment of nuclear, CCS, and wind power.
First, the penetration of advanced electric sector technologies does not necessarily yield a
significant criteria pollutant benefit. Under all but the most radical departures from BAU electric
sector carbon trajectories, for instance, nuclear power and CCS merely replace the new coal and
gas capacity that are needed to meet increasing electricity demand. The model maintains its
existing coal plants through their useful lifetime, and criteria pollutant emissions remain near
their CAA limits as a result. This picture changes when electric sector carbon emissions depart
furthest from their BAU levels and CO2 emissions from baseload plants declines. Even in these
cases, however, the decrease in NOx emissions is more modest than that seen in SO2.
Second, CCS is dominated by investment in new nuclear technologies as well as additional gas-
fired generation and wind power. Only when favorable CCS cost and efficiency assumption
combine with high nuclear investment cost and gas price scenarios does CCS approach its
modeled growth constraints over the full time horizon in which it is available (2015-2030).
Furthermore, given that CO2 capture and geologic sequestration are both untested at even the
modest scales adopted here, the model results are probably optimistic.
Finally, even with a rudimentary conception of wind resources, wind generation in the
EPANMD competes for a significant share of electric power output when the model incorporates
Xlll
-------
more favorable growth constraint assumptions for new turbine investment. These growth limits
are modest compared to some predictions, though a more realistic representation of wind turbine
dispatch is needed to capture the technology's intermittency and evaluate its actual potential.
While the model results provide insight into future energy technology pathways, they are
nonetheless based on a model, and all models have limitations that introduce caveats. The use of
national rather than regional inputs to MARKAL is perhaps the most significant limitation
affecting the analysis. The EPANMD is a national database that does not contain region-specific
data. The lack of regional data manifests itself in three key limitations. First, important regional
differences in resource supplies, energy service demands, and technology availability cannot be
represented. Second, the EPANMD does not include transportation costs for coal or other
resources. Third, air pollution regulations that have been implemented on a state or regional
level must be modeled at a national level within EPANMD. ISA-W is currently developing a 9-
region MARKAL database that will account for variations in these factors. Beyond its lack of
region-specific detail, the EPANMD does not model unit dispatch, and contains a very generic
representation of the existing fleet of U.S. coal-fired power plants. Equally important, the
analysis excludes consideration of efficiency improvements as well as non-traditional generating
technologies which could supply a large share of the nation's electricity in the coming decades.
Likewise, the analysis does not consider the impact of radical technological innovation, such as
the emergence of "nano-bio-info" technologies, on future energy demand.
xiv
-------
Section 1
Electricity Generation and Air Quality
The U.S. EPA contributes to the U.S. Climate Change Science Program (CCSP) by working to
develop an understanding of the potential human health, ecosystem, and socioeconomic impacts
of global change. A central part of this work builds on traditional EPA expertise by examining
the connection between climate and air quality. Climate variability will likely result in changes
in regional meteorology. These meteorological changes, in turn, may affect air pollution levels
by altering atmospheric chemical fate and transport processes as well as biogenic and
anthropogenic emissions. To characterize these changes, EPA's Office of Research and
Development (ORD) is working on a multiyear Global Change Air Quality Assessment (AQA)
to explore the potential consequences of global change on criteria pollutant emissions through
midcentury. The research has advanced through a series of projects geared towards building the
ability to analyze the relationship between global change and air quality and involves research
teams with expertise in meteorological, emissions, and air quality modeling, as well as
technology assessment. The research also takes into account demographic, economic, and
technological changes that would be expected to occur independent of global change.
EPA's Integrated Systems Analysis Workgroup (ISA-W) contributes to the AQA by crafting
scenarios of plausible technology change and assessing how this evolution might impact air
pollutant emissions. These activities focus on the two economic sectors that impact air quality
the most: transportation and energy (other sectors are considered to capture important system
effects). Combined, the sectors account for roughly two-thirds of the pollutants that impact air
quality and are areas where significant technological changes are expected to occur over the next
several decades. EPA researchers will use the results of ISA-W s technology assessments and
corresponding emission growth rates from the scenario analyses to calculate the future emission
profiles needed as input to the EPA's air quality models. In 2004, ISA-W produced a report
which outlined its approach for technology assessments and provided initial results for the
transportation sector (Gage et al., 2004). The present technology assessment provides a
companion piece focused on electricity generation.
1.1 Electricity Production and Demand
Reliable access to electricity is fundamental to economic productivity, quality of life, and the
comforts of daily existence. Since its development in 1882 as a power source to operate 800 of
Thomas Edison's electric light bulbs (Edison Electric Institute, 2006), the U.S. electric power
generating capacity has increased to over 1,051,000 megawatts. Electricity is now the primary
power source for most commercial and residential building end-use services, including lighting,
space heating and cooling, water heating, ventilation, and refrigeration. The technologies on
which the information economy rest are also dependent on electricity, and more traditional
manufacturing industries like chemical and paper production remain major consumers of electric
power.
-------
Demand for electricity has a strong historical linkage to the development of the U.S. economy.
Figure 1.1 illustrates the relationship between annual change in electricity use and annual growth
rate of Gross Domestic Product (GDP) for the past 55 years. Over this period, the energy used to
produce electricity climbed from about 5,300 PJ in 1950 to almost 42,100 PJ in 2005 (EIA,
2005b). As a result, the relative energy intensity of the economy doubled between 1950 and 1977
(see Figure 1.1). While both electricity use and GDP have continued to grow, energy intensity
has declined by 33% since its mid-1970s peak (BEA, 2006; EIA, 2005b). Improvements in end-
use energy efficiency and the shift from a manufacturing- to a service-based economy are largely
responsible for this decline.
14%
12%
GDP
Electricity Use
Intensity
1.6
1.4
-4%
1950
1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
Figure 1.1: Energy Intensity and Growth Rate Trends for U.S. GDP and Electricity Use;
Sources: BEA (2006), EIA (2005b)
A variety of fuels serve as primary energy sources for U.S. electricity production, as Figure 1.2
illustrates (EIA, 2005b). Coal, which is abundant in the U.S. has dominated electricity
production and currently accounts for 50% of U.S. generation. Oil was an important source until
the 1970's when U.S. production began declining and the Arab oil embargo caused import
shortages. Oil now fuels 3% of electricity generation. Natural gas use increased by nearly half
during the 1990s, though more recent volatility and sustained high prices are discouraging
expanded use.
Fuel choice for electricity generation is also regionally dependent (Edison Electric Institute,
2006b). In both the West and East North Central census regions, coal is the dominant energy
fuel for electricity production, with generating shares of 77% and 70% respectively. In New
England, natural gas dominates at 38%. Hydroelectric dominates in the contiguous Pacific
-------
region at 41%; and oil, at 52%, dominates in the noncontiguous Pacific. In the Middle Atlantic
region, nuclear and coal account for 35% and 36%, respectively.
Demand for electricity is governed by the end-use energy services of the other sectors of the
economy: residential, commercial, industrial, and transportation. Figure 1.3 shows the
distribution of electricity use across these sectors for the U.S. in 2004. Residential and
commercial demand each account for about one-third of electricity use, while industrial,
including non-utility direct usage, accounts for another 30%. The U.S. transportation
infrastructure, in contrast, relies on a negligible amount of electricity. Demand for end-use
services (lighting, cooling, manufacturing, etc.) will increase further as the economy and
population expand. Projections show the U.S. GDP growing by 2.9 to 3.0% annually over the
next 25 years (EIA, 2006c) while the U.S. population expands by 29% (U.S. Census Bureau,
2006). This growth will impact electricity demand and potentially air quality.
22,000
20,000
18,000
16,000
14,000
^ 12,000
GO
c
1 10,000
8,000
6,000
4,000
2,000
Coal
Nuclear
1950 1955 1960 1965
1970
1975 1980 1985 1990
1995 2000 2005
Figure 1.2: Fuel Sources for U.S. Electricity Generation; Source: EIA (2005b)
-------
Non-Utility
Direct Use Transportation
J >0%
4%
Figure 1.3: U.S. Distribution of Electricity Usage in 2004; Source: EIA (2005b)
1.2 Electricity and Air Quality
Electricity is produced primarily from the combustion of fossil fuels. It is this combustion
process which causes the production and release of a variety of atmospheric pollutants, including
SO2, NOX, particulates, CO2, and mercury (U.S. EPA, 2005, 2006a, and 2006b). All of these
pollutants impact human health and the environment. Table 1 shows the percent of total
pollutant emissions which are produced from electricity generating technologies. SO2 and NOx
(which contribute to acid rain) can also contribute to PM2.5 and ozone formation, respectively,
further degrading ambient air quality.
Table 1.1: Percent of Total Emissions from U.S. Electric Generation Technologies; Sources:
U.S. EPA (2005), (2006a), and (2006b)
Impact
Pollutant
Emission %
Ambient Air Quality
PMio
16
PM2.5
3
Acid Rain
SO2
65
NOX
20
Toxic
Hg
43
Climate
Change
CO2
38
N2O
4
EPA has developed regulations for several of these pollutants which have already produced
significant emission reductions. The Acid Rain Program of the Clean Air Act (CAA) of 1990,
-------
for instance, regulates 862 and NOx from electric generation. This program entered Phase II of
the reduction requirements in 2000. By 2010, SC>2 emissions from power plants will be capped
at 8.95 million tons per year, a reduction of more than half from 1980 emission levels. After
implementation of Phase II, NOX emissions from power plants have declined by a total of 2.1
million tons/year from 1980 levels. In addition to the Acid Rain Program, SC>2 and NOx
emissions are affected by the Clean Air Interstate Rule of 2005 which impacts 28 Eastern states
and the District of Columbia. The Phase II caps for SO2 and NOX in 2015 will be 2.5 and 2.2
million tons, respectively.
How these caps will be met and maintained under growing demand for electricity is yet to be
determined. Options include pre- or post-combustion control of emissions from fossil fuel
sources, finding "cleaner" technologies for producing electricity, or implementing efficiency
improvements on both supply and demand.
1.3 Technology Assessment, Scenario Analysis, and the Use ofMARKAL
Technology assessment lies at the core of ISA-W's electric sector analysis. Such assessments
are multidimensional in that they require a detailed characterization of individual technologies as
well as the economic and institutional factors that both drive and constrain their use. Resource
supply costs, demand estimates for energy-related services, emission considerations, and other
drivers, for instance, combine to determine how existing and new technologies compete. Data
for these assessments comes primarily from government agencies, academic studies, other
published literature, industry studies, and individual consultations. Mid-range or consensus
estimates furnish the starting point for inputs to the typical assessment, though sensitivity
analyses examine the effects of more divergent assumptions.
A systems perspective is needed to capture the interaction of the diverse factors required by a
complete technology assessment. Such a perspective goes beyond lifecycle assessment's
"cradle-to-grave" analysis of resource needs and environmental impacts for a given technology
to examine how multiple technologies compete with each other to meet demand, and how
resource constraints affect this balance. A systems focus, therefore, captures both direct
environmental impacts (e.g., the emissions generated by a particular technology) as well as
indirect effects (e.g., how reducing demand for a fuel in one economic sector might lower its cost
and consequently increase its use—and associated emissions—in another). These indirect
feedbacks can be counterintuitive, and a systems perspective, therefore, helps capture
unanticipated consequences. In addition, real-world trade-offs in technological and economic
feasibility often emerge only at the systems level, an example being that of a high capital cost
technology (e.g., a coal-fired power plant) that only makes sense to build if its use is sufficient to
recover the initial investment.
Technology assessment within a systems framework provides the structure for ISA-W's
research. The use of scenarios guides the actual analysis of technological futures and assists in
understanding how complex systems may evolve. Scenarios are internally consistent depictions
of how the future may unfold, given assumptions about economic, social, political, and
technological developments as well as consumer preferences (Schwartz, 1996). Scenarios
explore plausible futures by using a model or models to generate an outcome (or set of
alternative outcomes) consistent with a set of motivating assumptions, sometimes called a
-------
"storyline." It is important to stress that these consequences should not be interpreted as
predictions, for example, about levels of new technology market penetration or emission
trajectories. Rather, the technology parameters and economic data used as inputs are best seen as
starting-point assumptions that reflect a range of reasonable estimates.
Scenario analysis aims to examine how changes in model parameters (inputs) affect outputs
across sets of related storylines, rather than focus on the results from a particular scenario. No
attempt is made to consider every possible future. These comparative analyses alternately look
forward ("What-if?") to examine how competing sets of input assumptions drive technology
adoption and emissions, and backward ("How-could?") to identify the energy technology
pathways available to meet some future environmental or technological goal. Scenarios,
therefore, facilitate assessment of the consequences of varying assumptions, the range of possible
futures, and trade-offs and branch points that govern choices among these futures. Results from
a selected set of scenarios will serve as input to the ORD Air Quality Assessment.
In order to investigate scenarios of future electric generation technologies and their impact on
future air pollutant emissions, ISA-W adopted the MARKAL energy-systems modeling
framework (see Shay et al. 2006). The Department of Energy's Brookhaven National
Laboratory created MARKAL (short for MARket ALlocation) in the late 1970s, and a strong
international users group has organized itself to support continuing applications and extensions.
MARKAL maps the energy economy from primary energy sources through their refining and
transformation processes to the point at which a variety of technologies (e.g., classes of light-
duty personal vehicles, heat pumps, or gas furnaces) service end-use energy demands (e.g.,
projected vehicle miles traveled, space heating). All economic sectors—industrial, residential,
commercial, and transportation—are covered. A large linear programming model, ISA-W's
MARKAL model, determines the least-cost pattern of technology investment and utilization
required to meet specified demands and model constraints and then calculates the resulting
criteria pollutant and greenhouse gas emissions.
MARKAL's strength lies in the fact that it is a systems model. The energy system is complex
and interactions must be considered, but feedbacks are not always intuitive; reducing demand for
a fuel in one economic sector, for instance, might lower its cost and, therefore, increase its use in
another. Such real-world trade-offs in technological and economic feasibility often emerge only
at the systems level. In addition, the readiness, costs, and performance of individual energy
technologies (e.g., wind turbines) must be considered in the full socio-economic context of their
use. MARKAL quantifies the system-wide effects of changes in resource supply and use,
technology availability, and environmental policy.
MARKAL is a data-driven energy-economic model. The user specifies the energy system
structure, including resource supplies, energy conversion technologies, end-use energy service
demands, and the technologies needed to satisfy these demands. The user must also provide data
to characterize individual technologies and resources, including their fixed and variable costs,
availability, performance attributes, and pollutant emissions. MARKAL is data-intensive.
Within the electric power sector, for instance, ISA-W's MARKAL database contains 24
generating technology options (both existing and future vintages). The database is divided into
-------
five-year periods that stretch from 1995 to 2035. The time horizon will be extended to 2055 in
future work to meet AQA needs.
Finally, ISA-W has also developed a suite of analytical and visual tools for evaluating sensitivity
to uncertainties in model parameters and inputs and exploring the solution space. Assessing the
sensitivity of results to assumptions is an important part of any scenario analysis. Do the results,
for instance, depend on a narrow range of input parameters? Or do several technology pathways
lead to equivalent outcomes? Confidence in model results depends on answering such questions,
and these answers may also have important policy implications.
1.4 Overview of the Electric Sector Technology Assessment
The future of air quality will be controlled by two primary factors: pollutant emissions
(influenced by technology use and adoption) and ambient temperature (influenced by climate
change). This report presents a series of technology assessments focused the first of these drivers
by examining different "classes" of power generation technologies. The assessment investigates
the barriers and potential limits to the market penetration through 2030 of both conventional and
alternative technologies, and examines the air emission impacts from their use. Information
from this assessment will aid the selection of scenarios for inclusion in ORD's 2010 AQA
synthesis report.
The following section provides a general overview of EPA's MARKAL model and presents
results for the business as usual (BAU) baseline scenario. Section 3 examines the factors driving
BAU results through an extended sensitivity analysis. The following two sections focus on
advanced generation technologies by investigating the air quality impacts of nuclear power
(Section 4) and carbon capture and sequestration (Section 5). The analyses treat these two
technologies separately in order to examine their independent impact on air quality relative to the
BAU results. Section 6 then presents a series of scenarios which allow competition between
these advanced technologies and a rudimentary representation of wind power. Finally, Section 7
discusses the assessment's limitations and outlines future work.
-------
Section 2
EPA National Model Baseline
The Electric Generation Sector in the EPA National MARKAL Database (EPANMD)
characterizes existing and new technologies available electricity generation. Based on sector-
specific electricity demand (residential, commercial, industrial, and transportation), fuel prices,
technology costs, and the environmental and operational constraints incorporated in the model,
MARKAL determines the least cost way to meet system electricity demand. This section
describes the EPANMD's representation of electric power generation technologies and presents
the baseline, or business as usual (BAU), model results. Shay et al. (2006) provides a complete
description of the EPANMD.
MARKAL is structured around a Reference Energy System (RES), a network diagram that
depicts an energy system from resource supply to end-use consumption. The RES divides an
energy system up into a series of elements, including primary energy resources plus process,
conversion, and demand technologies. These technologies feed into a final stage consisting of
end-use demands for useful energy services. End-use demands include items such as residential
lighting, commercial space conditioning, and automobile passenger miles traveled. Energy
carriers interconnect the stages. A large linear programming model, MARKAL determines the
least-cost means of meeting end-use demand over the model's time horizon, 1995-2030. (For a
detailed description of MARKAL, see ETSAP, 2004).
The electric generation sector specific RES consists of imported electricity resource technologies
and conversion technologies (e.g., existing coal steam) which output electricity to the system.
"Dummy" emissions process technologies, which have no costs, track emissions through the
system. Figure 2.1 provides a generic representation of the electric sector RES structure.
energy carriers Process Technologies energy carriers Conversion Technologies
Resource Technologies
ELC
Figure 2.1: Generic Electric Sector RES
2.1 Technology Characterization
The EPANMD contains twenty-four electric generation technologies, sixteen of which have
existing capacity in the U.S. For a complete description of the model's technology
representation, see Shay et al. (2006). EPANMD is calibrated to the 2002 Annual Energy
Outlook (EIA, 2002a), except where noted.
-------
Available Technologies
Table 2.1 lists the existing electric generation technologies represented in the EPANMD with
their initial capacities. Note that the EPANMD does not explicitly model the costs associated
with retiring these technologies.
Table 2.1: Existing Electric Generation Technologies
Technology
Coal Steam Plants
Natural Gas Steam Turbine
Conventional Nuclear LWR
Conventional Hydropower
Natural Gas Combustion Turbine
Diesel Combustion Turbine
Hydropower Pumped Storage
Residual Fuel Oil Steam
Natural Gas Combined Cycle
Biomass Combined Cycle
Diesel Internal Combustion Engine
Municipal Solid Waste
Geothermal
Wind Central Electric
Solar Central Thermal
Solar Photovoltaic
GW Existing Capacity (1995)
307.8
115.0
102.0
78.5
31.7
25.2
22.0
17.0
13.6
7.6
3.5
3.4
3.0
1.7
0.3
0.1
New Technologies
The new electric generation technologies available to the model in the base case include:
• Advanced Coal—Integrated Gasification Combined Cycle (IGCC)
• Advanced Coal—Pressurized Fluidized Bed
• Advanced Natural Gas Combined Cycle
• Advanced Natural Gas Combustion Turbine
• Distributive Generation—Baseload
• Distributive Generation—Peak
Technologies are characterized in the model by costs (investment, operation, and maintenance),
efficiencies, lifetimes, emissions, hurdle rates, fractions in peak equations, and availability.
Hurdle rates refer to discount rates applied to the investment costs of new technologies which are
meant to mimic hesitancy on the part of the purchaser to invest in a newer technology over an
established technology. The fraction in peak equations refers to the fraction of the technology's
total capacity in a specified period which can be counted on to be available to meet peak demand
and reserve margin requirements. Availability refers to the percentage of the year that a
technology is on-line and available accounting for forced and scheduled outages.
The database draws from five primary data sources, which are listed below in order of
precedence.
-------
1. Annual Energy Outlook 2002 (EIA, 2002a)
2. "Supporting Analysis for the Comprehensive Electricity Competition Act" (DOE, 1999)
3. Technical Assessment Guide (EPRI, 1993)
4. 1997 DOE MARKAL database (Tseng, 2001)
5. National Energy Technology Laboratory (NETL) (Boilanger, 2002)
Tables 2.2 and 2.3 list the main parameters and their values for the available technologies. All
values are in model base year (1995) dollars; post-2010 values remain constant.
In addition to the costs of each electric generation technology, all electricity generated by the
sector (except that from distributed generation) is subject to the transmission and distribution
costs shown in Table 2.4.
2.2 Model Constraints
Several constraints are active during the baseline run, including:
1. A 30% reserve capacity (the amount by which the installed electricity generating capacity
exceeds the average load of the season and time-of-day division of peak demand).
2. A hurdle rate of 18% (applied to all new generation technologies).
3. Nuclear capacity bound set at 2005 AEO levels (EIA, 2005a).
4. IGCC investment in IGCC limited to the 2002 AEO projections (EIA, 2002a).
5. Investment in renewables fixed at levels based on analysis by in-house researchers and 2005
AEO (EIA, 2005a) estimates (see Table 2.5).
