You are on page 1of 10

Environmental Modeling and Assessment 7: 207–216, 2002.

 2002 Kluwer Academic Publishers. Printed in the Netherlands.

NEMS and MARKAL-MACRO models for


energy-environmental-economic analysis: A comparison
of the electricity and carbon reduction projections
S.C. Morris a G.A. Goldstein b V.M. Fthenakis c
a Energy and Environmental Consultant Port Jefferson, NY 11777, USA
b International Resources Group Washington, DC, USA
c Department of Environmental Sciences, Brookhaven National Laboratory, Upton, NY 11973, USA

The Annual Energy Outlook forecasts published by the United States Energy Information Administration (EIA) of the Department
of Energy are based on results from the National Energy Modeling system (NEMS). This paper compares NEMS, which is used only in
the U.S., with the U.S. version of MARKAL-MACRO (USMM) model, which is used in more than thirty-five countries. The two models
predict similar results for the base 1999 US Annual Energy Outlook (AEO), but their results with carbon constraints are quite different. The
differences of the models and those of their predictions are explained. USMM can be used to provide an alternative and complementary
approach to projections of renewable technologies penetration and their potential in reducing carbon dioxide emissions in the USA.

1. Introduction 2. Characterization of the models

Following are brief summaries of the two models. More


There are several models in use for integrated energy- details are provided in tables 1, 2 and 3.
environmental-economic analyses, but few direct compar-
isons among them. Comparisons are difficult because mod- 2.1. NEMS
els differ in structure, in character, in their input assump-
tions, in the level of detail they include, in their solving The National Energy Modeling system (NEMS) was ini-
mechanisms, and other factors. The form of inputs to differ- tiated in 1993 to develop the forecasts for the U.S. Energy In-
ent models varies. While some models may require a para- formation Administration’s Annual Energy Outlook (AEO).
meter as input, others may generate that parameter endoge- The purpose of NEMS is “. . . to project the energy, eco-
nomic, environmental, and security impacts on the United
nously. Some models use discrete data as input, others make
States of alternative energy policies and of different assump-
extensive use of pre-constructed curves as inputs. Often the
tions about energy markets”. NEMS consists of 12 interac-
inputs to one model must be adjusted in some judgmental
tive modules and an integrating module. The various mod-
way to fit the needs of another model. ules operate differently; the electricity module operates on
This paper compares The United States version of a cost minimization premise. NEMS analyzes the U.S. on
MARKAL-MACRO (USMM) with the National Energy a regional basis. The number of regions varies by energy
Modeling system (NEMS). NEMS is used by the Energy sector. The electric sector includes 15 regions [5–7].
Information Administration (EIA) of the Department of En-
ergy to produce the Annual Energy Outlook (AEO) for the 2.2. USMM
United States. About 50 developed and developing countries
and smaller regions use USMM. In this paper we focus on USMM is an outgrowth of MARKAL, which was de-
the electric sector; however, the entire energy system was veloped as an energy systems model in the 1976–81 pe-
modeled. We compare two cases. One is the 1999 Annual riod in an effort involving analysts from 17 nations and
two international organizations. The goal was to provide
Energy Outlook (AEO) reference case. The other is a case
an open architecture that would allow nations the ability to
with carbon emissions constrained to seven percent below
include a wide variety of energy technologies and policies.
1990 levels [5]. The motivation for this comparison was to
The model is currently used for energy planning in over 35
provide an alternative and complementary approach to pro- countries. MARKAL is a demand-driven, multi-period, lin-
jections of renewable technologies. Furthermore, USMM ear programming model that minimizes total energy system
can be used for smaller geographical areas and it is easier cost [8]. In 1989–91, MARKAL was augmented with a
to add new technologies than NEMS, and therefore, can be macroeconomic component and regional capabilities, creat-
used to study the effect of, e.g., a step-up capacity of photo- ing a new energy-environmental-economic model, USMM.
voltaics in a given state or region. USMM is a non-linear program maximizing consumer util-
208 S.C. Morris et al. / A comparison of the electricity and carbon reduction projections

Table 1
Comparisons of various model characteristics.