10
-------
Table 2.2: Electricity Generation Technologies and Associated Cost Parameters
Technology Name
Existing Coal Steam
Coal Steam - 2000
Coal Steam - 2005
Coal Steam -2010
Integrated Coal Gasif. Combined Cycle - 2000
Integrated Coal Gasif. Combined Cycle - 2005
Integrated Coal Gasif. Combined Cycle - 2010
Pressurized Fluidized Bed
Coal Gasification Molten Carb Fuel Cell
Distributed Generation-Base-2005
Distributed Generation-Base-201 0
Distributed Generation-Peak
Existing Natural Gas Combined Cycle
Natural Gas Combined Cycle-2000
Natural Gas Combined Cycle-2005
Natural Gas Combined Cycle-2010
Natural Gas Advanced Combined Cycle-2005
Natural Gas Advanced Combined Cycle-2010
Natural Gas Steam
Existing Natural Gas Combustion Turbine
Natural Gas Combustion Turbine-2000
Natural Gas Combustion Turbine-2005
Natural Gas Combustion Turbine-2010
Natural Gas Advanced Combustion Turbine-2005
Natural Gas Advanced Combustion Turbine-2010
Conventional Nuclear
Diesel internal combustion engine
Residual Fuel Oil Steam
Distillate Oil Combustion Turbine
Hydroelectric Pumped Storage
Biomass Gasification Combined Cycle
Geothermal Binary Cycle and Flashed Steam
Hydroelectric
Municipal Solid Waste-Landfill Gas
Solar Central Thermal
Central Photovoltaic
Photovoltaic-Residential
Local Wind Turbine
Wind Central Electric
Heat Rate
(BTU/kWh)
11990
9419
9253
9087
7969
7469
6968
9228
7575
10991
9210
10620
8030
7687
7343
7000
6639
6350
9500
11900
11467
11033
10600
8567
8000
10800
13648
9500
11900
10280
8911
32173
10280
13648
10280
10280
10280
10263
10280
Capital Costs
(1995 million
$/kW)
n/a
1119
1110
1083
1338
1315
1287
1570
2683
623
623
559
434
456
453
448
572
526
959
322
339
336
333
446
384
3445
376
959
322
1615
1725
1746
929
1429
2539
3830
7519
1246
982
Variable O&M
Costs
(1995 million
$/kW)
2.78
3.10
3.10
3.10
0.73
0.73
0.73
3.10
24.83
13.87
13.87
21.20
0.48
0.48
0.48
0.48
0.48
0.48
0.48
0.09
0.09
0.09
0.09
0.09
0.09
0.29
8.07
0.48
0.09
2.40
2.66
0.00
4.07
0.01
0.00
0.00
0.00
7.45
0.00
Fixed O&M
Costs
(1995 million
$/kW)
14.21
21.48
21.48
21.48
29.98
29.98
29.98
38.11
34.49
3.69
3.69
11.53
14.31
14.33
14.33
14.33
13.27
13.27
28.61
5.91
5.92
5.92
5.92
8.41
8.41
77.05
0.78
28.61
5.91
15.18
41.25
64.31
12.90
88.39
43.93
9.04
118.28
8.51
23.44
11
-------
Table 2.3: Lifetime and Availability Parameters for Electricity Generation Technologies
Technology Name
Existing Coal Steam
Coal Steam - 2000
Coal Steam - 2005
Coal Steam -2010
Integrated Coal Gasif. Combined Cycle - 2000
Integrated Coal Gasif. Combined Cycle - 2005
Integrated Coal Gasif. Combined Cycle - 2010
Pressurized Fluidized Bed
Coal Gasification Molten Carb Fuel Cell
Distributed Generation-Base-2005
Distributed Generation-Base-201 0
Distributed Generation-Peak
Existing Natural Gas Combined Cycle
Natural Gas Combined Cycle-2000
Natural Gas Combined Cycle-2005
Natural Gas Combined Cycle-2010
Natural Gas Advanced Combined Cycle-2005
Natural Gas Advanced Combined Cycle-2010
Natural Gas Steam
Existing Natural Gas Combustion Turbine
Natural Gas Combustion Turbine-2000
Natural Gas Combustion Turbine-2005
Natural Gas Combustion Turbine-2010
Natural Gas Advanced Combustion Turbine-2005
Natural Gas Advanced Combustion Turbine-2010
Conventional Nuclear
Diesel internal combustion engine
Residual Fuel Oil Steam
Distillate Oil Combustion Turbine
Hydroelectric Pumped Storage
Biomass Gasification Combined Cycle
Geothermal Binary Cycle and Flashed Steam
Hydroelectric
Municipal Solid Waste-Landfill Gas
Solar Central Thermal
Central Photovoltaic
Photovoltaic-Residential
Local Wind Turbine
Wind Central Electric
Technical
Lifetime
(years)
40
40
40
40
30
30
30
40
30
30
30
30
30
30
30
30
30
30
40
30
30
30
30
30
30
40
20
40
30
50
30
30
60
30
30
30
20
20
30
Availability
Fraction
0.85
0.85
0.85
0.85
0.85
0.85
0.85
0.85
0.87
0.84
0.84
0.84
0.91
0.91
0.91
0.91
0.91
0.91
0.85
0.92
0.92
0.92
0.92
0.92
0.92
0.80
0.84
0.85
0.92
varies by
timeslice
0.80
0.64
0.44
varies by
timeslice
varies by
timeslice
varies by
timeslice
varies by
timeslice
varies by
timeslice
varies by
timeslice
Fraction* of
unavailability
that is forced
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.50
0.80
0.63
0.63
0.63
0.57
0.57
0.57
0.57
0.57
0.57
0.37
0.47
0.47
0.47
0.47
0.47
0.47
0.42
0.63
0.37
0.47
n/a
0.80
1.00
0.10
n/a
n/a
n/a
n/a
n/a
n/a
Fraction of
Capacity
for Peak
and
Reserve
0.96
0.96
0.96
0.96
0.90
0.90
0.90
0.80
0.60
0.96
0.96
0.96
1.00
0.94
0.94
0.94
0.86
0.86
0.96
0.96
0.96
0.96
0.96
0.94
0.94
0.85
0.96
0.98
0.96
0.95
0.84
0.63
0.94
0.90
0.30
0.50
0.30
0.30
0.30
*This fraction is applied to (1- availability fraction).
12
-------
Table 2.4: Electricity Transmission and Distribution Costs
Transmission Investment Cost
Transmission O&M Cost
Distribution Investment Cost
Distribution O&M Cost
Transmission Efficiency
228.75
0.1
496
0.736
93.5
million U.S. $ perGW
million U.S. $ per PJ
million U.S. $ perGW
million U.S. $ per PJ
%
Table 2.5: Total Allowed Generating Capacity of Renewable Technologies in GW
Technology
Biomass Combined Cycle
Municipal Solid Waste
Geothermal
Hydropower
Solar Central Photovoltaic
Solar Photovoltaic - Residential
Wind Central Electric
2000
7.6
3.4
3.0
79.3
0.1
0.1
2.5
2005
8.3
3.8
3.5
79.3
0.2
0.2
9.1
2010
9.0
4.2
4.1
79.3
0.2
0.2
11.0
2015
9.8
4.6
4.8
79.3
0.4
0.4
13.2
2020
10.7
5.2
5.6
79.3
0.6
0.6
15.8
2025
11.7
5.7
6.6
79.3
0.8
0.8
19.0
2030
12.7
6.4
7.7
79.3
1.3
1.3
22.8
Emissions Constraints and the Coal Steam Retrofits
The Clean Air Act (CAA) imposes emissions constraints on SO2 and NOx emissions from the
electric power sector. In order to keep model results in line with these emissions levels,
constraints on SO2 and NOx in the electric sector were developed from DOE projections and
applied in the model (Table 2.6). These constraints reflect future emissions under the existing
trading schemes, though MARKAL does not model these trading dynamics internally. Note that
the current EPANMD does not represent mercury emissions; future database revisions will add
the necessary emission coefficients and control technologies.
Table 2.6 Electric Sector NOX and SO2 Emission Constraints in the EPANMD
Emission (kt/PJ)
NOx emissions in the Electric Sector
SO2 emissions in the Electric Sector
2000
5875
11400
2005
5300
10500
2010
3500
9900
2015
3500
9000
2020
3500
9000
2025
3500
9000
2030
3500
9000
The technology-specific CO2 emission factors used in EPANMD (Table 2.7) were derived from
Table A-15 of the Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2001. SO2 and
NOX factors were taken from the 1997 DOE MARKAL database with updates (EIA, 2004; OAR,
2002).
13
-------
Table 2.7: Emission Factors for CO2 and SO2 and NOX in the EPANMD; see Shay et al. (2006)
for details.
CfifctadcKDn
(KW^J)
Coal Steam Plants
Advanced Coal Plants
Natural Gas Steam Turbine
Natural Gas Combustion Turbine and Combined Cycle
Diesel Combustion Turbine
Residual Fuel Oil Steam
25.2
25.2
15.2
15.2
19.7
21.2
NOx (kT/PJ)
Existing Controls
0.0215
0.043
0.1075
0.043
0.1075
0.1075
Improved Controls
n/a
n/a
0.0108
0.0043
0.0108
0.0108
Coal Steam Plants
Advanced Coal Plants
Diesel Combustion Turbine
Residual Fuel Oil Steam
Bit
Lignite
Sub-Bit
Bit
Lignite
Sub-Bit
SO2 (kT/PJ)
High Sulfur
0.231
0.179
n/a
4.62
3.4
n/a
Existing Controls
0.024
0.492
Medium Sulfur
0.09
0.098
0.071
1.8
1.96
1.42
Improved Controls
n/a
0.024
Low Sulfur
0.046
0.04
0.033
0.92
0.8
0.66
Bit = bituminous
Existing Steam Electric Retrofit Technologies
Coal plants account for roughly 41% of total installed U.S. electric generation capacity in 1995
and produce most of the electric sector's SO2 and NOx emissions. A proper characterization of
the amount of coal capacity with pre-existing controls as well as the cost and removal efficiency
of new control retrofits is therefore important to the model's overall performance (Table 2.8).
The EPANMD contains three coal types: bituminous, sub-bituminous, and lignite as well as
three sulfur levels: high, medium, and low. Flue gas desulfurization (FGD) costs and
efficiencies reflect the fact that FGD performance varies by both coal type and sulfur level.
Because NOx emissions are insensitive to sulfur level, the EPANMD includes three unique NOx
control processes (one for each coal type). The following NOx control processes are available to
the model:
• LNB (Low NOX Burner)
• SCR (Selective Catalytic Reduction)
• SNCR (Selective Non-Catalytic Reduction)
• LNB-SCR combination
• LNB-SNCR combination
Note that SCR and SNCR can be installed by themselves, at the same time as LNB, or after LNB
controls are in place.
14
-------
Table 2.8: Emission Control Retrofit Data
SO2 Control Summary
Bituminous
2000 Residual Capacity (PJ/yr)
Equilibrium Emissions Rate (kT/PJ)
High Sulfur
FGD
3160
0.1186
No Retrofit
2.3723
Medium Sulfur
FGD
1313
0.0440
No Retrofit
1.0536
Low Sulfur
FGD
1093
0.0148
No Retrofit
0.2950
Sub-Bituminous
2000 Residual Capacity (PJ/yr)
Equilibrium Emissions Rate (kT/PJ)
High Sulfur
FGD
N/A
N/A
No Retrofit
N/A
Medium Sulfur
FGD
1684
0.0360
No Retrofit
0.7182
Low Sulfur
FGD
1718
0.0208
No Retrofit
0.4144
Lignite
2000 Residual Capacity (PJ/yr)
Equilibrium Emissions Rate (kT/PJ)
High Sulfur
FGD
149
0.0966
No Retrofit
1.9294
Medium Sulfur
FGD
1047
0.0666
No Retrofit
1.3356
NOX Control Summary
Bituminous
2000 Residual Capacity (PJ/yr)
Equilibrium Emissions Rate (kT/PJ)
Pass Through | LNB | SCR (after LNB) | SNCR (after LNB) | SCR only | SNCR | LNB-SCR | LNB-SNCR
15,600
0.3464 0.2300 0.0477 0.1194 0.0767 0.1916 0.0460 0.1150
Sub-Bituminous
2000 Residual Capacity (PJ/yr)
Equilibrium Emissions Rate (kT/PJ)
Pass Through | LNB | SCR (after LNB) | SNCR (after LNB) | SCR only | SNCR | LNB-SCR | LNB-SNCR
3600
0.3511 0.2080 0.0307 0.0766 0.0693 0.1733 0.0416 0.1040
Lignite
2000 Residual Capacity (PJ/yr)
Equilibrium Emissions Rate (kT/PJ)
Pass Through | LNB | SCR (after LNB) | SNCR (after LNB) | SCR only | SNCR | LNB-SCR | LNB-SNCR
470
0.2006 0.1284 0.0392 0.0981 0.0428 0.1070 0.0257 0.0642
For a complete description of the EPANMD retrofit technology methodology, see Shay et al.
(2006).
2.3 Baseline Results
The baseline (or BAU) results presented in this section are MARKAL results generated running
the baseline scenario of the EPANMD and cover 1995 (the base year) to 2030.
Electric Generation
Total electricity use increases 1.3% annually from 13,378 PJ in 2000 to 19,622 PJ in 2030
(Figure 2.2). Annual growth in electricity demand varies between 1.0% in the residential sector,
to 2.1% in the commercial and 1.5% in the industrial sectors of the U.S. economy (Figure 2.3).
15
-------
25000
LU
0
2000 2005 2010 2015 2020 2025 2030
Figure 2.2: Baseline Electric Generation
o>
to
u
"s_
"5
UJ
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
•Residential
•Commercial
Industrial
2000
2005
2010
2015
2020
2025
2030
Figure 2.3: Electricity Use by Demand Sector
16
-------
Capacity Additions
To meet this demand growth, the model adds a total of 293 GW of new electric generation
capacity between 2000 and 2030. More than 76% of the new capacity is in the form of natural
gas technologies, with 61% being natural gas combined cycle and 15% being natural gas
combustion turbines (Figure 2.4). New conventional coal-fired power plants are not added until
2020 and later, though a small amount of integrated gasification combined cycle generation
comes on-line in 2015. Renewables add 34 GW of capacity, with 61% coming from wind power
generation, 15% from biomass combined cycle, and 14% from geothermal.
350 i
D existing capacity - 2000
• capacity additions 2005-2030
, ^
' / / ^
<7^ ^»
Figure 2.4: Current Capacity and Capacity Additions
Electricity Generation by Fuel
Coal-fired power plants continue to generate the most electricity (Figure 2.5). As electric sector
emissions constraints tighten over time, more of the original plants are retrofitted with NOX and
SC>2 controls, and 20 GW of new capacity is built. New coal capacity includes FGD for 862 and
SCR for NOx and therefore does not need retrofit technologies for emissions controls. Nuclear
power capacity increases slightly, but this is due to capacity upgrades at existing facilities. No
new nuclear plants are built in the BAU case. Overall, coal electric generation grows 0.5%
annually, natural gas grows 3.8%, and renewables grow 2.7%.
17
-------
25000
20000
15000
10000
5000
2000
2005
2010
2015
2020
2025
2030
Figure 2.5: Electric Generation by Fuel Type
Use of Coal Retrofits
The EPANMD representation of retrofits is simplified in that it averages emissions over the
whole country and ignores regional issues related to air quality. As a result, the retrofit coal
capacity is smaller than it actually is (or will be), and when retrofits are selected they tend to be
lower cost technologies with lower removal efficiencies.
Retrofits for SC>2 control are currently installed on 4.8% of existing coal fired power plants.
Tightening of emissions constraints leads to FGD units being installed on over 14% of plants
(Figure 2.6). Low NOx Burners (LNB) for NOx control are currently on 80% of existing plants.
Over time, SNCR units are added to the existing LNB units for increased emissions reductions.
In later years, SCR units play a larger role (Figure 2.7); SCR would be favored earlier if the
EPANMD included emissions trading.
18
-------
16%
14%
12%
10%
£
o
•g 8%
a:
s?
6%
4%
2%
0%
14.2%
14.1%
8.3%
4.5%
13.7%
13 4%
2000
2005
2010
2015
2020
2025
2030
Figure 2.6: SO2 Retrofits
o%
2000
2005
2010
2015
2020
2025
2030
I LNB a SNCR after LNB n SCR after LNB n LNB-SCR
Figure 2.7: NOX Retrofits
Emissions
CC>2 emissions in the electric sector grow 1% annually from 689,296 kt in 2000 to 927,302 kt in
2030 (Figure 2.8). The model constrains NOx and SC>2 starting in 2010 and 2015, respectively,
to conform with DOE projections of the impacts from the Clean Air Act regulations. These
constraints are binding over the applicable time periods and emissions do not drop below the
upper limits (Figure 2.9).
19
-------
w
w
w
UJ
1200000
1000000
800000
600000
400000
200000
2000
2005
2010
2015
2020
2025
2030
Figure 2.8: Carbon Emissions in the Electric Sector
12000
0
2000
2005
2010
2015
2020
2025
2030
Figure 2.9: NOx and SOi Emissions in the Electric Sector
20
-------
Section 3
Sensitivity Analysis of the EPANM Database
The MARKAL Business As Usual (BAU) case presented in Section 2 provides a projection of
the evolution of the U.S. energy system from 1995 through 2030. The BAU case was generated
using best estimates for the values of model inputs, such as the characteristics of current and
future technologies, energy service demands, and regulations on criteria pollutant emissions.
Since the true values for many of these inputs are unknown, the BAU case represents only one of
many possible outcomes. Further, it does not itself convey information regarding the sensitivity
of the energy system to alternative input assumptions.
This section describes the application of formal sensitivity analysis techniques to evaluate the
model's response to changes in input assumptions. The results aid in characterizing and
communicating the drivers that lead to such outcomes as: the penetration of particular
technologies, a decrease in pollutant emissions, or a reduction in fossil fuel imports. Sensitivity
analysis also allows one to view the BAU case in the context of the range of possible future
energy scenarios that may occur.
3.1 Background - Sensitivity Analysis
To evaluate sensitivities within the EPANM database and model, both global and parametric
sensitivity techniques were used. Global sensitivity analysis techniques typically are applied in
practice when the goal is to characterize the relationships among model inputs and outputs over a
wide range of input conditions. In contrast, parametric sensitivity analysis, also known as local
sensitivity analysis, is used to evaluate the response to a change in a single input, holding all
other inputs constant. This subsection provides background information about sensitivity
analysis techniques within these categories. For more information about sensitivity analysis, see
publications by Saltelli et al., (2000) and Cullen and Frey (1999).
Global sensitivity analysis involves perturbing multiple model inputs simultaneously and
evaluating the effects of each input or of combinations of inputs on model outputs. Inputs are
often perturbed via Monte Carlo simulation. In Monte Carlo simulation, statistical or empirical
distributions are assumed for inputs of interest. A value is sampled from each distribution, and
the resulting set of values is fed into the model. The values of relevant outputs are recorded. The
combination of a set of inputs and the corresponding outputs constitutes one potential
"realization." Typically fifty to several hundred realizations are generated in a Monte Carlo
simulation. These realizations are then evaluated using visualization and statistical techniques to
characterize the nature and strength in the relationships among inputs and outputs. *
1 Monte Carlo techniques are also used in uncertainty and risk analyses. In uncertainty analysis, the distributions of
model outputs are characterized to estimate the uncertainty associated with each. In risk analysis, the output values
are compared with a particular metric to determine the likelihood that the metric will or will not be exceeded. In
these techniques, much effort is typically put into developing the input distributions since these will have an impact
on the output distributions. In global sensitivity analysis, however, uniform distributions may suffice since the goal
is to characterize the range of potential combinations of inputs.
21
-------
A variety of global sensitivity analysis techniques are available, with the selection of the best
technique for a particular application being a function of factors such as the degree of linearity in
the input-output relationships, the number of realizations and their coverage of the possible
outcomes, and the type of sensitivity information desired. One straightforward, readily applied
approach is correlation analysis. Correlation coefficients can be calculated between any
combination of inputs and outputs to the model. The resulting coefficients indicate whether there
is a strong linear relationship between each pair and whether this relationship is positive or
negative. Correlation coefficients cannot characterize the impact of combinations of inputs on an
output and may provide misleading results if applied to nonlinear relationships. In contrast,
scatterplots and various exploratory visualization techniques can be used to identify linear or
nonlinear relationships. These techniques typically require human expertise to visually identify
relationships, and thus tend to be qualitative instead of quantitative.
Linear regression techniques also are used in global sensitivity analysis. Regression can evaluate
the impact of multiple inputs simultaneously. Further, if inputs are normalized along their
ranges, the resulting regression coefficients indicate the relative impact of each input on an
output. Like correlation coefficients, linear regression may not be well suited for characterizing
nonlinear relationships or those in which the input-output relationships involve discrete behavior.
Nonlinear regression techniques, regression trees, and statistical analysis of variance (ANOVA)
approaches may address these issues for some problems.
Parametric sensitivity analysis, in which one input is perturbed while others are held constant, is
very useful in characterizing incremental responses to changes in inputs from a base or reference
case. These responses can be characterized quantitatively, such as with a sensitivity metric or
empirical derivative, or graphically. While parametric techniques do not characterize responses
over combinations of inputs, they often play an important role both in preliminary analyses, as a
cursory means to identify sensitivities of interest, and in more detailed analyses of input-output
responses.
Global and parametric sensitivity techniques can be used independently or together. This
analysis uses two global sensitivity analysis techniques, correlation analysis and normalized
linear regression, to make broad observations regarding input-output relationships. Global
sensitivity analysis was followed by a parametric analysis to provide more detailed information
about impacts of changes in specific inputs in the context of the BAU case.
3.2 Methodology
Steps taken to carry out the sensitivity analysis of MARKAL include:
• Key model outputs were identified, as were the input assumptions expected to impact
those outputs;
• For each input, a range was estimated;
• Monte Carlo simulation was carried out, with the inputs and outputs of each of 1000
realizations tracked;
• The Monte Carlo inputs and outputs were analyzed using correlation analysis to provide
insight regarding correlations and tradeoffs;
22
-------
Monte Carlo results were also evaluated with normalized multiple linear regression to
estimate the relative impact of each input value on each output; and
From the normalized regression analysis, the inputs with the highest influence on outputs
were identified, and parametric sensitivity runs were used to evaluate model responses in
more detail.
3.3 Selecting Inputs and Outputs
With a focus on the electric sector, key outputs for the year 2030 were identified. These
included: the market penetrations of various electricity generation technologies; fuel use in
generating electricity; the summer day peak marginal electricity price; NOx, SC>2, and CC>2
emissions from the electric generation sector; utilization of existing electricity generation
capacity; and use of NOx controls on coal-fired boiler emissions. Since the electric sector
interacts with the rest of the energy system, system-wide fuel use, emissions, and fossil fuel
imports were also selected. In the category of system-wide fuel use, the fuels that were tracked
include coal, natural gas, oil, petroleum, uranium, and renewables, which include solar power,
wind power, hydropower, geothermal, and biomass. Oil and petroleum are differentiated because
oil includes domestic and imported crude oil, both of which are refined domestically, while
petroleum includes imported petroleum products.
Inputs included in the analysis were: the future costs of natural gas, coal, and oil; the hurdle rate
for new electric generation technologies; future nuclear capacity; growth bounds on renewables
(for biomass, landfill gas, hydropower, geothermal, solar, and wind); and the availability factor
and maximum capacity of coal gasification technologies. Each of these inputs was expected to
have an impact on the selected outputs.
The next step in the analysis was to characterize ranges for the inputs. Each of the inputs, which
were assumed to be independent of each other, was represented with a uniform distribution. The
bounds of the distributions are shown in Table 3.1 and reflect modeler's judgment. Fuel inputs
were allowed to range from approximately -20% to +100% of their 2030 values in the BAU case.
The range was biased toward the high end to represent uncertainties related to political instability
and concerns about resource limitations. One half of the fuel cost change was implemented in
2015, with the remaining half coming into effect in 2020.
The hurdle rate for new electricity generation technologies ranged from 0.05, the current system-
wide discount rate, to 0.20. Nuclear power capacity was allowed to range from 1995 levels to
125% of 1995 levels. One third of any increase in nuclear capacity was specified to come online
in 2015, with the remainder entering in 2020. In our modeling, nuclear power capacity is
constrained to specific levels instead of allowing nuclear to penetrate via economics. This
modeling approach was taken to reflect the assumption that policy is the major driver that limits
or expands nuclear capacity.
Ranges for growth rates for renewables were selected to encompass the default values. The
availability factor for coal gasification ranged from 82.5% to 87.5%, and the overall capacity
limit for the technology ranged from 0 to 1000 PJ.
23
-------
Table 3.1: Ranges for inputs used in the Monte Carlo simulation
Input
Natural gas cost
increase
Oil cost
increase
Coal cost
increase
New ELC
technology
hurdle rate
Nuclear power
capacity
multiplier
Biomass
growth rate
bound
Geothermal
growth rate
bound
Hydropower
growth rate
bound
Landfill gas
growth rate
bound
Solar PV
growth rate
bound
Solar thermal
growth rate
bound
Wind growth
rate bound
IGCC
availability
factor
IGCC
maximum
capacity bound
Units
$M/PJ
$M/PJ
$M/PJ
None
None
% growth
per 5-yr
period
% growth
per 5-yr
period
% growth
per 5-yr
period
% growth
per 5-yr
period
% growth
per 5-yr
period
% growth
per 5-yr
period
% growth
per 5-yr
period
None
Gigawatts
Default
0.0
0.0
0.0
0.18
1.0
9.0
17.0
0.0
11.0
50.0
3.0
21.0
0.846
16.0
Low
-0.896
-0.712
-0.5
0.05
1.0
0.0
0.0
0.0
0.0
2.5
0.0
0.0
0.825
0.0
High
8.86
7.12
1.5
0.20
1.25
25.0
20.0
10.0
25.0
50.0
10.0
50.0
0.875
1000.0
Description
Cost added to the natural gas and
liquid natural gas supply curves
Cost added to the supply curves
imported and domestic fuel oil,
as well as to imported petroleum
fuels
Cost added to the coal supply
curves
Technology-specific discount
rate applied to all electric sector
technologies except renewables
Multiplier used to increase future
nuclear capacity
Growth rate on biomass use in
electricity generation
Growth rate on geothermal
power for electricity generation
Growth rate on hydropower for
electricity generation
Growth rate for the combustion
of landfill gas from municipal
solid waste landfills for
electricity generation
Growth rate for solar
photovoltaics in electricity
generation
Growth rate for solar thermal
technologies in electricity
generation
Growth rate for wind turbines in
electricity generation
Fraction of time during which
IGCC is operational
Peak capacity for IGCC
technologies in 2030
24
-------
One thousand Monte Carlo realizations were performed and the inputs and outputs tabulated.
Correlation analysis and normalized multiple linear regression were then applied.
3.4 Global Sensitivity Results - Correlation Analysis
Correlation coefficients were calculated for each combination of model inputs and outputs for
the year 2030. The resulting coefficients are provided in Tables 3.2 through 3.5. In each table,
correlation coefficients with absolute values of less than 0.1 are not shown, while those with
absolute values greater than 0.7 are emphasized via bold type and light shading.
Table 3.2: Correlation coefficients between model inputs and electric sector outputs
Inputs
Outputs
Electricity Generated by
Various Fuels
Pulverized Coal
Gasified Coal
Oil
Natural Gas
Nuclear Power
Geothermal
Biofuels
Landfill Gas
Solar Power
Wind Power
Use of NOX Controls on Coal-
Fired Power Plants
Peak Electricity Price
Emissions from the Electric
Sector
C02
NOX
SO2
to
o
0
(/)
O ^ to
ro o O
3 ° To
ra = o
-z. O o
-0.42 -0.14
0.3 -0.67
-0.55 0.49
0.19
0.19
0.12
0.24
0.64
-0.44 -0.45
"ro
a:
to
^
I
D)
O
O
c
O
_i
LU
3
^
-0.29
-0.12
-0.13
1
-0.33
0) dJ 0) ni
"ro "S "ro tt::
rv Ct rv fp r-
oj D- LL m jy ^
^ £ ^ £ ^ S
.c o 2 ° £ CD Jj
30 0 o | ra ^
O — i- ,n " C £
o 2 * % ® * %
c > CD >a ^ o
— OJ Q. ^ CL \— O
14— "o "O ^ ^ ^ c
.2 - ra o o •>
m O x _i w w S
-0.26 -0.25
-0.11
-0.12
1
0.86
0.36 0.33
0.76
-0.21 -0.26
.^
1
.£• — '
5 1
< 0
0 0
o o
0 0
Table 3.2 shows the correlation coefficients between each input and tracked electric sector
output. Only five of the input-output pairs have correlations greater than 0.7. The lack of a
larger number of strong linear relationships is indicative of the complexity of the system, which
allows for discrete behavior associated with selecting technologies and fuel switching.
25
-------
Many of the correlations that are shown in the table verify anticipated behavior. For example,
the new electricity generation technology hurdle rate and peak electricity price had a correlation
of 0.76, indicating that inhibiting uptake of new technologies for electricity generation resulted
in an increase in energy prices. Similarly, although the correlation is only 0.56, there appeared to
be a relationship between the hurdle rate and the use of controls on NOx emissions from coal-
fired power plants: as it became more difficult for new, cleaner electric generation technologies
to be adopted, reliance on retrofits such as selective catalytic reduction increased.
While it is not a strong correlation, there does appear to be an inverse relationship between the
cost of natural gas and the use of coal. This relationship can be attributed to the effect of the
limit on electric sector NOx emissions. Increased natural gas costs led to decreased use of
natural gas in electricity generation. Because natural gas-fueled technologies have lower NOX
emission rates than coal-fueled technologies, however, natural gas could only be substituted for
by coal if increased NOx controls were used on coal-fired boilers. The model instead, based upon
a comparison of the relative costs, opted to fill this gap with oil-fueled technologies. While oil
has fewer NOX emissions than coal, its emissions are greater than natural gas. Reductions in
coal use were required to compensate. These relationships are evident in the results of the
normalized multiple linear regression, shown in Figures 3.1 through 3.9.
Table 3.3 shows correlation coefficients between inputs and system-wide outputs. Only eight of
these coefficients exceeded 0.7. Many of the coefficients again confirmed expected behavior.
For example, natural gas cost increases resulted in decreased system-wide use of natural gas,
compensated for by increased use of renewables and petroleum.
An interesting result was that oil costs were much more highly correlated with NOx and SO2
emissions than natural gas or coal costs were. This is because the primary use of natural gas in
the model is in electricity generation, but the representation of the Clean Air Act emission
constraints places limits on the emissions of NOx and SO2 from that sector. In contrast, the
transportation sector is the primary user of oil, and this sector is not subject to NOx and SO2
emissions constraints.
The lack of strong correlations for some pairs was also insightful. For example, over the growth
rate constraints evaluated, market penetrations of various renewables did not appear to have had
a strong impact on system emissions or fuel use.
26
-------
Table 3.3: Correlation coefficients between model inputs and system-wide outputs
Inputs
Outputs
System- Wide Emissions
CO2
NOX
S02
Fuel Inputs to System
Renewables
Coal
Petroleum
Oil
Natural Gas
Liquid Natural Gas
Uranium
Net Imports
to
O
O
(/)
O to
E 8 5
13 c 5 —
ii w ra
ro = o
Z O 0
-0.29 -0.59
0.14 -0.82
0.3 -0.74
0.84 0.19
-0.5 0.28
0.87
-0.12 -0.79
-0.91
-0.82
0.68 -0.5
a>
to
a:
,
D)
O
O
c
.c
o
O
_i
LU
^
0.5
-0.16
-0.16
0.11
0.27
0.16
.£•
o
ra
CL
ro
0
ra
GJ
O
3
^
-0.29
-0.14
1
.22
0) ji <1) .S
to « to Q-
m ce ce ce .£ £
•^ *- .c r- ro ^
2 | 1 | | 1 JB
.c o 2 E £ (!) /S
"S 5 0 0 | ro _£.
O — ^ (/) i- C -i_i
»- (D OJ m fl E ^
0 E 3 5 ^ oi I
-^ oJ §. = CL K O
3 ^ 2 ~n *~ ^^ T3
•t; O -o c — — c
.2 <" >- m O O 3
03 CD X _J W W >
-0.17 -0.24
-0.11
0.14 0.14 0.18
-0.17 -0.12
±±
E
= 3r-
'ra Q.
< 0
O O
O O
o o
Table 3.4 shows correlation coefficients between electricity sector outputs and all tracked
outputs. Very few of the values in this table show a strong degree of correlation (values of 1.0
typically correspond to comparisons of the same output (e.g., pulverized coal vs. pulverized
coal). These correlation coefficients plus those repeated elsewhere in the table are shaded. The
coefficients in the table provide an indication that fuel-switching was occurring between natural
gas and oil in electricity generation.
System-wide NOx and SC>2 emissions showed a high degree of correlation with the use of oil in
electricity generation. Increased system-wide emissions were not a result of the electric sector,
however, since its NOx and SC>2 emissions are constrained. Instead, they were a result of fuel and
technology switching within the transportation sector in response to drivers of favorable oil
costs.
Net imports of fossil fuels appeared to be correlated with oil use in electricity generation, but
inversely correlated with natural gas use. These correlations indicate that increased demands for
oil resulted in increased fossil fuel imports. At the quantities modeled, most of the natural gas
was supplied domestically, and thus had a negative correlation with imports.
27
-------
Table 3.4: Correlation coefficients between electric sector and all outputs
Electric Sector Outputs
All Tracked Outputs
Electricity Generated by Various Fuels
CO i_
O CD
0 en 3 - -
S J £ | | g
•g E S 1 « ? £
> 3 "o 'o 5 co ^
=! = CO=!CD.00g
Q-OZZOCOC/JS
Electricity Generated by Various Fuels
Pulverized Coal 1 1 .(X^M6 0.12-0.29 -0.24
Oil
Natural Gas
Nuclear Power
Geothermal
Biofuels
Solar Power
Wind Power
1.00 -0.65 -0.12 -0.21
1.00 -0.13
1.00
1.00
1.00
1.00 0.18
1.00
Use of NOX Controls on Coal-Fired
Power Plants
Peak Electricity Price
CO2 from Electric Sector
System- Wide Emissions
C02
NOX
S02
Fuel Inputs to System
Renewables
Coal
Petroleum
Oil
NGA
NGL
Uranium
Net Imports
0.41 0.51 -0.27 -0.29 -0.32 -0.42 -0.29
0.85 -0.60 -0.24
0.88 -0.57 -0.23 0.16
-0.51 -0.34 0.15 0.21 0.29 0.26
0.76 -0.62 0.36 -0.14 -0.19
-0.32 0.40 -0.54 0.12
0.65 -0.47 -0.15
0.32 -0.36 0.55 -0.35 -0.20
0.27 -0.29 0.55 -0.13
-0.29 -0.12 -0.13 1.00
-0.23 0.73 -0.74 -0.12
c-ffi o
0 £ °
(/) CO CD CD
0°- £ ^
'c CD °- •;=
00 o a
O ~& t3 r-
^ CD CD E
O _i_ _^ "*"
Si o CD O
3 O CL O
-0.38 0.55
0.19 0.38
-0.48
-0.33
-0.33
0.61 -0.35 -0.39
0.17 -0.17 -0.29
1.00 -0.38 -0.55
1.00
1.00
-0.57 0.20 0.96
-0.11 0.13 0.46
0.13 0.27
0.34 0.37 -0.67
-0.18 -0.22 0.33
0.17 0.64 -0.25
-0.13 -0.11 0.47
-0.47 -0.35 0.51
-0.19 -0.55 0.31
-0.33
0.53
28
-------
Table 3.5 shows correlation coefficients between combinations of system-wide outputs. Among
the notable relationships are the inverse correlations between system-wide natural gas use and
both renewables and petroleum use.
Table 3.5: Correlation coefficients between system outputs and all outputs
System-Wide Output
System-Wide Output
System- Wide Emissions
CO2
NOX
SO2
Fuel Inputs to System
Renewables
Coal
Petroleum
Oil
Natural Gas
Natural Gas Liquids
Uranium
Net Imports
Emissions
CM X CM
000
o z w
1 .00 0.63 0.40
1.00 0.89
1.00
Fuel Inputs to System
CO
;o
!^
^ ,- ra ra
.a £ O O
ro ^ £
3 ^ ra ra =!
CD _ O i= i= 'F
c ra i: 335
o> o a) = ro ro ii:
CC O CL O Z Z Z)
-0.56 0.16 -0.11 0.60 0.38 0.16 -0.29
-0.49 0.25 0.80 -0.20 -0.22
0.14 -0.63 0.40 0.63 -0.42 -0.30
1.00 -0.50 0.66 -0.22 -0.87 -0.66
1 .00 | -0.48 -0.22 0.51 0.37 -0.14
1.00^^3 -0.77 -0.91
1.00
1.00^(^5
1.00
1.00
CO
•c
o
CL
'<£
z
0.29
0.69
0.70
0.41
-0.55
0.82
0.45
-0.62
-0.83
1.00
3.5 Global Sensitivity Results - Normalized Linear Regression
To evaluate the relative magnitude of impact (as opposed to correlation) of each input on the
tracked outputs, normalized linear regressions were used. Normalization was carried out so that
the coefficients on each term would be directly comparable. Regression results are depicted
graphically in the tornado diagrams provide in Figures 3.1 through 3.9. Tornado diagrams were
used to present the regression results since these diagrams visually convey the sign and relative
magnitude of influence of each input. R2 values are provided to give an indication of the quality
of the fit, indicating the ratio of explained variance to total variance.
29
-------
In Figure 3.1, the factor with the greatest influence on the quantity of coal used was natural gas
cost. As seen in Table 3.2, increased natural gas costs led to decreased coal use. The correlation
between coal and nuclear power in electricity generation is much more straightforward since
nuclear power would offset coal in supplying base load electricity. In comparison, the
correlations between nuclear and natural gas and between nuclear and oil were much weaker.
Natural Gas Cost I -0.239
Nuclear Capacity Multiplier I -0.169
Biofuels Growth Rate I „„,.,.[
r, , I -0.1 4D
Bound i I
Hydro Growth Rate Bound I -0.138^
New Electricity Generation | 01271
Technology Hurdle Rate ! I
Oil Cost | -0.059 |
Geothermal Growth Rate i nnicf
_ . i -U.Uob
Bound i I
Wind Growth Rate Bound I R2=0.67 -0.035
-0.3 -0.25 -0.2 -0.15 -0.1 -0.05
Normalized Regression Coefficient
Figure 3.1: Factors affecting use of coal in electricity generation
30
-------
Figure 3.2 provides an indication of the relative sensitivities of natural gas use in electricity
generation to the Monte Carlo inputs. As might be expected, natural gas use decreased with
increases in natural gas costs, but increased with higher oil costs. The impact of the hurdle rate,
which inhibited the introduction of new, more efficient natural gas technologies, was only about
a third as important as either of these two cost factors.
Natural Gas Cost
Oil Cost
New Electricity Generation
Technology Hurdle Rate
Nuclear Capacity Multiplier
Biofuels Growth Rate
Bound
Hydro Growth Rate Bound
Wind Growth Rate Bound
-0.231
-0.073
-0.053
-0.047
-0.036
-0.012 F
0.213
-0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
Normalized Regression Coefficient
Figure 3.2: Factors affecting use of natural gas in electricity generation
-------
Figure 3.3 shows the sensitivities of oil use in electricity generation to the various Monte Carlo
inputs. Oil cost increases were the major factor affecting penetration. Natural gas cost increases
had approximately half the impact. The influence of nuclear capacity is less than one half that of
natural gas cost.
Oil Cost
Natural Gas Cost
Nuclear Capacity Multiplier
Wind Growth Rate Bound
Hydro Growth Rate Bound
-0.555
-0.106
-0.040
-0.031
0.259
-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1
Normalized Regression Coefficient
0.2
0.3
Figure 3.3: Factors affecting use of oil in electricity generation
Figure 3.4 indicates that CC>2 emissions from the electric sector were impacted the greatest by
increases in natural gas and oil costs. For natural gas, this relationship represents the same
behavior observed in Table 3.2: natural gas cost increases led to decreased penetration of natural
gas-fueled electricity generation technologies. These technologies had fewer NOX emissions than
pulverized coal plants, and, as a result, the amount of coal used in electricity generation
decreased to meet NOx constraints. Because coal has higher CC>2 emissions than the other fuels,
this decrease in coal use resulted in a decrease in CC>2 emissions as well. Similar behavior
appeared to be occurring with increased oil costs.
The impact of the hurdle rate was only slightly lower than the cost of natural gas or oil. By
making it more difficult for new, more efficient technologies to penetrate the electricity
generation market, the hurdle rate led to increases in CO2 emissions.
32
-------
Natural Gas Cost
Oil Cost
New Electricity Generation
Technology Hurdle Rate
Nuclear Capacity Multiplier
Hydro Growth Rate Bound
Biofuels Growth Rate
Bound
Wind Growth Rate Bound
Geothermal Growth Rate
Bound
-0.275
-0.266
-0.192
-0.128
-0.126
-0.051
-0.020
10.256
R2=0.88
-0.4 -0.3 -0.2 -0.1 0 0.1
Normalized Regression Coefficient
0.2 0.3
Figure 3.4: Factors affecting COi emissions from electricity generation
Another expected result was that increased nuclear capacity led to decreased CC>2 emissions.
This influence was not as strong as that of natural gas costs, oil costs, and hurdle rate, however,
even though nuclear power has no CC>2 emissions. Again, the results were influenced by the NOx
limit on electricity generation. Increased use of nuclear power to meet base load electricity
generation effectively freed space under the NOx limit. The result was that fewer NOx controls
were placed on coal-fired power plant emissions and additional coal capacity was added, at the
expense of natural gas and oil technologies.
The new technology hurdle rate had the greatest impact on the adoption of NOx retrofit controls,
such as selective catalytic reduction, at coal-fired power plants (Figure 3.5). Increases in the
hurdle rate effectively required that additional existing coal-fired power plants remain online in
2030. To achieve electric sector NOx constraints with these plants, greater use of NOx controls
was required.
33
-------
New Electricity Generation
Technology Hurdle Rate
Natural Gas Cost
Oil Cost
Hydro Growth Rate Bound
Nuclear Multiplier
IGCC Availability Factor
-1.220
0.432
0.138
~~| 0.060
3.057
GO,
0.044
R2=0.82
-1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
Normalized Regression Coefficient
Figure 3.5: Factors affecting use of NOx controls on coal-fired power plants
Figure 3.6 illustrates that increasing oil costs had the effect of reducing system-wide CC>2
emissions. Contributing factors were fuel and technology switching in the transportation sector.
Additionally, fuel-switching within the electric generation sector may have played a role. By
favoring existing coal-fueled electricity generation technologies, the hurdle rate had the effect of
increasing system-wide carbon emissions.
34
-------
Oil Cost
New Electricity Generation
Technology Hurdle Rate
Natural Gas Cost
Nuclear Capacity Multiplier
Hydro Growth Rate Bound
Biofuels Growth Rate
Bound
Wind Growth Rate Bound
Geothermal Growth Rate
Bound
-0.359
-0.181
-0.167
-0.106
-0.103
-0.041
-0.017
0.311
R2=0.67
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Normalized Regression Coefficient
Figure 3.6: Factors affecting system-wide COi emissions
The regression coefficients shown in Figure 3.7 indicate that SC>2 emissions were most sensitive
to oil and natural gas costs. While 862 emissions from the electric sector were constrained,
changes in SC>2 emissions were occurring in other sectors, primarily in the transportation sector.
Decreased oil costs led to increased use of gasoline and diesel within the transportation sector.
Diesel, in particular, has the potential to increase SO2 emissions, although this impact will be
reduced when new regulations on sulfur content in diesel fuels come into effect. These
regulations are not currently represented in the MARKAL model.
The influence of increased natural gas costs was approximately half that of oil costs. One of the
reasons that increased natural gas costs resulted in increased 862 emissions was because of
switching from natural gas-fueled transportation technologies to oil-fueled technologies.
Changes in SC>2 emissions may also have been influenced by resulting technology changes
within the electricity sector, and resulting interactions with other sectors, although additional
analyses would need to be carried out to characterize these impacts.
35
-------
Oil Cost
Natural Gas Cost
New Electricity Generation
Technology Hurdle Rate
Nuclear Multiplier
Hydro Growth Rate Bound
-0.537
-0.119
-0.085
-0.0284
-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1
Normalized Regression Coefficient
0.243
0.2 0.3
Figure 3.7: Factors affecting system-wide SOi emissions
36
-------
Figure 3.8 presents the relative impacts of various inputs on system-wide NOx emissions. As
with SC>2, inexpensive oil inhibited the transition to more efficient vehicle technologies. The
result was an increase in NOx emissions. The impact of other drivers was relatively small.
Oil Cost
Natural Gas Cost
Nuclear Capacity Multiplier
New Electricity Generation
Technology Hurdle Rate
-0.633
-0.067
0.13
0.066
-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
Normalized Regression Coefficient
0.1
0.2
Figure 3.8: Factors affecting system-wide NOx emissions
37
-------
In MARKAL, reductions in fossil fuel imports can be achieved through either switching to more
efficient technologies or switching to fuels that are supplied domestically instead of imported.
Fuels that are largely supplied domestically in the model include coal, uranium, natural gas, and
renewables. Oil and other petroleum products, in contrast, are largely imported fuels. Thus, in
Figure 3.9, the two factors with the highest impact on imports were natural gas costs and oil
costs, which had opposite effects on imports
Natural Gas Cost
Oil Cost
New Electricity Generation
Technology Hurdle Rate
Nuclear Capacity Multiplier
Wind Growth Rate Bound
Hydro Growth Rate Bound
Geothermal Growth Rate
Bound
-0.446
-0.071
-0.023 f
-0.022 [
0.567
0.109
0.02
-0.6 -0.4 -0.2 0 0.2 0.4
Normalized Regression Coefficient
0.6
0.8
Figure 3.9: Factors affecting fossil fuel imports
3.6 Parametric Sensitivity Analysis
From the global sensitivity analysis, it was evident that many of the key outputs were most
responsive to changes in natural gas and oil costs, nuclear capacity, and the hurdle rate for new
electricity generation technologies. There were also cross-sector effects between the electricity
generation and transportation sectors that had an impact on technology adoption, fuel use, and
emissions.
A parametric sensitivity analysis was carried out to evaluate the effect of incremental changes in
each of these inputs on the primary outputs of interest: fuel use within the electricity generation
sector, system-wide emissions, and net imports. By evaluating changes parametrically,
sensitivities around the baseline run were characterized. Results of the parametric runs are
provided in Figures 3.10 through 3.16.
38
-------
Figure 3.10 shows the change in electricity generation by fuel type as natural gas cost changed in
relation to the baseline natural gas price in 2030. In this figure, as well as in Figures 3.11
through 3.13, electricity generation categories with outputs of less than 500 PJ are not shown.
Decreases in natural gas costs led to an increase in natural gas use in electricity generation, at the
expense of coal and oil. The range of decreased natural gas costs that was examined was not
sufficient to drive compressed natural gas vehicles to penetrate the light-duty vehicle market, so
there was limited interaction with the transportation sector.
Increasing natural gas costs by 35% or more appeared to drive oil use in electricity generation
while at the same time resulting in decreased use of coal. Natural gas use decreased with cost,
although the decrease was comparatively modest. These observations suggest that the increased
natural gas cost may have triggered an emissions-related modeling tipping point that favors oil
use over coal.
9000
7000
Pulverized Coal
£
01
o
01
c
Ol
O
o
o
m
ee-
Natural Gas
o o
5000
Nuclear Power
2000
O 0 I 0
Oil
Hydropower
-40% -20%
0% 20% 40% 60% 80% 100% 120% 140%
Change in Natural Gas Cost
Figure 3.10: Impact of natural gas cost on electricity generation by fuel type.
39
-------
Figure 3.11 shows that a decrease in the oil costs resulted in additional oil use in electricity
generation, accompanied by decreases in natural gas and coal use. Coal use decreased to offset
the loss of the low NOx-producing natural gas technologies. Increased oil prices led to decreases
in oil and coal use, with the corresponding demand met instead by natural gas. Coal decreases in
this case were likely adopted to offset the loss of low-NOx-producing oil technologies. Changes
in coal use corroborate the hypothesized effect of the emissions constraint suggested in the
discussion of Table 3.3 and Figure 3.4.
9000
CL^
Ol
o
13
01
O
7000
5000
2000
o10S° i
Pulverized Coal
Natural Gas
Nuclear Power
Hydropower
-100%
-50%
50% 100%
Change in Oil Cost
150%
Oil
—^*
200%
250%
Figure 3.11: Impact of oil cost on electricity generation by fuel type.
Figure 3.12 shows the response of the electric sector to increased nuclear capacity. The primary
fuel offset was coal, with only minor changes in natural gas use. Interestingly, the total amount
of electricity generated, the sum across fuel types in the future, appears to increase with
additional nuclear capacity. An examination of residential energy demands suggests that
additional nuclear capacity resulted in fuel switching within that sector from natural gas and oil
to electricity (e.g., to meet demands for home heating). Similar fuel switching may have
occurred in the commercial and industrial sectors.
40
-------
9000
1.15 1.2 1.25 1.3 1.35 1.4
Nuclear Capacity Multiplier
Figure 3.12: Impact of nuclear capacity on electricity generation by fuel type.
Increasing the hurdle rate from 0.05 to 0.25 had the effect of decreasing the uptake of new
electricity generation technologies as shown in Figure 3.13. This decrease forces the cost of
generating electricity to increase. The net result was a decrease in the use of oil, coal, and natural
gas in electricity generation, as well as an overall decrease in the total amount of electricity
generated. In response, MARKAL results suggested that the residential sector, commercial, and
industrial sectors would experience some degree of fuel switching from electricity to other fuels
in meeting demands such as space heating.
41
-------
10000
9000
^ 8000
Q_
g; 7000
01
3
6000
o 5000
S
ai
c
ai
O
1
m
4000
3000
2000
1000
0
Pulverized Coal
Natural Gas
Nuclear Power
Hydropower
Oil
0.05 0.1 0.15 0.2 0.25
Hurdle Rate for New Electricity Generation Technologies
0.3
Figure 3.13: Impact of the new electricity generation technology hurdle rate on electricity
generation by fuel type.
Figure 3.14 shows percent changes in CC>2 emissions from the electric sector in response to each
of the four inputs. This is a different type of parametric sensitivity graph that shows the
responses to parametric changes in more than one input on an output.
The results suggest that nuclear power capacity had the most direct impact on CO2 emissions
from the electric sector. A 35% increase in nuclear capacity yielded less than a 6% reduction in
CO2, however. Oil cost had a similar impact with respect to magnitude of CO2 response.
The overall magnitude of changes in electric sector CC>2 emissions is less than might be
expected. One of the primary reasons for the moderate level of these changes is that the
renewable technologies with negligible CO2 emissions are not sufficiently economically
competitive that they are able to offset large amounts of fossil fuel usage. Instead, the primary
response of the electric generation system to changing inputs is fuel switching from one fossil
fuel to another. This has some implications on CO2 emissions, as reflected in the small changes
observed.
42
-------
2.0%
m -1CD%
p
O
o
_c
01
O)
c
ra
^
O
100% 150% 200% 250%
Natural Gas Cost
- Oil Cost
ELC Hurdle Rate
"~ Nuclear Capacity
-6.0%
Change in Input
Figure 3.14: Changes in COi emissions from the electric sector in response to changes in
inputs.
In Figure 3.15, the system-wide impacts of parametric changes to each input on CC>2 emissions
are shown. The functions observed are similar to those in Figure 3.14, although the overall
magnitude of CO2 changes is less, ranging from approximately +1% to -2.75%.
Sensitivity diagrams for electric sector NOx and SC>2 emissions are not shown here. The electric
sector NOx and SC>2 emissions limits were binding in all runs. System-wide impacts were
therefore largely dependent on the transportation sector.
43
-------
2.0%
in
O
'in
in
c -1CD%
LJJ
CM
O
O
01
;c
3i
Ol
u
OT
Ol
O)
ra
.n
O
100% 150% 200% 250%
Natural Gas Cost
- Oil Cost
ELC Hurdle Rate
"~ Nuclear Capacity
-6.0%
Change in Input
Figure 3.15: Changes in system-wide COi emissions in response to changes in inputs.
Figure 3.16 shows the changes in net imports. Increased natural gas costs had the greatest
impact on increasing reliance on imports, with a 100% increase in natural gas costs resulting in
approximately a 17% increase in imports. The impact of oil costs on net imports was
considerably less, reaching a maximum decrease in imports of about 5%, corresponding to a
150% increase in oil costs.
44
-------
20.0%
•c
o
Q.
E
in
in
O
01
O)
c
ra
^
O
-1CO%
Natural Gas Cost
- Oil Cost
ELC Hurdle Rate
Nuclear Capacity
25|)%
-10.0%
Change in Input
Figure 3.16: Changes in net imports in response to changes in inputs.
3.7 Summary of Observations from Sensitivity Analysis
The combination of correlation coefficients, normalized multiple linear regression, and
parametric analysis of the BAU case provides considerable insight into the inner-workings of the
MARKAL model and the response of the model to alternative input assumptions. In most cases,
the results of these analyses confirmed expected behavior in the model. Additionally, the results
provided insight into the complicated response of the system to criteria pollutant emission limits.
In this subsection, we summarize many of the key observations from the sensitivity analysis.
Much of the EPANMD electric sector's behavior appears to be influenced by whether specific
technologies and fuels meet base or peak load electricity demands. The predominant base load
technologies are coal-fired power plants and nuclear power plants. Coal is the most competitive,
and has the largest market share.
Natural gas- and oil-fueled technologies are used to meet peak electricity demands. Both fuels
experience some cross-sector interactions with the transportation sector because natural gas and
refined oil products can be used within vehicles. While there was some evidence of cross-sector
45
-------
interactions, these were largely of secondary importance to in-sector technology and fuel
competition.
The electric sector emissions constraints have interesting effects on the electric generation mix.
For example, when natural gas became more expensive and gas technology utilization decreased,
use of coal also decreased. This behavior, which may at first seem counter-intuitive, is explained
by the response of the system to the electric sector NOx constraint. Since coal technologies had
higher NOX emissions than natural gas technologies, and since NOX constraints on electric
generation were binding, the model opted to replace natural gas with oil. Oil combustion also
leads to greater NOx emissions than that from natural gas, thought it is less than coal. Some coal
was therefore displaced by oil. This fuel-switching and related technology change had
implications on CO2 emissions as well.
The electric sector NOx constraint also affects the model's response to increased nuclear
capacity. By introducing electricity generation capacity that does not have NOx emissions, coal-
fired power plant emissions are less constrained in meeting the electric sector NOX constraint. In
response, the fraction of coal-fired plants projected to make use of NOx controls, such as
selective catalytic reduction, decreases. The same behavior can be expected from increased
electricity generation from low- or zero-NOx renewables, such as solar or hydropower.
The electricity generation hurdle rate has an additional impact on future-year energy sector
technologies. Increasing this rate effectively makes it more difficult for new, more efficient
technologies to penetrate the electricity generation market. This difficulty, in turn, has the effect
of increasing the marginal peak electricity price, increasing CO2 emissions from electricity
generation, and decreasing the penetration of renewables. Thus, addressing hesitancy to adopt
new technologies through some approaches for hedging risk may yield a more efficient
electricity generation system.
Imports of fossil fuels are correlated with the use of oil in electricity generation, but inversely
correlated with natural gas use, reflecting that much of the natural gas demand is being met by
domestic supplies in the model. Cross-sector fuel switching resulting from changes in natural
gas and oil consumption in the transportation sector may also have an impact, but further
analysis is needed to characterize this behavior.
Finally, changes in system-wide CO2 emissions in response to variation in model inputs were
minor, with decreases of less than 3% observed. This output is influenced by the inability of low-
CO2 emitting technologies, and in particular, renewables, to achieve high market penetrations.
When these technologies do penetrate, the potential reductions in CO2 emissions are often offset
by increased use of coal and other fossil fuels, made possible via the room under the NOx limit
created by the renewables.
Sensitivity analysis such as is presented here has been a useful component of the MARKAL
model database development and quality assurance. By identifying key drivers and interactions,
sensitivity analysis facilitates an understanding of how the model responds to alternative input
assumptions which, in turn, aids model refinements.
46
-------
Section 4
The Future Role of Nuclear Energy in the U.S.
4.1 Introduction
With all nations facing enormous challenges related to energy security, sustainability and
environmental quality, nuclear power will likely play an increasingly important role in the future.
In particular, the life-cycle emissions of criteria pollutants and greenhouse gases from nuclear
power plants are significantly lower than from conventional fossil-fueled plants, renewing
interest in nuclear power as a low emissions source of electric power. In order for nuclear power
to emerge as a key future technology, four basic challenges must be met: cost, safely,
proliferation prevention, and waste management (Ansolabehere et al., 2003). While
acknowledging significant challenges related to the latter three items, this analysis focuses on an
engineering-economic assessment of future nuclear power in the U.S.
This section focuses on the potential role of nuclear power in the U.S. electric sector over the
next 30 years by analyzing results from the U.S. EPA National MARKAL Database model
(EPANMD). The section first describes the implementation of conventional and advanced
nuclear technologies in the EPANMD. In the next subsection, modeling results are presented and
analyzed. Finally, implications of the penetration of nuclear technologies on the emissions from
the power sector are drawn.
4.2 Nuclear Technology Representation in MARKAL
In the EPANMD, all nuclear technologies draw on a single uranium supply curve. The uranium
supply curve is based on estimates of global uranium reserves and the cost of extraction
(OECD/IAEA, 2002). Because the energy density (energy per unit weight) of uranium is high,
transport costs were ignored.
The nuclear fuel cycles included in EPANMD were determined by careful consideration of the
nuclear technologies most likely to be deployed in the U.S. over the next three decades. The
analysis considers the following technologies: light water reactors operating on a once-through
fuel cycle, mixed-oxide reactors, heavy water reactors, fast breeder reactors, and high
temperature gas-cooled reactors. Although not a comprehensive list of nuclear technologies, they
represent the broad technical thrusts in the nuclear industry.
In an increasingly competitive world of deregulated electricity generation markets, other things
being equal, systems with lower upfront costs and shorter construction times are likely to be
preferred by investors over those with higher upfront costs and longer construction times. Light
water reactors (LWRs) operating on a once-through fuel cycle (no reprocessing) currently have
the lowest cost among commercially available reactors. LWRs have been widely adopted
globally and still currently serve roughly 20 percent of U.S. electricity demand.
Heavy water reactor technology typically calls for larger plants with higher construction and
capital costs, as compared to light water reactor plants. Moreover, the large amount of heavy
water (deuterium) required to run these plants also necessitates significant infrastructure
investments. A heavy water reactor's key advantage is its ability to use natural uranium, as
47
-------
compared to enriched uranium required by light water reactors. However, the demand-supply
equilibrium is such that most analysts expect enough enriched uranium to be available at
reasonable price for at least the next half-century. As a result, heavy water reactors are not
included in the EPANMD.
Enough supplies of uranium exist to build and operate roughly 1000 reactors in a once-through
LWR nuclear fuel cycle. If new uranium resources do not become available and existing
resources are depleted at a rapid pace, then breeder reactors may emerge as a viable option to
meet long-term energy supply goals. Breeder reactors not only fission uranium, but also convert
'7^8 O'lO 91Q '711
fertile materials (primarily U and Th ) into fissile products (primarily Pu and U ).
Breeder reactors are designed to produce more fissile material than they fission. Breeder reactors
are not an economically attractive option in the wake of prevailing enriched uranium prices, at
least in the short-term. Therefore, we did not include breeder reactors in the EPANMD.
In the U.S., LWRs are likely to remain the dominant nuclear technology because there is
significant experience with design, construction and operation of these plants. However, MOX
(mixed oxide) reactors were also included in the model, since it is at least plausible that
plutonium recycling would be considered in the future, despite the high costs and risks of
proliferation. MOX fuel is created by first extracting the fissionable uranium and plutonium from
the spent LWR fuel in a process known as PUREX (plutonium and uranium extraction). The
resultant fuel must be blended with depleted uranium (a byproduct of the uranium enrichment
process) to obtain the correct proportion of fissionable material in the fuel.
Costs and performance characteristics for LWRs were drawn from Ansolabehere et al. (2003)
and DOE (2001). It should be noted that the Ansolabehere et al. (2003) cost estimates for MOX
recycling are significantly higher than European estimates (Ansolabehere et al., 2003) and should
be interpreted as conservative. Table 4.1 below shows the cost associated with the light water
reactor fuel cycle used in EPANMD.
Table 4.1: Cost and performance estimates for LWRs used in EPANMD.
Capital Cost* ($/kW)
O&M Cost ($/kWh)
Burnup (MWd/kg IHM)
Capacity Factor (%)
Efficiency (%)
Fuel Cost ($/kWh)
Lifetime (years)
Average Cost ($/kWh)
Data Source
Existing LWRs
N/A
0.0125
50
85
33
0.0051
40
0.018
NEI, 2003
LWRs in 2010
1440
0.005
50
90
36
0.0051
40
0.036
DOE, 2001
MOX Reactors
2000
0.0077
40
85
33
0.022
40
0.070
Ansolabehere et al., 2003
*The levelized costs for new plants are calculated with a 15% discount rate, representing a
private investor's expected rate of return.
48
-------
Figure 4.1 presents a schematic diagram of the LWR fuel cycle in the EPANMD. The
enrichment step includes the cost for uranium ore purchase, conversion, enrichment, fabrication,
and storage and disposal of nuclear waste. Note that spent uranium can be reprocessed and
fissioned in a MOX (mixed oxide) reactor.
Mined Uranium
Natural U
Conversion, Enrichment
and Fabrication
Depleted U
Stockpile of
Depleted U
Enriched U
Nuclear power plant
Electricity
to grid
Spent U
Hi PUR
1 Pu
PUREX
waste
Stockpile of
PUREX Waste
MOX Fabrication
MOX reactor
Electricity
to grid
MOX
waste
Figure 4.1: Schematic diagram of the light water reactor (LWR) fuel cycle in the
EPANMD.
LWRs have several technical drawbacks that have limited their deployment, particularly in the
U.S. Construction of the reactor vessel and associated fuel handling equipment is materials-
intensive and requires large capital outlays. To defray the cost, LWRs exploit economies of
scale: LWR plants are typically 1 GW or larger and usually take 10 years to approve and build.
From a finance perspective, investors may be unwilling to bear the risk of ordering a plant that
will not go online for many years, by which time others may have added new capacity in the
same region. Although large nuclear plants provide relatively low per kWh cost, they have the
drawback of exceeding demand growth in smaller-demand networks.
In addition, the low thermal efficiency (approximately 33%) and burnup (50 GWd/TIHM) of
LWRs create a significant amount of radioactive waste that requires storage and disposal. The
ultimate disposition of spent uranium from LWRs presents a serious technical challenge and a
public policy concern. Finally, LWRs rely on active safety systems and human judgment and
intervention to prevent core meltdowns, which can fail, as evidenced by the accident at the Three
Mile Island nuclear facility outside of Harrisburg, Pennsylvania.
High temperature gas-cooled reactors (HTGRs) address many of the shortcomings of LWRs and
are therefore included in the EPANMD. HTGRs are built in modular units ranging from 100-300
MWe, making nuclear a feasible option in smaller markets. The smaller size and modular design
can also allay investor concerns about long construction times and the associated financial risk.
In addition, HTGRs have higher burnup and thermal efficiency, resulting in a proportional
reduction in spent fuel. Finally, HTGRs incorporate a passive safety design, such that heat
generated during fission can be thermally conducted to the ground without resulting in a core
meltdown.
49
-------
In order to dissipate heat passively through the ground, HTGRs must have a smaller core density
than LWRs, so multiple HTGR units are required to compete with a single LWR. At first glance,
it would seem that HTGRs would be prohibitively expensive because the design does not take
advantage of the same economies of scale as LWRs. HTGR designers are depending on a
different economic scaling law that will make HTGRs competitive with LWRs: factory
manufacturing of modules, shorter construction schedules, and sequential completion of units (1
year apart). Because HTGRs are currently in the demonstration stage, this economic strategy
remains unproven.
There are two main competing HTGR designs: the Pebble Bed Modular Reactor (PBMR) and the
Gas Turbine - Modular Helium Reactor (GT-MHR). A common feature of these two designs is
uranium oxide particles coated with pyrolytic carbon to contain the fission products. In the
PBMR design, the fuel particles are embedded in carbon spheres roughly 2.5 inches in diameter,
which allows the continuous removal and reloading of fuel spheres without shutting down the
reactor. Also, both the PBMR and GT-MHR use helium as a coolant and run modified
combustion turbines directly on the high temperature helium. A schematic diagram of the HTGR
fuel cycle without recycling is shown in Figure 4.2. Note that while in practice it is possible to
extract enriched uranium and plutonium from the spent fuel particles, it is prohibitively
expensive at present.
Mined Uranium
Natural U
Conversion, Enrichment
and Fabrication
Depleted U
Uranium fuel pellets
PBMR or GT-MHR
Sf
Spent pellet fuel
Electricity
to grid
Figure 4.2: Schematic diagram of the high temperature gas-cooled reactor (HTGR) fuel
cycle in the EPANMD.
Because there has been no commercial development of HTGRs, the cost estimates used in
EPANMD are speculative. The capital costs represent nth-of-a-kind estimates, which assume the
units are being mass produced. The speculative nature of the estimates and the nth-of-a-kind cost
assumptions make the characterizations of HTGRs in the EPANMD optimistic (see Table 4.2).
50
-------
Table 4.2: Cost and performance estimates used in EPANMD for HTGRs (advanced nuclear).
Capital Cost* ($/kW)
O&M Cost ($/kWh)
Burnup (MWd/kg IHM)
Capacity Factor (%)
Efficiency (%)
Fuel Cost ($/kWh)
Lifetime (years)
Average Cost ($/kWh)
Data Source
PBMR
1250
0.0025
80
95
40
0.005
60
0.03
DOE, 2001
GT-MHR
1122
0.0036
112
90
48
0.0077
60
0.033
DOE, 2001
* The levelized costs for new plants are calculated with a 15% discount rate, representing a
private investor's expected rate of return.
4.3 MARKAL Analysis
The EPANMD includes the following nuclear technologies: LWRs, MOX plants, PBMRs, and
GT-MHRs (the latter two being the high-temperature gas-cooled reactors). In the results that
follow, the LWR and MOX plants are grouped together and presented as "conventional nuclear"
and the PBMR and GT-MHR are grouped together and presented as "advanced nuclear".
Growth Rates
The choice of a nuclear growth rate constraint is a key assumption, as it prevents MARKAL
from building an unrealistic amount of nuclear capacity. Ideally, the growth rate constraint
would be based on the amount of capacity installed in the previous time period, but MARKAL
does not provide the capability to define a capacity-dependent growth rate constraint. Instead,
MARKAL allows the specification of growth rate constraints by time period only. The maximum
growth rate decreases in later time periods under the assumption that nuclear capacity will grow
at a slower pace as the nuclear capacity base grows larger. In addition, another growth parameter
was used to allow a small increment of capacity to be added above the upper bound growth
constraint. For conventional LWRs, the incremental capacity allowance is 4 GW-representing
roughly four new plants-and the capacity allowance is 3 GW for advanced nuclear-representing
roughly 15 new units. This incremental capacity provides an upper bound on the nuclear capacity
the first year it enters the model, and allows the 3 - 4 GW addition over and above the specified
growth rate in all subsequent periods. Growth constraint specifications are shown in Table 4.3.
Conventional nuclear can first enter in 2010, and 4 GW max can be built. Note that these growth
rates apply to each individual technology; there are two advanced and two conventional nuclear
technologies.
Table 4.3: Annual maximum growth rates (%) by model time period and incremental capacity
(GW) allowable over the growth rate.
Incremental GW 2015 2020 2025 2030
Conventional 4 25 25 18 12
Advanced 3 0 25 25 18
51
-------
While the choice of future growth constraints is somewhat arbitrary, the growth constraints
shown in Table 4.3 are plausible. From the inception of Eisenhower's Atoms for Peace Program
in the early 1950s to the mid-1980s when the last U.S. nuclear units came online, roughly 100
GW of nuclear capacity was built (EIA, 2003). This benchmark indicates the U.S. built 100 GW
over 30 years, or roughly 3 GW/yr assuming linear capacity additions. With the upper bound
growth constraints in Table 4.3, the maximum conventional nuclear capacity would be
approximately 130 GW by 2030. Likewise, the maximum advanced nuclear capacity would be
approximately 80 GW by 2030. Though this upper bound on growth allows for the
unprecedented expansion of nuclear, a mature nuclear industry and improved technology make
such growth feasible.
Model Results
Electricity production is shown in Figure 4.3 when EPANMD is run with both conventional and
advanced nuclear technologies. The results include 16 GW of new conventional nuclear and 50
GW of new advanced nuclear.
25000
2025
2030
• Advanced Nuclear
ffl New Pulverized Coal
D Other
H Renewables
D Natural Gas
• Conventional Nuclear
• Residual Coal
Figure 4.3: Electricity generation in a baseline run of EPANMD with both conventional
and advanced nuclear technologies available.
The new nuclear capacity did not reach the upper bound growth constraints shown in Table 4.3.
In 2030, nuclear power accounts for 17.4% of the electricity generated, which maintains the
current amount of nuclear power on a relative share basis. Much of the growth in electricity
demand over the model time horizon was met by new combined-cycle natural gas (248.8 GW).
Because fossil fuel prices are a key future determinant of electric sector technology, parametric
analysis of natural gas and coal prices was performed to observe the impact on nuclear power
installations. Figures 4.4 and 4.5 show the change in the nuclear share as coal and gas prices are
increased, respectively. Doubling the coal price (+2 $/GJ) only increases the nuclear share by
approximately 1% (Figure 4.4). While doubling the natural gas price (+5 $/GJ) increases the
share of nuclear by approximately 5% (Figure 4.5).
52
-------
c o^/o -
g
TO
5 28% -
c
0)
O O/IO/
X 24% -
~
o
£ 20% -
o
0)
LLJ
TO 16% ~
0
**— 1 2%) ~
Si
^ 8% -
w
S 4% -
o
3
"Z. no/.
24. 1% 24.5% 25.0% ^ 25.3% ^
I
,
,
Reference
$0.50 $1.00
Coal Price Increase ($/GJ)
$2.00
Figure 4.4: Change in the 2030 nuclear share of total electricity production as coal price is
increased parametrically.
c tu/o -
.g
5 35% -
0)
^ 30% -
| 25%-
-------
Figures 4.4 and 4.5 demonstrate that nuclear power can provide a limited hedge against high
fossil fuel prices in the electric sector. An additional scenario was run that included both an 8
$/GJ markup on natural gas and 2 $/GJ markup on coal prices to examine the synergistic effect
on nuclear capacity. The result was a nuclear share of total electricity that was approximately
37% (approximately 3% more than an 8 $/GJ mark up on natural gas alone). The high natural gas
and coal prices together did not push conventional nuclear to its growth limits, which would
allow nuclear to achieve a maximum share of approximately 45% in 2030. Because EPANMD
supply curves for coal and natural gas are based on EIA's Annual Energy Outlook, which
employs very conservative assumptions regarding fuel prices, the range of costs tested in the
sensitivity analysis above is plausible.
Surprisingly, the availability of nuclear has no effect on SO2 and NOX emissions. In all
scenarios, the Clean Air Act emission constraints on electric sector emissions are in effect.
Because air pollutant emissions are driven largely by pre-existing coal plants, the emissions
constraints require much of this capacity to be retrofitted to reduce emissions. However, new
capacity is largely SO2 and NOX emissions free: nuclear has zero operating emissions, new
pulverized coal capacity includes FGD for 862 control and SCR for NOx control that eliminate
more than 90% of emissions, and natural gas turbines have low emissions. Because the
incremental cost to retrofit existing coal plants is low, the pre-existing coal plants are retrofitted
in all model scenarios and continue to run over the entire model time horizon. The availability of
new nuclear capacity only affects the construction of new plants, which has little effect on air
pollutant emissions.
Figure 4.6 shows that CO2 emissions are also minimally impacted. The "baseline" scenario
assumes no new nuclear capacity additions and the "advanced + conventional nuclear" scenario
allows the addition of both LWRs and HTGRs. Between 2000 and 2030, 66 GW of new nuclear
(conventional + advanced) is built. By 2030, new nuclear capacity results in only a 7% reduction
in electric sector carbon emissions compared with a model scenario that does not allow new
nuclear capacity to be built. This modest emissions benefit is partly explained by the fact that the
availability of nuclear results in 20 GW of total additional generating capacity over the constant
nuclear case, presumably because some service demands can be met more cost-effectively by
electricity when nuclear is a supply option. The nuclear capacity additions have a negligible
effect on carbon emissions in sectors other than electric; the modest electric sector emissions
benefit is largely obscured in the system-wide carbon emissions.
54
-------
0.6 -
o.4-
- Advanced + Conventional Nuclear
- Baseline
§
80.6
-Advanced + Conventional Nuclear
- Baseline
2000 2005 2010 2015 2020 2025 2030
Year
2000 2005 2010 2015 2020 2025 2030
Year
Figure 4.6: System-wide (left) and electric sector (right) COi emissions relative to baseline
scenario.
In the analyses above, nuclear was unable to gain a significant market share and, as a result,
emission reductions were minimal. This is a consequence of the CAA restrictions being tighter
than the beneficial role nuclear played in generating electricity. Circumstances that depart even
further from BAU model assumptions, however, could provide stronger incentive for new
nuclear investment.
Advanced nuclear generation technologies, for instance, may play a larger role when electric
sector CC>2 emission trajectories depart significantly from BAU assumptions (see Section 2).
Alternate (non-BAU) carbon trajectories, though they represent a realistic stimulus for new
nuclear investment, primarily offer a means to flex the EPA MARKAL model and facilitate a
more convincing examination of how this investment might impact criteria pollutant emissions.
The trajectories used in this analysis include: (1) electric sector carbon emissions limited to 1995
levels from the BAU scenario (Section 2) starting in 2015, (2) electric sector carbon emissions
limited to 80% of 1995 levels from 2015 on, and (3) electric sector carbon emissions limited to
50% of 1995 levels from 2015 on. Note that these low carbon scenarios apply only to the
electric sector, not system-wide.
These carbon trajectory scenarios are deliberately arbitrary in that they do not reflect known
projections, proposed policy, or a preferred carbon emissions profile. The non-BAU trajectories
merely serve to force a signal—in this case, the adoption of advanced nuclear generating
technologies. One, of course, could constrain the model to produce a certain fraction of its
electricity from new nuclear capacity and examine the effects on criteria pollutant emissions.
Doing so, however, would be equivalent to using MARKAL as a calculator and would fail to
take advantage of its strength as an energy systems model. The non-BAU carbon trajectories
provide an incentive not only for new nuclear investment, but also for the addition of competing
technologies like natural gas and renewables. It such this competition that will drive criteria
pollutant emissions in any scenario, and accounting for these emergent system effects provides a
more realistic picture of the corresponding air quality implications.
55
-------
Figure 4.7 presents electric sector technology penetrations for the three alternative carbon
trajectories. In general, pre-existing coal capacity is forced off-line, and combined-cycle natural
gas turbines appear to be the most effective technological response in carbon limited scenarios.
However, nuclear plays an increasingly important role as the carbon trajectory departs from
BAU results. Even with high costs for the MOX fuel cycle, MOX plants play a role in the lowest
carbon scenario. Figure 4.8 illustrates how much conventional and advanced nuclear capacity
was built compared with the capacity limits set by the growth constraints. The low carbon
trajectories are the same as shown in Figure 4.7. Note that the growth in advanced nuclear is
steeper because the capacity of both PBMR and GT-MHR have been added together.
56
-------
25000
20000
Carbon Limited to 1995 Level
• Advanced Nuclear
H Other
H Renewables
D Natural Gas
D Conventional Nuclear
• Residual Coal
25000
2025
2030
• Advanced Nuclear
H Other
H Renewables
D Natural Gas
Q Conventional Nuclear
• Residual Coal
25000
5000
2000
2005
2025
2030
• Advanced Nuclear
nMOX Nuclear
H Other
H Renewables
D Natural Gas
Q Conventional Nuclear
• Residual Coal
Figure 4.7: Electricity generation by technology from non-BAU carbon trajectories and
both conventional and advanced nuclear technologies are available.
57
-------
140
120 -
100
o
o
60 -
Conventional Nuclear 2030 Maximum
Advanced Nuclear
-Conventional Nuclear
2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030
Carbon 50% of 95 Level
Carbon 80% of 95 Level
All Scenarios
Carbon Limited to 95
Level
Baseline
Figure 4.8: Conventional and advanced nuclear capacity over time, compared with the
maximum capacity allowed by the growth rate constraints.
Figure 4.8 demonstrates that advanced nuclear (HTGRs) are a compelling option for the future.
In all scenarios, the model builds the maximum allowable advanced nuclear capacity. While the
costs for HTGRs in the model are highly speculative, the results can be interpreted as
prescriptive: if developers can meet the currently projected nth-of-a-kind cost targets for HTGRs,
they are likely to play an important role in the future U.S. electricity system. Conventional
nuclear only hits the growth constraint limit in the scenario with the tightest carbon trajectory.
The higher capital costs for LWRs put them at an economic disadvantage relative to HTGRs.
An interesting result is how advanced technologies in the U.S. electricity sector do have ancillary
benefits for air quality. Figure 4.9 presents SC>2 and NOx emissions under the three non-BAU
carbon trajectories shown in Figure 4.7.
58
-------
1.2
.= 1
o 0.8
0.6
« 0.4
E
UJ
8
0.2
*-_
—*—Carbon: Baseline
—o—Carbon: 95 Levels
- -*- - Carbon: 80% of 95 Levels
- - -x- - - Carbon: 50% of 95 Levels
2000
2005
2010
2015
Year
2020
2025
2030
1.2
m
.E 1 *
o 0.8
>
«
1 0.6
« 0.4
«
UJ
>< 0.2]
—*—Carbon: Baseline
—o^ Carbon: 95 Levels
_ -a- _ Carbon: 80% of 95 Levels
- - -x- - - Carbon: 50% of 95 Levels
2000
2005
2010
2015
Year
2020
2025
2030
Figure 4.9: SOi (top) and NOx (bottom) emissions relative to the baseline scenario under
the low carbon scenarios presented in Figure 4.7.
The alternative CC>2 trajectories have a significant impact on 862 emissions, with deep
reductions in emissions under the two scenarios that depart furthest from B AU assumptions. The
NOx emissions constraint, however, is binding except under the 80 and 50 percent trajectory,
when 87 percent of the pre-existing coal-derived electricity is replaced by technologies with
lower emissions. As a result, low carbon trajectories in the electric sector will only reduce NOx
emissions when most of the pre-existing coal capacity is displaced
Suppose there is no limit to the rate at which nuclear capacity can grow. Figure 4.10
demonstrates the mix of generation sources (with nuclear growth rate constraints removed) under
a carbon trajectory that is 20% below 1990 levels from 2015 to 2030. In this case, advanced
nuclear plays a central role in electricity production after 2020.
59
-------
25000
20000
o
1
-------
Increased costs for natural gas and coal do not have a significant impact on the penetration of
nuclear. As a result, nuclear only provides a modestly effective hedge against high fossil fuel
prices in the electric sector, at least over the price range explored here. These modest penetration
levels contribute to meeting the electric sector's CAA limits for 862 and NOX.
Although both conventional and advanced nuclear are economical in the base case, nuclear
generation only has a modest impact on system-wide carbon emissions. When the electric sector
follows non-BAU carbon trajectories, new nuclear capacity plays a more significant role and
leads to SC>2 and NOX reductions below CAA constraints. In all cases, the model prefers
advanced nuclear (the high temperature gas-cooled reactors) to conventional nuclear (the light
water and MOX reactors). The issues governing this preference include whether HTGR
developers can meet the model's optimistic nth-of-a-kind cost assumptions and how rapidly units
can be manufactured and deployed over the next few decades.
61
-------
Section 5
The Air Quality Implications of Carbon Capture and Sequestration in U.S.
Electric Markets
With its abundant coal reserves, the U.S. has built an electric power infrastructure dominated by
coal-fired generation. Coal plants produce roughly half the nation's electricity, with natural gas
and nuclear plants each contributing nearly twenty percent and renewables making up the
difference (EIA, 2006a). Given the long-lived nature of this infrastructure and the lack of an
economically-competitive substitute technology, the nation's dependence on coal is unlikely to
cease. The U.S. Department of Energy (DOE) is therefore assessing technological options to
reduce the environmental impacts of coal plants. Technologies included in this assessment
include coal gasification and carbon capture and sequestration (CCS). By itself, coal gasification
would nearly eliminate atmospheric SC>2, NOX, and Hg emissions; integrated with CCS, the
combined technology could also make potentially significant reductions in electric sector CCh
emissions, should the need to do so arise (NETL, 2005). Integrated gasification-CCS
technologies could also provide the technological underpinnings of a future hydrogen economy
(FutureGen Industrial Alliance, 2006).
This section examines the air quality benefits of CCS in an energy systems context. While
climate concerns would drive the adoption of CCS, the technology could yield important air
quality benefits—benefits that would depend on the rate at which CCS units enter the market and
the technologies they displace. In addition, CCS would compete with an expanded use of natural
gas, nuclear power, and renewable energy sources. A systems level assessment is needed to
examine how these routes to electric sector CC>2 reduction interact, how this interaction affects
the economic attractiveness of CCS, and how these dynamics collectively impact criteria
pollutant emissions.
This section begins to meet the need for a comprehensive assessment. The first two subsections
define CCS as examined here, list its prominent advantages and disadvantages, briefly describe
the current state of the technology, and discuss its implementation in the U.S. EPA National
MARKAL model. The following subsection presents the analysis by identifying scenarios in
which CCS technologies enter, given business as usual assumptions about competing
technologies. Section 6 of this report broadens the assessment to look at how CCS might
compete with other supply-side abatement alternatives under more optimistic scenarios about the
latter. An evaluation of CCS relative to end-use efficiency improvements is left for future work.
The assessment in this section focuses on CCS from an energy system perspective and seeks to
identify scenarios—ranges of CCS costs and performance factors, for instance—that lead to its
adoption. The section examines generic classes of CCS power generation units and does not
evaluate the merits of particular CCS technologies. The analysis also concentrates on CC>2
capture and considers only the aggregate cost of sequestration—not the many other significant
issues surrounding underground injection of the gas. Finally, the analysis does not evaluate or
propose policies related to CC>2 control. CCS technologies, of course, will not be adopted
without the need to reduce electric sector CC>2 emissions. The analytical approach employed
here examines how CCS technologies fare relative to other emission abatement options given a
62
-------
CC>2 emission trajectory expressed as some percentage of business as usual projections in order
to examine the impact on air quality.
5.1 CCS Technology Background
CCS provides a means of reducing electric sector CC>2 emissions. In this context, CC>2 capture
works with any power generation unit utilizing coal or natural gas as its primary fuel, and may
take place before or after combustion. Sequestration occurs by injecting the captured CC>2 into a
suitable geological formation after transportation - most likely by pipeline - from the plant.
Deep saline aquifers, as well as active and depleted oil and gas reservoirs are potential
sequestration sites. CO2 may also be captured from industrial processes, and "CO2
sequestration" is often used to refer to the uptake of atmospheric CC>2 in biomass and soils.
Neither of these options is considered here. Recent publications from the International Energy
Agency (IEA, 2004) and the Intergovernmental Panel on Climate Change (IPCC, 2005) provide
up-to-date and comprehensive background information on CCS.
The value of CCS lies in its potential to ease the world's dependence on an energy infrastructure
dominated by fossil fuel consumption (Johnson and Keith, 2004). The long life and slow
turnover of this infrastructure and the current unavailability of an economic substitute support
the continued use of fossil energy—especially coal for power generation. CCS is compatible
with the electric sector as it exists today. On the generation side, power plants with CC>2 capture
(new or retrofit) would have the same capacity as their conventional counterparts, and follow
similar dispatch rules. Construction and management expertise at the plant level would also
transfer. Beyond the plant, the electric power distribution network would remain the same and
electricity consumers would experience no change in how they use energy (though it would
likely cost more). Finally, the utilities and other energy companies that currently supply
electricity would continue to do so, and the ability to exploit niche markets and take advantage of
synergies (e.g., selling captured CC>2 for enhanced oil recovery) might increase their competitive
advantage.
CCS, of course, is not without significant drawbacks. Although a potentially valuable transition
technology, CCS is an "end-of-pipe" solution that does not address the more fundamental need
to move away from an energy system reliant on carbon-intensive—and, in the case of petroleum
and natural gas, ultimately limited—fossil energy resources. In addition, successful adoption of
CCS could ease the pressure to develop more sustainable energy alternatives, and resources
invested in CCS technologies are resources not invested in renewable options (i.e., an
opportunity cost). CC>2 capture also requires considerable energy, which increases both the cost
of generating electricity and the amount of CC>2 produced per kWh. As a result, while a sudden
release of sequestered CC>2 near a low-lying inhabited area poses the vivid risk of asphyxiation, a
more general failure to contain CC>2 could—in the extreme—result in higher atmospheric
concentrations of the greenhouse gas than would have been the case if CCS had not been
pursued. These disadvantages, along with technological uncertainties and the need to address
important issues related to regulation, liability, long-term monitoring, and public acceptance,
pose obstacles to the adoption of CCS (Palmgren, et al., 2004; Wilson, 2004).
In terms of promise, however, the potential benefits of CCS are seen as offsetting its
disadvantages, and development of the technology has consequently progressed to the
63
-------
demonstration phase. The DOE's FutureGen Initiative, for instance, is the most visible U.S.
attempt to develop a large-scale power production facility with CC>2 capture and sequestration—
one that will also produce H2 and serve as a joint industry-government sponsored research
testbed within the coming decade (FutureGen Industrial Alliance, 2006). Utilities are also
beginning to evaluate the merits of designing new, retrofit compatible coal-fired or coal
gasification power generation units should they eventually decide to pursue CC>2 capture (EPRI,
2005). On the sequestration side, several ventures are operational, and three of these—in
Norway, Canada (with CO2 captured in the U.S.), and Algeria—have achieved injection rates
close to 1 MtCC>2 per year (IPCC, 2005). These examples provide an indication of the resources
being devoted to CCS-related projects, and are only a reflection of the growing level of
international research activities (see IPCC, 2005, for a review).
The initial success of CCS will depend on the integration of disparate, but largely mature,
component technologies, and scaling the resulting system up to a level able to handle the amount
of CC>2 generated annually by a typical power plant (roughly 1 MtCCVyear for a 500 MW unit).
Niche applications—in which an electric power plant with capture would provide a supply of
CC>2 for enhanced oil recovery (EOR), for instance—might help lower initial costs and lead to
learning-related cost reductions and performance improvements. Long-term success will rest on
improvements in capture technology, development of a legal framework to govern sequestration,
and public acceptance. This section focuses on the first of these three requirements (capture),
though the other two are likely to be equally significant.
CC>2 capture as currently envisioned may take place along one of three general routes (Figure
5.1). The first of these is equivalent to traditional "smoke stack" controls for SO2 and NOX and
involves post-combustion separation of CC>2 from the remaining flue gases. Applicable to both
coal-steam and natural gas combustion turbines, this approach would likely be the preferred
means of retrofitting existing power generation units (short of a complete repowering, as
discussed below). Existing post-combustion capture technology relies on chemical absorption of
CC>2 using a monoethanolamine solvent—a mature industrial process that has provided CC>2 for
use in food, beverage, and chemical production since the 1950s. Amine separation can remove
up to 95% of the CC>2 from a gas stream, though removal efficiencies in the 80% range would
likely be more common in practice. Table 5.1 (below) summarizes cost and performance data.
Several technical issues, however, create disincentives to the use of post-combustion capture
processes relying on amine separation. Solvent regeneration, steam requirements, and the need
to compress the captured CC>2 (which constitutes only 3-15% of power plant flue gas by volume),
for instance, impose an energy penalty on the order of 10-40% of the unit's output. In addition,
the amine separation process requires the use of scrubbers to remove SC>2, NOx, particulate
matter, and other flue gas impurities prior to CC>2 capture. Scale is yet another issue, with
contemporary commercial amine capture systems more than an order of magnitude smaller than
that required for a typical coal-fired power plant. Amine-based capture is therefore unlikely to
be economically competitive relative to other electric sector CC>2 emission abatement
alternatives. Post-combustion CC>2 capture from air-fired power plants may be viable if an amine
substitute becomes available. Research is currently focusing on absorption using novel solvents
(both liquid and solid), the development of adsorption processes, and the use of membranes
64
-------
(IPCC, 2005). The advantages of finding a "cheap" retrofit option for existing coal-fired power
plants provide incentives.
Post combustson
Doma^i
AiiVQ,
Ceal I Stean
BiOITBSS f X \
hTB COmDUSIIOn ijjmiwson
Qas.Oil — ^
DOil
Oxyfuel . L>as •
Air
Coal
Industrial processes G« ^«
Biarrtas?
CO,
v 1
-X I
Air
COj
4o,
Air ^~^Uir Separation ^^^^^^^^^J
'^
n COf
\x
con
/A Dal:ydialioii
— JT
/
1 /
/
^
Riw rraUriil
Gas, Ammonia, Steel
Figure 5.1: Schematic of the three generic routes to carbon capture and sequestration;
industrial process are shown, but not included in this analysis; Source: IPCC (2005)
Amine-based CC>2 capture is conceptually similar to post-combustion controls for SC>2 and NOx;
except for the energy penalty, operation of a power plant would not "look" different with flue
gas CC>2 capture. So-called oxy-fuel processes offer a second post-combustion route to CC>2
capture, but one that would require more significant changes in power plant design. Rather than
combusting coal or natural gas in air, an oxy-fuel system would use a mix of pure O2 and
recycled flue gas CC>2. As combustion in 62 produces mainly CC>2 (at least 80% of flue gas
volume) and water vapor, capture would require little more than drying the flue gas and
compressing the CC>2. An oxy-fuel system would therefore capture nearly all CC>2 produced
(again, see Table 5.1 for cost and performance data). Small-scale oxy-fuel combustion processes
have found industrial applications, but do not exist as integrated systems at the scale needed for
power generation. Like amine separation, an oxy-fuel power plant would operate at a lower
efficiency than its conventional counterpart, with the C>2 production process (a mature
technology) responsible for much of the added energy requirement. Applied to existing power
plants, oxy-fuel conversion is considered to be a repowering option (i.e., replacing a plant's core
energy generating technologies), rather than a retrofit modification.
The last generic route to electric sector CO2 capture differs from the first two by separating the
carbon from the fuel stream prior to combustion. The approach to separation most likely to
shape the design of new power plants with CC>2 control, pre-combustion capture is a mature
process that the fk, synthetic fuel, and chemical industries use routinely. The process begins
with either steam reforming or partial oxidation of natural gas, or gasification of coal, to produce
65
-------
H2 and CO (other byproducts—such as H2S from coal gasification—need to be removed). A
water gas shift reaction then produces additional H2 while converting CO into a high pressure
CO2 stream. The higher pressure simplifies the CO2 capture process (which is typically
accomplished via physical absorption) and reduces its energy requirements, improving overall
system efficiency. The H2 is available for use in a combined turbine and steam cycle power
generation unit, though it could also be used in a fuel cell to produce power or sold as a
transportation fuel or industrial feedstock. Coal-based IGCC (integrated gasification combined-
cycle) plants with CO2 capture are the most frequently mentioned pre-combustion CCS
technology in the literature, both for new plant construction and existing unit repowering. IGCC
technologies (with or without capture), however, are currently limited to high-ranked coals. Like
oxy-fuel conversion, pre-combustion CO2 capture applied to existing power plants would involve
significant repowering rather than a retrofit add-on.
Table 5.1 compares cost and performance estimates for new technologies representing the three
generic routes to CO2 capture described above, and contrasts these with their non-CCS
counterparts. With the benefit of research and learning-by-doing, newly-built plants with CO2
capture in 2020 are expected to achieve costs and efficiencies similar to their non-capture
equivalents today (IPCC, 2005).
Table 5.1: Estimated cost and performance ranges from the literature for new CCS technologies
as summarized by the IPCC (2005).
Technology
Post-Combustion
PC*
PC + CCS
Pre-Combustion
IGCC
IGCC + CCS
NGCC**
NGCC + CCS
Oxyfuel
Oxyfuel
Oxyfuel + CCS
Capital Cost
($/kW)
1161-1486
1894-2578
1169-1565
1414-2270
515-724
909-1261
1260-1500
1857-2853
Cost of
Electricity
($/MWh)
43-52
62-86
41-61
54-79
31-50
43-72
44-45
58-83
Thermal
Efficiency (%
LHV)
41-45
30-35
38-47
31-40
55-58
47-50
37-44
25-35
CO2 Capture
Efficiency
n/a
85-90
n/a
85-91
n/a
85-90
n/a
Insufficient data
*PC = pulverized coal
**NGCC= natural gas combined-cycle
The costs of retrofitting existing plants, in contrast, are highly uncertain and are more likely to be
site-specific. The IPCC, for instance, notes that "[tjhere has not yet been any systematic
66
-------
comparison of the feasibility and cost of alternative retrofit and repowering options for existing
plants" (IPCC, 2005, p. 344). The energy penalty associated with amine-based post-combustion
retrofits renders this option economically unattractive; oxy-fuel or pre-combustion repowering of
coal units, however, may be worthwhile as the end result is essentially a new—and likely
larger—plant without the obstacles associated with developing a greenfield site. A further
distinction lies between retrofitting or repowering plants that exist today, and designing new non-
capture power plants for future conversion. The cost of the latter, of course, will be cheaper and
more predictable.
The analysis that follows focuses on the penetration of electric sector CC>2 capture technologies.
The sequestration component of CCS enters the assessment only as a cost, which consists of CC>2
transport, injection, and monitoring components. The first two elements involve mature
technologies and processes with which the oil and gas industries have considerable experience.
Land-based pipeline transport as currently practiced, for instance, costs 1-5 $/tCC>2 per 100 km.
Applied to CCS, future injection cost estimates range more widely—from 0.5 to 8.0 $/tCC>2,
depending on the site—with the need to monitor CO2 containment adding 0.1 to 0.3 $/tCO2
(IPCC, 2005). Note that sequestration capacity is not likely to pose a problem. Deep saline
aquifers (i.e., below 800 m) alone may hold 1000 to 10,000 GtCO2—several orders of magnitude
greater than yearly global CC>2 emissions (which are on the order of 25 GtCCVyear; IEA, 2004).
Sequestration in depleted oil and gas reservoirs provides another option and, as noted, the ability
to sell CC>2 for enhanced oil recovery (EOR) might provide a profitable and early niche market—
and a route to achieving learning-by-doing technology improvements. EOR operations in the
U.S., for instance, have historically paid 10-16 $/tCC>2, depending on oil and gas prices (LEA,
2004).
While the cost of CC>2 sequestration is likely to be significantly less than that of capture, the
institutional uncertainties associated with injection and storage are greater. The necessary legal
frameworks (both domestic and international) are not in place, risk management trade-offs are in
need of resolution, and public acceptance of large-scale underground CC>2 storage remains
untested. In addition, site-specific factors will play a much greater role in determining
sequestration (and, hence, emission abatement) potential in a given region. The future of CCS
rests as much on resolving institutional issues such as these as its does on solving technical
problems (Wilson, Johnson, and Keith, 2003).
Integrated assessments, however, suggest that CC>2 capture technologies could sequester nearly
50% of projected global CC>2 emissions by mid-century at a cost ranging between 25 and 50
$/tCO2 (IEA, 2004; IPCC, 2005). While CO2 capture would add at least 2-3 cents/kWh to
electricity prices today, this premium would likely drop by half over a few decades. CCS could
therefore be an important element in a portfolio of electric sector emission abatement options.
The remainder of this section examines the place of CCS in this portfolio from an air quality
standpoint.
67
-------
5.2 Implementation of CCS in the U.S. EPA National MARKAL Model
The analytical approach adopted in the remainder of this section examines generic classes of
electric sector CCS technologies (e.g., IGCC with CO2 capture) rather than specific proposed
designs (e.g., a Texaco oxygen-blown, quench-based gasifier). Reference values from the range
of studies surveyed by the IPCC (2005) and summarized in Table 5.1 provide baseline cost and
performance data for these technology representations—a starting point to more informative
uncertainty analyses. Anderson and Newell (2004), IEA (2004), Johnson and Keith (2004), and
EIA (2006b) provide additional data. Table 5.2 lists the CCS-related technologies included in
the U.S. EPA National MARKAL model database, and provides details about their
specifications. EPA (Shay,et al., 2006) provides documentation on the base MARKAL model.
All CCS technologies become available in the 2015 model period, have a 40 year useful life, and
share an 18 percent investment hurdle rate.
Table 5.2: CCS technology parameters as implemented in the EPANMD. All figures were
converted into common units from their actual MARKAL values.
Technology
Retrofits
Existing Coal
Retrofit
New PC Retrofit
IGCC Retrofit
NGCC Retrofit
New Integrated
Technologies
IGCC+CCS
NGCC+CCS
Capital Cost
($/kW)
1414
1345
966
763
1873
1021
Variable
Operating
Cost ($/kWh)
0.0093
0.0085
0.0045
0.0022
0.0040
0.0027
Fixed
Operating
Cost ($/kW)
26.24
21.84
8.99
5.77
41.44
18.12
Thermal
Efficiency
(%)
65
70
80
80
40.0
42.9
CO2 Capture
Efficiency
(%)
85
85
90
85
90
90
The retrofit parameterization requires a few words of explanation. First, the incremental retrofit
costs are higher than published estimates for amine-based systems (see, e.g., Simbeck and
McDonald, 2001; IPCC, 2005) to account for the fact that the MARKAL model employs base
plant (i.e., conventional PC, IGCC, and NGCC capacity) cost figures that are more optimistic
than the retrofit studies typically assume. To prevent MARKAL from immediately retrofitting
newly-built conventional capacity instead of building a new integrated capture plant, the model
adopts retrofit cost parameters that ensure that the combined investment and operating costs of
retrofit capacity (base plant costs + retrofit costs) are at least as great as those of the
corresponding integrated capture plant. Second, the retrofit "efficiency" is perhaps better
interpreted as an energy penalty (i.e., a base plant output derating). The inverse of the retrofit
efficiency (expressed as a decimal fraction) is the increase in input energy required per unit of
68
-------
retrofit energy output. The corresponding energy penalty is then this increase divided by 1 plus
itself [energy penalty = increase/(l+increase)]. Finally, since all retrofits sit in the fuel chain
leading into the base plant, the MARKAL database adjusts the retrofit parameters to account for
the base plant efficiency.
Note that the model does not include new integrated PC capture plants or an oxyfuel option.
These technologies are not sufficiently different from new IGCC capture units to be
meaningfully distinct in MARKAL. IGCC is generally seen as the least-cost coal-fired capture
alternative and the model therefore adopts the label for its new coal CCS technology; investment
in IGCC capture technologies could therefore represent one of several technologies.
In keeping with this aggregate technology representation and the analytical focus on supply-side
CC>2 abatement options, the augmented model uses a single figure (28 $/tC, or approximately 7.5
$/tCC>2) to represent the cost of CC>2 transport, injection, and long-term monitoring. This value
reflects the upper end of published estimates, and assumes that geological injection sites are
located within 300 km of all central station power generating units (IPCC, 2005).
This generic representation of CCS technologies is compatible with the level of detail
characterizing other electric sector power generation technologies in the EPA National
MARKAL database (e.g., three model plants represent all existing U.S. coal-fired units). This
level of detail is also compatible with both the uncertainty inherent in CCS technologies, the
nature of a linear programming optimization model like MARKAL, and the scenario-based
analytical strategy that guides the following assessment.
Choices made in modeling carbon capture retrofits provide a useful illustration of this point. As
discussed, CC>2 capture using solvent absorption is a mature industrial process. The costs and
performance issues associated with retrofitting existing coal-fired power plants at the scale
necessary to make a significant reduction in electric sector emissions, however, are uncertain.
Contemporary amine-based technologies would impose such a significant energy penalty that
alternative technologies (e.g., membranes) would likely be needed for the post-combustion
retrofit option to become economically feasible. All retrofit technologies would be subject to
uncertain learning, and site-specific details such as the space available for the capture equipment
would drive installation costs. Differences in the specifications of proposed retrofit technologies
easily fall within the range of this uncertainty, but an optimization model like MARKAL will
always pick the technological option with the lowest overall costs, inflating the apparent
significance of that technology and possibly leading to rapid and unstable period-to-period shifts
in favored options. Modeling identifiable but marginally different retrofit schemes would
therefore produce meaningless and potentially misleading results. (Note that these schemes may
have significantly different technological potential; unless they also have substantially different
costs and efficiencies, however, they essentially look the same within MARKAL. The most one
could do is lower a technology's availability to reflect decreased reliability. Radically different
retrofit designs, of course, should be modeled independently.)
Given that small differences in parameter values often lack practical significance in a modeling
context like this, a reasonable alternative would look only at broad technology classes and then
use sensitivity analysis around their reference parameter values to identify ranges of these values
69
-------
that influence results. MARKAL characterizes CCS retrofits, for instance, by their costs and
energy penalties. One could use parametric sensitivity analysis to examine how variations in
model retrofit specifications affect retrofit adoption, and then trace "successful" parameter
combinations back to specific retrofit designs. The difference in approach between modeling a
suite of technologies and using sensitivity analysis in conjunction with a representative example
may seem subtle, but the generic strategy—for reasons described above—is more defensible for
this type of analysis and is consequently employed below.
5.3 CCS Results and Analysis
The remainder of this section does for CCS what the previous section did for nuclear power
generating technologies. Building on the business as usual (BAU) scenario results (Section 2),
the analysis examines the extent to which CCS technologies contribute to meeting electricity
demand and its impact on air quality under the same set of alternative electric sector CO2
trajectories that framed the nuclear assessment. It is worth repeating that these non-BAU
trajectories are arbitrary and merely a modeling device used to stimulate CCS investment, and do
not represent policy or endorse a particular carbon emissions profile.
Outcomes of interest from this analysis include: patterns of investment in CCS technologies and
their electricity output; the balance between new, integrated CCS units and existing capacity
retrofits; the share of coal- and natural gas-fueled CCS technologies; the competition between
coal-to-gas fuel switching and CCS; and ultimately the impact of CCS penetration on electric
sector criteria pollutant emissions. Scenario drivers within the CO2 trajectory framework include
variations in CCS parameters (especially costs and efficiencies), natural gas prices, and
sequestration costs. This analysis assumes BAU nuclear and renewable generating capacity.
Section 6 allows new nuclear and renewable technologies into the model and concludes the
report with a look at the competition between these alternatives to fossil-fuel based power
generation and CCS.
Tables 5.3 and 5.4 summarize electricity generation by technology class and new capacity
investment, respectively, for a model run with CCS technologies (as described above) and an
electric sector CO2 trajectory that holds emissions constant at 1995 levels from 2015 on
(approximately 2.4 Gt CO2 per five-year period). Figure 5.2 illustrates how different
technologies contribute to meeting increasing power demand over time, and Figure 5.3 shows the
impact of CCS technology penetration on electric sector CO2, SO2, and NOx emissions, as well
as economy-wide CO2 output. Compared to the non-CCS BAU analysis (Section 2), coal-fired
generation actually increases slightly at the expense of gas, with new integrated IGCC+CCS
capacity providing the CO2 emissions reduction. Neither NGCC units with CO2 capture nor
retrofits of any types enter. Note that the level of investment in conventional gas technologies is
not significantly different from the non-CCS base case.
70
-------
Table 5.3: Baseline CCS scenario electric power generation (in PJ/period) by technology class
and time.
Technology
Nuclear
Existing Coal
New PC
IGCC
Existing Coal
Retrofit
New PC
Retrofit
IGCC Retrofit
IGCC+CCS
Gas
NGCC Retrofit
NGCC+CCS
Renewables
Other*
Electricity Generation (PJ/period)
2000 2005 2010 2015 2020 2025 2030
2734
7485
0
0
0
0
0
0
1424
0
0
702
550
2788
8095
64
0
0
0
0
0
1661
0
0
781
481
2831
7900
64
0
0
0
0
0
2676
0
0
827
439
2874
6452
64
0
0
0
0
674
3767
0
0
1369
132
2916
6225
64
0
0
0
0
2012
4402
0
0
1445
146
2916
5972
64
0
0
0
0
2250
5003
0
0
1583
143
2916
5558
64
0
0
0
0
3201
5713
0
0
1685
220
*Other includes natural gas-fired microturbines (distributed generation), diesel combustion
engines, and generation from municipal solid waste and landfill gas.
Understanding the results involves analyzing the relationship between the ways in which
different technologies meet electricity demand, air quality goals, and the need to recover capital
investment. In general, coal-fired power plants have high capital requirements, but low
operating costs relative to gas turbines (the latter due largely to differences in fuel costs). This
cost difference, combined with dissimilar operating characteristics (i.e., coal plants are not
simply "turned on"), explain why coal units typically supply baseload power, while natural gas
units are especially suited for load-following during peak demand hours. Coal, of course, is the
most carbon-intensive fossil fuel and baseload generation (to the extent that it is not met by
nuclear plants) is therefore especially carbon intensive. Hence, CCS enters with coal (here,
IGCC). This pattern also makes sense from a CCS investment standpoint. CC>2 capture is costly.
If a utility invests in CCS technology, it will want to use that capacity to the limits of its
availability (reliability), since the per kWh costs of the technology decline with increased power
production. CCS therefore enters the dispatch order to meet baseload demand, while peaking
units simply do not generate enough electricity to make CCS worthwhile (see Johnson and Keith,
2004).
71
-------
Table 5.4: Baseline CCS scenario new capacity investment (in GW/period) by technology class.
Technology
Nuclear
Existing Coal
New PC
IGCC
Existing Coal Retrofit
New PC Retrofit
IGCC Retrofit
IGCC+CCS
Gas
NGCC Retrofit
NGCC+CCS
Renewables
Other
New Capacity Investment (GW/period)
2000
0
0
0
0
0
0
0
0
38
0
0
0
4
2005
0
0
2
0
0
0
0
0
77
0
0
9
4
2010
0
0
0
0
0
0
0
0
0
0
0
1
4
2015
0
0
0
0
0
0
0
25
98
0
0
7
4
2020
0
0
0
0
0
0
0
50
43
0
0
4
4
2025
0
0
0
0
0
0
0
9
31
0
0
5
4
2030
0
0
0
0
0
0
0
35
32
0
0
6
4
20
O *f.
o 15
o
§ 10
£H
O
O
'£
o
2°000
Conventional
Retrofit CCS
New CCS
Renewable
Gas
Coal
Nuclear
2010
2020
2030
Figure 5.2: Electricity generation per-period by technology class for the baseline CCS
scenario (electric sector COi emissions held constant at 1995 levels from 2015 on, with no
CCS growth constraints).
72
-------
The emissions numbers (Figure 5.3) tell an interesting story. Electric sector CC>2 emissions
decline to nearly two-thirds of their BAU levels by 2025, while system-wide carbon output
decreases by about 10 percent. Electric sector NOX emissions, however, remain essentially
unchanged at their constrained BAU levels (which account for Clean Air Act mandated limits),
while SC>2 declines roughly 20 percent for two periods before hitting its cap once again. The
explanation for this behavior lies with the deployment of existing coal plant criteria pollutant
control retrofits (FGD especially), which are used approximately 20 percent less in the CCS
scenario. IGCC plants with CC>2 capture also reduce criteria pollutant emissions, easing the
burden on coal plant retrofits to meet emission constraints.
1.20
0.60
2000
2005
2010
2015
2020
2025
2030
O - - System-wide CO2
-A—Electric Sector NOX
-Electric Sector CO2
-Electric Sector SO2
Figure 5.3: Electric sector COi, SOi, and NOx and economy-wide COi emissions as a
function of time for the baseline CCS scenario.
One objection to these results is the rate at which CCS investment takes place, starting in 2015
(Table 5.4). MARKAL, as a "perfect foresight" modeling framework, assumes the position of a
rational decision maker with complete information about the future. The model thus determines
the least-cost means of meeting demand for energy services over the entire time horizon, making
all investment and operating decisions simultaneously. Actual investors do not have the luxury
of clairvoyance, or necessarily the long-lead time represented here (20 years from 1995 to 2015),
to pursue research and development and ready new technology (such as IGCC with CO2 capture)
for the market. Results would likely look different in a myopic model that made investment
decisions on a period-by-period basis without consideration of future operating constraints.
Adoption of novel technologies like IGCC would be more gradual, at least until reliability issues
were resolved, sufficient learning occurred, and economies of scale were achieved.
Picking a growth rate, however, is difficult for the non-BAU world represented here. Achieving
an emissions trajectory similar to the CCS scenario's would induce technology change in an
73
-------
unpredictable manner. With sufficient leadtime, levels of technology adoption as shown in
Table 5.4 would not necessarily be unreasonable, though one might expect to see IGCC
appearing gradually from an earlier date (with CC>2 capture used when needed). Furthermore, the
observed model behavior does not seem far out of line when one considers that sequestered CO2
represents 6 percent of the CCS scenario's electric sector total carbon emissions in 2015 (5
percent of the BAU scenario's), and increases to only 30 percent (21 percent of the BAU level)
in 2030. Finally, as noted earlier, the model labels its integrated coal-fired, CC>2 capture
technology "IGCC" as current thinking tends to assume that this combination would be cheaper
than new pulverized coal with capture. Advanced PC plants with capture, however, could
achieve costs and efficiencies similar to those used here to represent IGCC and would be nearly
indistinguishable in MARKAL. The large per-period investment in what appears to be a single
technology could actually be a mix of advanced coal-based designs. The following set of
scenarios explore the growth rate issue in tandem with the other CC>2 trajectories. One must keep
in mind, however, that results should always be interpreted as indicating what would be optimal
from a least-cost perspective - a target, for instance, for allocating research, development, and
deployment resources.
The initial CCS scenario added a total of 119 GW of new IGCC units with capture through 2030,
starting with 25 GW in 2015. Restricting CCS growth to an initial 10 GW per technology
(including retrofits) with a 10 percent annual growth rate limit reduces the cumulative CCS
installation over the same timeframe to 106 GW, with 94 GW IGCC (a binding constraint) plus
12 GW NGCC; 40 GW of new conventional NGCC, entering between 1995 and 2030, makes up
the difference. A 5 GW initial period limit with the same growth rate results in 89 GW, split
nearly evenly between new IGCC and NGCC with capture. Both restrictions lead to a small
investment in existing coal plant retrofits (< 10 GW). Note that a growth rate restriction similar
to that employed in the nuclear analysis (i.e., a 3 or 4 GW first-period limit and initial annual
growth rate of 25 percent) does not affect cumulative CCS installation. The remainder of the
section (except where noted) uses a 10 GW per technology initial investment limit with a 10
percent annual growth rate. The more restrictive 5 GW initial limit may seem more realistic but,
as argued above, induced technology change is difficult to predict and the looser limit reduces
the chance that modeling assumptions will arbitrarily drive results.
Growth rate restrictions play a more important role with electric sector CC>2 trajectories that
depart further from BAU assumptions. Figures 5.4 and 5.5 depict the least-cost mix of
generating technologies that attain emission trajectories corresponding to 80 and 50 percent of
1995 level, respectively, starting in 2015. Both scenarios adopt a mix of IGCC (94 GW in each)
and NGCC capture units (52 and 68 GW, respectively), with a minor retrofit investment. As
before, the IGCC limits are binding. New NGCC units also play a larger role in each scenario,
with nearly 500 GW of new (conventional) capacity installed in the most restrictive CC>2
trajectory between 1995 and 2030. Compared to the initial CCS scenario (which held CC>2
emissions at 1995 levels), 862 emissions decline significantly below BAU output (Figure 5.6)
for both emission trajectories examined here. Electric sector NOx emissions decline below their
constrained values only for the lowest carbon trajectory.
74
-------
Conventional
Retrofit CCS
Renewable
Gas
Coal
Nuclear
2000
2010
2020
2030
Figure 5.4: Electricity generation per-period by technology class for the 80 percent
emissions scenario (electric sector COi emissions held constant at 80 percent of 1995 levels
from 2015 on, with CCS growth constraints in place).
Conventional
Retrofit CCS
New CCS
Renewable
Gas
Coal
Nuclear
2000
2010
2020
2030
Figure 5.5: Electricity generation per-period by technology class for the 50 percent
emissions scenario (electric sector COi emissions held constant at 50 percent of 1995 levels
from 2015 on, with CCS growth constraints in place).
75
-------
1.20
0.
2000
2005
2010
2015
2020
2025
2030
O - • System-wide CO2
-A—Electric Sector NOX
-Electric Sector CO2
•Electric Sector SO2
(a)
0.00
2000
2005
2010
2015
2020
2025
2030
D - - System-wide CO2
-A— Electric Sector NOX
-Electric Sector CO2
•Electric Sector SO2
(b)
Figure 5.6: Electric sector COi, SOi, and NOx and economy-wide COi emissions as a
function of time for (a) the 80 percent emissions scenario and (b) the 50 percent emissions
scenario.
The relative consistency of these results across emission scenarios points to the need to examine
how CCS technology specifications affect model behavior. Tables 5.5 and 5.6 show how
different generating technologies compete to meet demand for two sets of parametric sensitivity
analyses under the original CC>2 trajectory. To avoid conflating growth rate dynamics with the
effects of interest, both sets of analyses exclude these constraints.
The first set of analyses (Table 5.5) varies the costs and input fuel requirements (the inverse of
efficiency) of all CCS technologies (new and retrofit) by ± 20 percent from their base line
76
-------
values. Two trends stand out. First, when CCS technologies perform better and cost less than
baseline assumptions, new integrated NGCC units with CC>2 capture enter to meet part of the
shoulder generation (between base and peak load) previously supplied by conventional NGCC
capacity. The overall share of gas technologies, however, remains constant (even though it is not
constrained) and existing coal units increase their output to meet part of the baseload generation
previously supplied by IGCC plants with capture. The efficiency improvement in this scenario
reduces the NGCC capture unit operating (fuel) cost disadvantage relative to IGCC. Next, when
efficiencies are low, but costs high, existing coal plant retrofits enter at the expense of IGCC
capture units. These results correspond to intuition: retrofits are at both a cost and efficiency
disadvantage to new integrated CCS units. When these shortcomings are removed, retrofits
become somewhat competitive. The next set of sensitivity analyses explores this dynamic
further. Note that the high CCS cost, high fuel requirement (low efficiency) scenario looks
similar to the base CCS run; differences are due to rounding and the slightly reduced centralized
electric power production in the former scenario.
Table 5.5: Share of electricity generation in 2030 by technology for the initial CCS scenario
(CC>2 emissions held constant at 1995 level from 2015 on) and four parametric analyses of
differences in CCS technology costs (investment plus all operating) and efficiencies (expressed
here as change in input fuel requirements) relative to baseline values.
Share of Electricity Generation in 2030
Fuel Input
Cost
Nuclear
Existing Coal
New PC
IGCC
Existing Coal Retrofit
New PC Retrofit
IGCC Retrofit
IGCC+CCS
Gas
NGCC Retrofit
NGCC+CCS
Renewables
Other
-20%
-20%
0.15
0.36
0.00
0.00
0.00
0.00
0.00
0.10
0.14
0.00
0.15
0.09
0.01
-20%
20%
0.15
0.30
0.00
0.00
0.09
0.00
0.00
0.06
0.30
0.00
0.00
0.09
0.01
20%
-20%
0.15
0.26
0.00
0.00
0.00
0.00
0.00
0.20
0.30
0.00
0.00
0.09
0.01
20%
20%
0.15
0.27
0.00
0.00
0.00
0.00
0.00
0.18
0.30
0.00
0.00
0.09
0.01
Base CCS
Scenario
0.15
0.29
0.00
0.00
0.00
0.00
0.00
0.17
0.30
0.00
0.00
0.09
0.01
The second set of parametric sensitivity analyses attempts to tease out the main drivers of CCS
retrofit behavior (Table 5.6). Once again, two trends are noticeable. First, energy penalty
improvements (decreases in input fuel requirements) help retrofits of existing pulverized coal
77
-------
plants and new IGCC capacity without CCS become economically competitive with integrated
IGCC capture units. As efficiency improves further, coal plant retrofits loose market share to
NGCC capture conversions. This is the same pattern seen above: the efficiency improvement
reduces the operating cost disadvantage gas plants experience as a result of their higher fuel
costs. New, integrated NGCC capture units are not affected by this sensitivity analysis and do
not enter the mix of generating plants. Equivalent improvements in retrofit costs do not have as
large an impact. As investment and operating costs decline, retrofits of existing coal plant
increase, reducing the share of power new IGCC capture units generate. Simultaneous
improvement in both retrofit costs and efficiencies combines these effects. Note that new
pulverized coal plant retrofits never become a competitive investment option.
Table 5.6: Share of electricity generation in 2030 by technology for the initial CCS scenario
(CC>2 emissions held constant at 1995 BAU levels from 2015 on) and four parametric analyses of
differences in CCS retrofit technology costs (investment plus all operating) and efficiencies
(expressed here as change in input fuel requirements) relative to baseline values.
Share of Electricity Generation in 2030
Fuel Input
Cost
Nuclear
Existing Coal
New PC
IGCC
Existing Coal Retrofit
New PC Retrofit
IGCC Retrofit
IGCC+CCS
Gas
NGCC Retrofit
NGCC+CCS
Renewables
Other
-20%
100%
0.15
0.30
0.00
0.00
0.09
0.00
0.06
0.00
0.30
0.00
0.00
0.09
0.01
-40%
100%
0.15
0.37
0.00
0.02
0.04
0.00
0.02
0.00
0.12
0.18
0.00
0.09
0.01
100%
-20%
0.15
0.28
0.00
0.00
0.03
0.00
0.00
0.14
0.30
0.00
0.00
0.09
0.01
100%
-40%
0.15
0.28
0.00
0.00
0.06
0.00
0.00
0.12
0.30
0.00
0.00
0.09
0.01
-20%
-20%
0.15
0.34
0.00
0.00
0.06
0.00
0.05
0.00
0.20
0.10
0.00
0.09
0.01
Base CCS
Scenario
0.15
0.29
0.00
0.00
0.00
0.00
0.00
0.17
0.30
0.00
0.00
0.09
0.01
These CCS sensitivity analyses showed that gas units with CC>2 capture could be competitive
with their coal-fired counterparts when efficiency improvements reduced the impact of the price
difference between natural gas and coal. More generally, fuel switching from coal to gas
provides an alternative to CCS. Hence, the impact of natural gas prices on technology
penetration is worth exploring via its own sensitivity analysis. Figure 5.7 compares the share of
generation in 2030 from conventional gas versus CCS units for the baseline CCS scenario plus
additional scenarios incorporating a 10 $/GJ markup in economy-wide natural gas prices as well
as a 3 $/GJ cut. The markup decreases the share of gas generation by 3%, and leaves CCS
78
-------
unchanged. The gas price cut affects neither. These results are partly an artifact of the
MARKAL model, which treats NGCC as non-baseload capacity and therefore prevents NGCC
capture units (new and retrofit) from substituting for their coal-fired counterparts. Gas price
markups smaller than 10 $/GJ had no effect on results. Note that these scenarios do not include
the CCS growth constraints; equivalent runs with the constraints in place prevented investment
in additional coal-fired CCS capacity (the preferred option) and only increased the overall share
of gas 4 percent by adding new NGCC plants with capture when gas prices were low
(conventional NGCC did not change).
0.40
0.30 -
0.20-
o.io -
0.00
III
• Gas
DCCS
Base CCS
10 $/GJ Markup
Gas Price Scenario
1 $/GJ Cut
Figure 5.1: Share of electricity generation for conventional gas and all CCS technologies
under three gas-price scenarios.
Finally, the analysis in this section has focused on the technological side of CC>2 capture.
Sequestration enters the analysis as a single 28 dollar cost per ton of carbon (approximately 7.5
S/tCCh) transported and injected, based on a moderate plant-to-well distance and a high-end
estimate for both costs derived from the IPCC synthesis report (2005). Actual sequestration
costs will be site specific. Moreover, as noted, an average cost figure fails to capture important
drivers such as uncertainty over public acceptance, the regulatory environment, and the need for
long-term monitoring. Where CC>2 has economic value, such as for enhanced oil recovery, the
gas could be sold for profit. Such opportunities could provide a niche market for capture
technologies, fostering learning-by-doing and lowering its costs. These issues require a more
detail analytical strategy and modeling environment than employed here. As a check on
sensitivity to model assumptions, however, two additional scenarios explored the impact of
halving and doubling sequestration costs - the likely range of variation for power plants
operating in the U.S. Neither change had an appreciable effect on results for the baseline CCS
scenario, supporting the view that capture would dominate CCS from an economic perspective
(TEA, 2004; IPCC, 2005).
79
-------
5.4 Conclusions
This section has taken an initial look at how carbon capture and sequestration technologies
function in the U.S. EPA National MARKAL energy systems model. The energy systems
context is important as CCS would compete with measures to reduce the carbon intensity of
power production and efforts to improve end-use efficiency. Driven by an assumed electric
sector CC>2 trajectory, the model adopts CCS mainly for baseload generation in the form of new
IGCC capture capacity. The analysis does not consider demand-side measures, but the model is
seen to rely on coal-to-gas fuel-switching as much as it does on CCS under the assumed carbon
trajectories. The next section examines CCS relative to nuclear power and renewable energy
sources, two other prominent means of reducing the carbon intensity of power production. Like
the previous section's nuclear power analysis, moderate levels of CCS adoption do not
significantly affect electric sector criteria pollutant emissions. CCS displaces new baseload
capacity under these circumstances, but does not significantly impact the operation of existing
coal-fired plants, which contribute most of the sector's Clean Air Act constrained 862 and NOx.
Several limitations of this section's analysis should be kept in mind. First, MARKAL is not a
dispatch model. Consideration of how different generation technologies operate over the course
of a typical day would help tease out how technology use affects investment patterns, as
discussed earlier (Johnson and Keith, 2004). Technology adoption—particularly the competition
between CCS and coal-to-gas fuel switching—would also look different in a model that treated
NGCC units as baseload capacity (a competitor to IGCC capture units). Likewise, the analysis
would benefit from a more detailed representation of existing coal-fired power plants; the retrofit
option, for instance, might look better with greater model resolution. Next, the model assumes
static cost and efficiency values for all CCS technologies. Since, learning and economy-of-scale
improvements in both sets of CCS parameters would likely occur, this assessment of CCS is
probably conservative (though the course of technology development is difficult to predict and
all electric sector technologies would be subject to leaning). Finally, CCS is not restricted to the
electric sector. IGCC capture plants, for instance, could produce electricity as needed during the
day, and use excess capacity during non-peak hours to produce H2 for use as a transportation
fuel. Other industrial applications of CCS are possible (see, e.g., IPCC, 2005). Future work will
explore these issues to see their full effect on air quality.
80
-------
Section 6
Nuclear Power, Carbon Capture and Sequestration, and Wind Generation:
The Competition Among Technology Alternatives
The previous sections built on the business as usual (BAU) analysis to examine how nuclear
power or CC>2 capture and sequestration (CCS) technologies might contribute to meeting demand
in U.S. electric markets and, in turn, impact air quality. Separate assessments enabled a higher-
resolution look at the factors that affect the adoption of these alternatives to conventional power
generation and the choice among the specific technologies that make up each class. This section
explores how nuclear and CCS compete in the EPA National MARKAL Model (EPANMD).
The analysis also loosens the constraints on wind generation to examine how these technologies
fare against a prominent renewable energy option. Once again, the focus is on technology
investment and how the resulting patterns of activity affect electric sector criteria pollutant
emissions.
Like the preceding sections, this analysis relies on the arbitrary non-BAU carbon trajectories as a
modeling device to stimulate investment in nuclear and CCS. Figure 6.1, for instance, shows
how CCS fares against conventional nuclear power (light water and mixed-oxide reactors),
natural gas, and wind when electric sector CO2 emissions level off at its 1995 level in 2015. All
technologies incorporate the baseline costs, efficiencies, and growth constraints assumed in the
previous sections. Nuclear power increases its contribution to baseload generation, with 126
GW of new investment through 2030 (bumping up against its growth constraint), while a more
modest 26 GW of new IGCC capture capacity comes on-line. Electric sector 862 and NOx
emissions remain constant at the bounds representing CAA rules. Offering advanced nuclear
technologies (gas turbine modular helium and pebble bed modular reactors) as investment
options increases nuclear power output slightly (from 33 to 34 percent of 2030 generation) at the
expense of CCS (which decreases from 4 to 2 percent); the addition does not affect gas output
(27 percent), wind (less than 10 percent), or criteria pollutant emissions.
81
-------
20
O -.,-
o 15
o
o
2000
Conventional
Retrofit CCS
New CCS
Renewable
Gas
Coal
Nuclear
2010
2020
2030
Figure 6.1: Electricity generation per period by technology class for the baseline CCS
scenario (electric sector COi emissions held constant at 1995 levels from 2015 on, with no
CCS growth constraints). Note that nuclear includes only conventional (LWR and MOX
reactors) technology.
A further departure from the electric sector CC>2 trajectory alters this picture slightly. Figure 6.2,
for instance, illustrates how different technologies contribute to meeting electricity demand for a
CO2 trajectory that holds steady at 50 percent of 1995 emissions from 2015 on. When nuclear
investment is limited to conventional technologies, nuclear power once again hits its growth
constraint. CCS investment, however, increases, with integrated IGCC CC>2 capture units
generating 7 percent of 2030's electrical output and NGCC with capture producing an additional
4 percent. The model also adds CCS retrofits to a modest part of the existing coal plant capacity.
Criteria pollutant emissions under this scenario decrease well below their upper bounds, as
generation in 2030 from the model's (nonretrofit) existing coal capacity drops from near 28
percent in the previous two scenarios to slightly more than 3 percent. SCh emissions are reduced
by about 90% below CAA restrictions, and NOX emissions are reduced by half (Figure 6.3).
Expanding investment options to include advanced nuclear technologies increases nuclear
generation from 34 to 42 percent of total 2030 output, while the contribution from both IGCC
and NGCC capture units drops by half. Conventional gas-fueled technologies in both scenarios
hovers around 30 percent.
82
-------
Conventional
Retrofit CCS
New CCS
Renewable
Gas
Coal
Nuclear
2000
2010
2020
2030
Figure 6.2: Electricity generation per-period by technology class for the 50 percent
emissions scenario (electric sector COi emissions held constant at 50 percent of 1995 levels
from 2015 on).
1.20
2000
2005
2010
2015
2020
2025
2030
•Electric SectorNOX
•Electric Sector SO2
Figure 6.3: Electric sector SOi, and NOx emissions as a function of time for the 50 percent
emissions scenario (electric sector COi emissions held constant at 50 percent of 1995 levels
from 2015 on).
83
-------
These results show that both nuclear and gas-fired generation dominate CCS, given baseline
technology assumptions. Figure 6.4 illustrates how generation by these technologies changes as
a function of assumptions about CCS costs and efficiencies, nuclear capital costs, and electric
sector natural gas prices. (Except where noted, electric sector CO2 emissions level out at 1995
level in 2015, and only conventional nuclear technologies are available.) Note that the BAU
wind constraints remain in place (and are preventing additional investment in wind turbines); this
assumption will be relaxed in the following set of scenarios.
PH 0.00
Better CCS Better CCS + Better CCS + Better CCS + Better CCS +
50%CO2 50%CO2+ 2xNucInv 2xNucInv +
NGA MU 50% CO2 +
NGAMU
in Renewable E Coal (No CCS) S Gas (No CCS) 0 CCS B Nuclear
Figure 6.4: Fraction of 2030 electric power generation by technology class for five
scenarios.
The Figure 6.4 scenarios attempt to improve the competitiveness of CCS relative to conventional
nuclear power. The first scenario ("Better CCS") lowers all CCS costs and input fuel
requirements (the inverse of efficiency) by 20 percent. Nuclear power's share of 2030
generation falls from 33 to 26 percent, while that of CCS increases from 4 to 11 percent - a mix
of IGCC and NGCC units with CC>2 capture. CCS growth constraints bind for only two periods
(2015 and 2020). Since the output from existing coal plants does not change, criteria pollutants
remain at their BAU levels (i.e., the limits of their regulatory constraints).
The second scenario ("Better CCS + 50% CO2") examines the same CCS technology
improvements under the 50 percent CC>2 trajectory. CCS maintains its higher share of output,
but nuclear returns to its previous (limited) 33 percent share at the expense of existing coal-fired
generation. The model also adds non-capture IGCC capacity, which contributes 12 percent of
2030 generation. The result is a significant decrease in electric sector SC>2 and NOx emissions,
which decline to 18 and 34 percent, respectively, of their 2030 BAU values. As this scenario
also sees an increase of (non-CCS) gas-fueled generation from 21 to 26 percent, it is worth
looking at the effects of a gas price increase. The third scenario ("Better CCS + 50% CO2 +
84
-------
NGA MU") adds 10 $/GJ to electric sector gas prices. The price increase lowers the output of
the model's gas capacity to its previous value; IGCC and NGCC units with CC>2 capture make up
the difference (nuclear power continues to hit its growth constraint). Existing coal plants reduce
their output further and non-capture IGCC increases its output to 19 percent, yielding an 88
percent reduction in 2030 862 emissions and a 48 percent drop in NOx from the electric sector.
The remaining two scenarios in Figure 6.4 attempt to advantage CCS further by increasing
nuclear investment costs. Nuclear construction costs have been notoriously difficult to predict,
and total project costs of nearly twice that of early estimates, for instance, mark the historical
record (EIA, 1986). The fourth scenario ("Better CCS + 2x Nuc Inv") is equivalent to the first,
with all (conventional) nuclear investment costs doubled. The results are similar, except that
nuclear power's share of 2030 generation drops to 15 percent, while that of CCS increased to
nearly 20 (with growth constraints binding through 2025); since existing coal capacity does not
change, criteria pollutant emissions remain at BAU levels. The last scenario ("Better CCS + 2x
Nuc Inv + 50% CO2 + NGA MU") adds the natural gas price markup and looks at technology
penetration assuming the lower electric sector CO2 trajectory. CCS maintains its higher output,
but nuclear generation increases to 24 percent; non-capture IGCC provides 17 percent of 2030
output and existing coal units once again drop below 5 percent. These dynamics result in the
lowest electric SC>2 emissions of any scenario (8 percent of BAU levels) and a NOx reduction to
approximately two-thirds of the modeled CAA limits.
Finally, these scenarios pitting nuclear and CCS are somewhat unrealistic in that they maintain
tight growth constraints on renewables. Wind generation is becoming economically competitive
with gas as natural gas prices remain at historically high levels and wind turbine technology
continues to improve. Wind therefore deserves consideration in this analysis. While nuclear and
CCS can be dispatched to supply baseload electricity, wind is an intermittent source of power
that tends to displace load-following units such as gas turbines. The EPANMD also lacks the
regional specificity needed to capture important differences in the availability of wind resources
(i.e., regional wind classes, which characterize typical wind speeds). In the EPANMD, wind is
simply available anywhere, subject to the assumed growth and availability constraints. A more
accurate representation would include a resource supply curve to reflect the increasing cost of
deploying wind turbines in less-suitable regions of the country. The same argument, of course,
holds for solar and other renewable technologies.
These caveats should be kept in mind while considering the final set of scenarios, which look at
the competition between wind generation, nuclear power, and CCS under the relaxed wind
growth constraints (these are equivalent to the Section 4 nuclear growth constraints). Figure 6.5
summarizes 2030 generation by technology class; all scenarios adopt the 50 percent of BAU CC>2
trajectory to stimulate alternative technology penetration.
85
-------
C l.UU
0
'•3
*-H r\ orv
u U.oU
S
w
O
o 0.60 -
CO
o
CN
<£ 0.40 -
s
o
•^H /\ <-)/\
•g 0.20 -
S
u.
n nn
••
W
^
/Si li
^ ii
w«
^
v
^
Conv Nuc
ConvNuc+NGA
MU + 2xNuc Inv
Adv Nuc + NGA
MU + 2xNuc Inv
Conv Nuc + NGA
MU + 2xNuc Inv +
Better CCS
OH Wind d Coal (No CCS) & Gas (No CCS) 0 CCS • Nuclear
Figure 6.5: Fraction of 2030 electric power generation by technology class for four
scenarios with the relaxed wind growth constraints.
The first relaxed wind scenario ("Conv Nuc") restricts nuclear investment to conventional
technologies. Wind and nuclear each contribute around 30 percent of 2030 generation, with gas
responsible for an additional 25 percent and existing coal units making up the difference (Figure
6.6). Electric sector 862 emissions fall to 10 percent of their constrained values, while NOx
declines to 75 percent (the increase in gas-fired generation largely explains this pattern). CCS
plays a trivial role. An increase in gas prices coupled with higher nuclear investment costs
("Conv Nuc + NGA MU + 2x Nuc Inv") increases wind's share of generation to 45 percent,
decreases nuclear output to 15 percent, and results in the growth of CCS to 12 percent by 2030.
Existing coal plants (18 percent) and gas (10 percent) contribute the rest. NOx emissions remain
at their constrained limits, while SC>2 falls to 40 percent of its upper bound. Opening nuclear
investment to advanced technologies, while maintaining the gas and nuclear investment cost
mark-ups, increases nuclear output to over 25 percent ("Adv Nuc + NGA MU + 2x Nuc Inv").
CCS once again plays a trivial role in meeting electric power demand. Electric sector criteria
pollutants remain essentially unchanged. Finally, improvements in CCS costs and efficiencies
("Conv Nuc + NGA MU + 2x Nuc Inv + Better CCS") lower gas-fueled generation somewhat
when only conventional nuclear investment is allowed, at the expense of higher electric sector
SO2 emissions (approximately two-thirds of their constrained value).
86
-------
fin
O
O
O
_
20
15
I 10
§
O
2°000
Conventional
Retrofit CCS
New CCS
Renewable
Gas
Coal
Nuclear
2010
2020
2030
Figure 6.6: Electricity generation per-period by technology class for the 50 percent
emissions scenario (electric sector COi emissions held constant at 50 percent of 1995 levels
from 2015 on) under the relaxed wind growth constraints with nuclear investment
restricted to conventional technologies (the "Conv Nuc" scenario in Figure 6.5).
These results provide an indication of how nuclear power, wind generation, and CCS might fare
against each other. The last set of scenarios, as noted, is tentative pending improvements in the
U.S. EPA's National MARKAL modeling framework. The results, however, suggest that wind
and nuclear together could displace a significant portion (60 percent) of fossil-fuel electric
generation, though the co-control emissions benefits of this substitution may not be that
substantial and depend largely on how existing coal plants are used. An important caveat is
whether and how system operators of electricity grids could manage the intermittent output of
wind serving such a large fraction of demand. Future work on the EPANMD includes a
representation of backup generation to complement intermittent resources.
In conclusion, several key results emerge from this section's integrated technology assessment:
• The penetration of advanced technologies does not necessarily yield a significant criteria
pollutant benefit. Under all but the most radical departures from BAU electric sector
carbon trajectories, for instance, nuclear power and CCS merely replace the new coal and
gas capacity that are needed to meet increasing electricity demand. The model maintains
its existing coal plants through their useful lifetime, and criteria pollutant emissions
remain near their Clean Air Act limits as a result. These existing plants, which are free of
amortized investment costs, remain too economical to retire as baseload capacity—an
observed trend noted in the literature (e.g., Ellerman, 1996). This picture changes when
electric sector carbon emissions depart furthest from their BAU levels and CC>2 emissions
87
-------
from baseload plants declines. Even in these cases, however, the decrease in NOx
emissions is more modest than that seen in SC>2.
CCS is dominated by investment in new nuclear technologies as well as additional gas-
fired generation and wind power. Only when favorable CCS cost and efficiency
assumption combine with high nuclear investment cost and gas price scenarios does CCS
approach its modeled growth constraints over the full time horizon in which it is available
(2015-2030). Furthermore, given that CO2 capture and geologic sequestration are both
untested at even the modest scales adopted here, these results are probably optimistic.
Even with a rudimentary conception of wind resources, wind generation in the EPANMD
competes for a significant share of electric power output when the model incorporates
more favorable growth constraint assumptions for new turbine investment. These growth
limits are modest compared to some predictions, though a more realistic representation of
unit dispatch is needed to capture the technology's intermittency (beyond lowering the
annual availability of wind turbines, as is done here) and evaluate its actual potential.
88
-------
Section 7
Limitations, Future Model Development, and Analysis Directions
While the model results provide insight into future energy technology pathways, they are
nonetheless based on a model, and all models have limitations that introduce caveats. This
section identifies and discusses some of the limitations associated with the EPANMD and
MARKAL model. Planned strategies and future work to address these limitations are interwoven
in the discussion. The report concludes with next steps towards fulfillment of ISA-W's role in
EPA's Air Quality Assessment.
The use of national rather than regional inputs to MARKAL is perhaps the most significant
limitation affecting the analysis. The EPANMD is a national database that does not contain
region-specific data. The lack of regional data manifests itself in three key limitations. First,
important regional differences in resource supplies, energy service demands, and technology
availability cannot be represented. An example of where this limitation would affect model
results is in an evaluation of renewables, such as biomass and wind. In the EPANMD,
renewable resources are represented by aggregated national supply curves, if at all. For example,
biomass resources are represented by a single 7-step supply curve that aggregates biomass
resources across the nation. Wind does not have a supply curve, but rather an assumed capacity
factor (which implies the strength of the wind resource) of 35 percent. These representations do
not account for widely varying resource quantities and costs from one region to another. To
address regional differences explicitly, ISA-W is developing a nine-region MARKAL database
(EPA9R) that will account for variations in supply, demand, and technologies between the nine
U.S. Census regions. We expect to use the EPA9R database to examine renewables in the future.
Second, the EPANMD does not include transportation costs for coal or other resources. As with
renewables, transportation costs will affect regional technology penetration, and by
extrapolation, the aggregate technology penetration at the national level. In particular, the
inclusion of transportation costs will have the greatest effect on energy resources with low
energy density, such as biomass or low grade coals. It is not cost-effective to ship these resources
on transnational scales, but such transport may be occurring implicitly within EPANMD
modeling. Transportation costs for energy resources will be included in EPA9R.
Third, air pollution regulations that have been implemented on a state or regional level must be
modeled at a national level within EPANMD. Air pollution is a problem that has multiple scales.
Damages from air pollution occur at the local scale, affecting human health, agriculture, and
ecosystems. Some of the pollutant emissions contributing to air pollution are from local sources,
but others typically are emitted from upwind sources both within and outside the state. In this
context, federal air quality regulations, including the Clean Air Act (CAA) and the Clean Air
Interstate Rule (CAIR), contain provisions that reduce emissions at the local and regional scales.
For example, the CAA introduces national ambient air quality standards that must be met at the
local level. States must develop plans for reducing emissions to these standards, so these plans
typically result in local- or state-level emissions reductions. To address regional- and national-
scale transport of emissions, the CAA also place NOx and 862 emissions limits on individual
boilers. Utilities are given the option of trading emissions credits so that the emissions reductions
89
-------
occur in aggregate, but, through purchasing emissions credits, individual boilers may exceed the
limit. CAIR was also developed to reduce the transport of pollutants from one state to another. In
CAIR, states within the eastern U.S. are assigned budgets that account for the effects of their
emissions on downwind states within the region. States have considerable freedom in
determining how to comply with their budgets.
The regional- and state-level complexities of these regulations complicate modeling emissions
regulations within EPANMD. Resulting projections for the use of emissions controls can be
quite different from what is seen "on the ground." As a result of CAIR, for example, most coal-
fired boilers east of the Mississippi are expected to use SCR controls for NOx. By not
representing the regional requirements of CAIR, however, EPANMD favors SNCR, which has
lower costs but also lower NOX removal efficiencies. A state or regional representation of control
requirements would produce results that are closer to what will actually be implemented. With
the development of EPA9R, ISA-W expects to be better able to represent regional-level
implementation of national rules. State-level differences will still be difficult to capture within
EPA9R.
Other limitations are unrelated to the lack of regional specificity in the EPANMD. First,
MARKAL is not an electric dispatch model: the user is limited to exactly six time slices in which
to specify all demands (see Shay, et al. 2006). The time slices represent two diurnal time periods
(day and night) as well as three seasons (summer, winter, and intermediate); the 6 time slices
come from the combination of times of day and seasons enumerated above. The specification of
demands according to these time slices results in a flat demand profile compared with the more
conventional load duration curve used to specify demand in electric dispatch models. The result
is that MARKAL tends to favor baseload units with high capital costs and low marginal costs
(e.g., coal and nuclear) over units with higher marginal costs that are better suited to meeting
peak and shoulder demand (e.g., gas turbines and wind). ISA-W is addressing this MARKAL
limitation by funding the development of flexible time slices, so the user can specify as many
time slices as desired to more accurately represent the profile of electricity load. The expanded
time slices will be included in the EPA9R model.
The EPANMD's simple representation of the nation's existing coal-fired generating
infrastructure is an additional limitation. The database contains just three types of pre-existing
coal plants: bituminous, sub-bituminous, and lignite steam. In reality, a diverse variety of coal
plants exist, with performance and emissions that are highly dependent on their vintage and
detailed characteristics of the coal used to supply the plants. Future development will include an
expanded representation these existing technologies.
Finally, the analysis excludes consideration of efficiency improvements as well as non-
traditional generating technologies, like combined heat and power, which could supply a large
share of the nation's electricity in the coming decades. End-use efficiency improvements, in
particular, are promising as they could significantly reduce electricity demand by 2030 (Laitner,
et al. 2005). Ignoring the potential for a significant demand reduction, as this analysis does,
biases conclusions in favor of advanced generating technologies like nuclear power and CCS
under certain scenarios. The non-BAU carbon trajectories, for instance, through their impact on
the cost of electricity, would likely spur as much investment in end-use energy efficiency
90
-------
measures as they do in alternative supply-side technologies. Likewise, the analysis does not
consider the impact of radical technological innovation, such as the emergence of "nano-bio-
info" technologies, on future energy demand (see Laitner, 2006, for an overview). Surprises in
principle are difficult to anticipate, and this assessment makes no attempt to cover all scenarios
that may play out in the future.
Future Work
With the transportation and electricity sector assessments complete, work will involve using the
EPANMD within a variety of applications. One such application is the development of future-
year emissions scenarios for the EPA Office of Research and Development's Global Change Air
Quality Assessment. These scenarios will be used to project criteria pollutant emissions out to
the year 2050. The emissions will then be used within an air quality model to characterize the
impacts of increased energy service demands and technological change on regional air quality.
ISA-W also plans to use the EPANMD to assess the prospects and implications of particular
technologies. Using formal sensitivity analysis techniques, MARKAL can be used to identify the
conditions in which a particular technology is competitive, the sectoral and cross-sector fuel
implications of the penetration of the technology, and the resulting emissions. An example of
such an analysis is the recently completed evaluation of hydrogen fuel cell vehicles (Yeh et al.
2006). Additional technologies that we expect to assess with the EPANMD include advanced
nuclear power, coal gasification, and plug-in gasoline-electric hybrids.
Another potential application for the EPANMD is in the area of risk assessment and risk
management. Within a probabilistic framework, MARKAL can be used to identify the
conditions that lead to poor air quality or depletion of specific fuel resources. The model can
then be used to identify technological pathways to minimize these risks.
Through these and other applications, we expect EPANMD, and later, EPA9R, to play an
important role in EPA's efforts to understand the linkage between energy and emissions and use
this understanding to protect the environment and human health into the future.
91
-------
Section 8
References
Anderson, S. and Newell, R. (2004). "Prospects for Carbon Capture and Storage Technologies."
Annual Review of Environment and Resources 29:109-42.
Ansolabehere, S., J. Deutch, et al. (2003). The Future of Nuclear Power. Cambridge, MA, MIT.
BEA (U.S. Bureau of Economic Analysis) (2006). http://www.bea.gov/bea/dn/gdplev.xls
(accessed July 2006).
Boilanger, F. (2002). National Energy Technology Laboratory, U.S. DOE, personal
communication.
Cullen, A. C. and H. C. Frey (1999). Probabilistic techniques in exposure assessment: A
handbook for dealing with variability and uncertainty in models and inputs. Plenum Publishing
Corp., New York, NY.
DOE (1999). Supporting Analysis for the Comprehensive Electricity Competition Act, U.S.
Department of Energy, DOE/PO-0059.
DOE (2001). A Roadmap to Deploy New Nuclear Power Plants in the United States by 2010, US
Department of Energy: Office of Nuclear Energy, Science, and Technology: 1-309.
Edison Electric Institute (2006a). "History of the Electric Power Industry,"
http://www.eei.org/industry_issues/industry_overview_and_statistics/hi story/index.htm
(accessed 12 July 2006).
Edison Electric Institute (2006b). Map: "Different Regions of the Country Use Different Fuel
Mixes to Generate Electricity",
http://www.eei.org/industry_issues/energy_infrastructure/fuel_diversity/diversity _map.pdf,
(accessed 30 Aug. 2006).
EIA (Energy Information Administration), Office of Integrated Analysis and Forecasting, U.S.
Department of Energy (1986). An Analysis of Nuclear Power Plant Construction Costs.
DOE/EIA-0485. Washington, DC: U.S. Government Printing Office.
EIA (2002a). Annual Energy Outlook 2002 with Projections to 2020, Energy Information
Administration, Department of Energy, DOE/EIA-0383(2002).
EIA (2002b). Assumptions to the Annual Energy Outlook 2002, Energy Information
Administration, Department of Energy, DOE/EIA-0554(2002).
EIA (2003). Annual Energy Review 2002. DOE/EIA-03 84(2002). Washington, D.C., Department
of Energy.
92
-------
EIA (2004). Assumptions to the Annual Energy Outlook 2004 With Projections to 2025.
DOE/EIA-0554(2004).
EIA (2005a). Annual Energy Outlook 2005 With Projections to 2025. DOE/EIA-0383(2005).
EIA (2005b). Annual Energy Review 2004. DOE/EIA-03 84(2004). Washington, D.C.,
Department of Energy.
EIA (2006a). Annual Energy Outlook 2006 With Projections to 2030. DOE/EIA-03 83 (2006).
EIA (2006b). Assumptions to the Annual Energy Outlook 2006 With Projections to 2030.
DOE/EIA-0554(2006).
EIA (2006c). International Energy Outlook 2006. DOE/EIA-0484 (2006).
Ellerman, A. D. (1996). The Competition Between Coal and Natural Gas: The Importance of
Sunk Costs. Resources Policy 22, 33-42.
EPA (2003). Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2001.,
Environmental Protection Agency [FRL-7449-1].
EPA (2005). AP 42, Fifth Edition Compilation of Air Pollutant Emission Factors, Volume 1:
Stationary Point and Area Sources, (available at http://www.epa.gov/ttn/chief/ap42/).
EPRI (1993). TAG Technical Assessment Guide, Vol. 2: Electricity End Use, Pt. 2: Commercial
Electricity Use, Electric Power Research Institute, Palo Alto, CA.
EPRI (2005). "Success Story: Engineering Economic Studies in Support of AEP's Strategic Plan
for Proposed IGCC Plant."
http://www.epri.com/targetHigh.asp?program=270683&value=05T066.0&objid=298291
(accessed 16 May 2006).
ETSAP (2004). Documentation for the MARKAL Family of Models. The Energy Technology
Systems Analysis Programme (ETSAP). (available at http:///www.etsap.org/ MrklDoc-
I_StdMARKAL.pdf).
FutureGen Industrial Alliance (2006). http://www.futuregenalliance.org/about.stm (accessed 16
May 2006).
Gage, C.L., T. Johnson, E. Wright, D. Loughlin, C. Shay, R. Spiegel, and L. Beck (2004).
"Demonstration of a Scenario Approach for Technology Assessment: Transportation Sector,"
EPA-600/R-04/135.
93
-------
IEA (International Energy Agency) (2004). Prospects for C02 Capture and Storage. Paris:
OECD/IEA.
IPCC (Intergovernmental Panel on Climate Change) (2005). Carbon Dioxide Capture and
Storage (IPCC Special Report). Metz, B., Davidson, O., de Coninck, H., Loos, M., and Meyer,
L. (Eds.). NY: Cambridge University Press.
Johnson, T.L. and D.W. Keith (2004). "Fossil electricity and CO2 sequestration: How natural gas
prices, initial conditions and retrofits determine the cost of controlling CC>2 emissions." Energy
Policy 32:367-382.
Laitner, J.A. (2006). "How Far Energy Efficiency? And, why current forecasting tools and
scenario exercises get us only part of the way there. . ." 2006 ACEEE Summer Study, Asilomar,
CA (August 2006).
Laitner, J.A., Hanson, D.A., Mintzer, I. and Leonard, J.A. (2005). "Adapting for Uncertainty: A
Scenario Analysis of U.S. Technology Energy Futures." Energy Studies Review: 14(1): 120-135.
NEI (2003). Nuclear Energy Institute, (available at http://www.nei.org).
NETL (National Energy Technology Laboratory, Office of Fossil Energy, U.S. Department of
Energy) (2005). Carbon Sequestration Technology Roadmap and Program Plan 2005:
Developing the Technology Base and Infrastructure to Enable Sequestration as a Greenhouse
Gas Mitigation Option.
http://www.fossil.energy.gov/sequestration/publications/programplans/2005/sequestration_road
map_2005.pdf (accessed 21 August 2006).
OAR (Office of Air and Radiation), U.S. Environmental Protection Agency (2002).
Documentation of EPA Modeling Applications (V.2.1) Using the Integrated Planning Model.
EPA 430/R-02-004.
OECD/IAEA (2002). Uranium 2001: Resources, Production and Demand. 19th Edition.
Palmgren, C., Morgan, M., Bruine de Bruin, W., and Keith, D. (2004). "Initial Public
Perceptions of Deep Geological and Oceanic Disposal of Carbon Dioxide." Environmental
Science and Technology 38:6441-6450.
Schwartz, Peter (1996). The Art of the Long View: Planning for the Future in an Uncertain
World. New York: Doubleday.
Saltelli, A., Chan, K., and E. M. Scott (Eds.) (2000). Sensitivity Analysis. NY: John Wiley and
Sons.
Shay, C. L., J. Decarolis, D. Loughlin, C. L. Gage, S. Yeh, S. Vijay, and E. L. Wright (2006).
EPA U.S. National MARKAL Database: Database Documentation. EPA/600/R-06/057.
94
-------
Simbeck, D.R. and M. McDonald (2001). "Existing coal power plant retrofit CC>2 control options
analysis." In: Williams, D.J., Durie, R.A., McMullan, P., Paulson, C.A.J. and Smith, A.Y. (Eds.),
Proceedings of the 5th International Conference on Greenhouse Gas Control Technologies,
Collingwood, Australia: CSIRO Publishing, pp. 103-108.
Tseng, P. (2001) U.S. Department of Energy, personal communication.
U.S. Census Bureau (2006). "Table la. Projected Population of the United States by Race and
Hispanic Origin: 2000 - 2050," http://www.census.gov/ipc/www/usinterimproj/ (accessed 13
July 2006).
U.S. EPA (2005). National Emissions Inventory: Air Pollutant Emissions Trend Data,
http://www.epa.gov/ttn/chief/trends/index.html (accessed July 2006).
U.S. EPA (2006a). Inventory of US Greenhouse Gas Emissions and Sinks, USEPA #430-R-06-
002.
http://yosemite.epa.gov/oar/globalwarming.nsf/content/ResourceCenterPublicationsGHGEmissio
nsUSEmissionsInventory2006.html (accessed July 2006).
U.S. EPA (2006b). EPAs Road Map for Mercury, EPA #HQ-OPPT-2005-0013.
http://www.epa.gov/mercury/pdfs/I_HgReleases.pdf (accessed August 2006).
Wilson, E.J., T.L. Johnson and D.W. Keith (2003). "Regulating the ultimate sink: Managing the
risks of geologic CC>2 sequestration." Environmental Science and Technology 37:3476-3483.
Wilson, E.J. (2004). Managing the Risks of Geologic Carbon Sequestration: A Regulatory and
Legal Analysis. Ph.D. Thesis, Department of Engineering and Public Policy, Carnegie Mellon
University, Pittsburgh, PA.
Yeh, S., D. H. Loughlin, C. L. Shay, and C. Gage. (2006). "An integrated hydrogen scenario
analysis: Impacts on transportation, energy, and air emissions." Accepted for publication in
Proceeding of the IEEE Special Issue: Hydrogen Economy.
95
------- |