NEMS USMM

Load segments for electricity demand


3 seasons, daytime, morning-evening and night, by region. 3 seasons, day and night (no regionalization in electric sector).
Reserve margin
13% for MAPP, STV and SPP; 9% for FL, 15% for NWP; for other regions, USMM calculates the reserve capacity as the amount by which the in-
NEMS determines the reserve margin by equating the marginal cost of stalled capacity exceeds the average demand in the time division with the
capacity and the cost of unserved energy. maximum demand. In the U.S. model that season is summer day and the
value is 0.5.
Plant retirements
Retires fossil plants when expected revenues are not sufficient to cover Assigns a “useful life” to all plants. Plants can be shut down before the
annual going forward costs. Retirement decisions are made nuclear plants end of their useful life, but fixed maintenance costs are still paid. The life
at 10 year intervals base on capital costs required to continue operation. of fossil plants can be extended at an additional capital cost.
Demand-side management
Energy efficiency improvements induced by rising energy prices, new ap- Competes demand side conservation and efficiency improvements directly
pliance standards, and utility demand-side management are represented in with electric power generation on marginal cost basis.
the end-use modules. Appliance decisions are based on a function of rela-
tive costs and performance characteristics.
Fuel price expectations
Fuel prices are derived using adaptive expectations; future fuel prices are Calculates fuel prices as the shadow prices of the cost of obtaining the
extrapolated from recent historical trends. fuels. These costs are based on classes of each fuel, the total amount avail-
able in the class, the unit cost of extraction, and the amount that can be
obtained annually. World oil prices are taken from the AEO.
Technology learning and optimism

Table 2
Additional model characteristics.
NEMS USMM

1-year periods 5-year periods


Myopic Perfect foresight
Solves year by year Solves as global optimum
Develops energy demands Draws demands from other sources (e.g., NEMS) and modifies based on
price elasticity
Capacity planning model uses linear programming to achieve least cost Entire energy system is solved in non-linear programming with the ob-
solution. Other parts of the electricity market module use process models. jective function to maximize consumer utility. This drives the technology
sectors to least cost.
NEMS is designed for forecasting USMM is designed for exploring possibilities
NEMS is regionalized with different number of regions for different mod- USMM is not regionalized, with the exception of residential heating and
ules cooling demands

ity [11,12]. USMM creates a competition between invest- 3. Approach for comparing the two models
ment in energy technology with investment in the rest of the
economy, which places demands for energy services. The The basic approach was to apply the input assumptions
result is an energy sector that is driven to a least-cost invest- of the AEO1999 in USMM to the extent possible and to
ment strategy. In 1998 regionalization and an endogenous compare the USMM results with those in the AEO (which
technology learning capability was added to MARKAL [13]. were produced by NEMS). As an additional comparison, we
then constrained USMM to limit carbon emissions to 1249
In 1999, a stochastic capability was added. The current
million metric tons per year from the 2010 period through
USMM model of the United States addresses building heat-
the 2020 period. This corresponds with the “1990 –7%”
ing and cooling by region, but currently models electricity in
case in [2]. The 1999 AEO reference case is referred to as
the U.S. as a single region. A new variant of the multi-region
“the reference case” while the corresponding USMM case is
version of MARKAL which incorporated elastic demands called the “base case”.
(where demand levels are responsive to own-price elastici-
ties) has been selected by the US Department of Energy – 3.1. Setting up USMM for the comparison
Energy Information Administration as the analytic frame-
work that will be behind the annual publication of the In- We calibrated USMM to the Energy Information Admin-
ternational Energy Outlook beginning with the 2002 report. istration’s (EIA) Annual Energy Outlook (AEO) to the extent
S.C. Morris et al. / A comparison of the electricity and carbon reduction projections 209

Table 3
Data on NEMS and USMM (EIA Information from EIA [6]).
NEMS USMM

Residential demand
14 end-use services 20 end-use services
3 housing types 2 housing types
34 end-use technologies 83 end-use technologies
Commercial demand
10 end-use services 5 end-use services
11 building types 1 building type
10 distributed generation technologies 2 distributed generation technologies
64 end-use technologies 54 end-use technologies
Industrial demand
7 energy intensive industries 3 energy intensive industries
8 non-intensive industries 15 technologies characterizing other technologies and
6 energy conservation and efficiency technologies
cogeneration 4 co-generation technologies
Transportation demand
6 car sizes 4 car sizes
6 light truck sizes 1 light truck size
59 fuel saving technologies for light-duty vehicles 3 mpg levels
15 fuels for light-duty vehicles 6 fuels for light duty vehicles
20 vintages for light-duty vehicles 3 vintages
8 types of aircraft 2 types of aircraft
12 types of freight trucks 1 type heavy truck
1 type bus
6 types of rail
1 ship type
Electricity
11 fossil technologies 24 fossil technologies
7 renewable technologies 16 renewable technologies
2 nuclear technologies 5 nuclear technologies
Generation capacity expansion Generation capacity expansion
Oil supply
4 oil supply technology types 8 oil supply types
Natural gas supply
7 natural gas supply technology types 7 gas supply types
Natural gas distribution
core vs. non-core
peak vs. off peak
pipeline capacity expansion
Refining
5 crude oil categories 13 crude oil categories
7 product categories
35 refining technologies 3 refinery technologies
refinery capacity expansion refinery capacity expansion
Coal supply
3 sulfur categories 4 sulfur categories
4 thermal categories 5 thermal categories
underground and surface mines

of applying world oil price, initial GDP growth rates, and to new technology and a lower cost for subsequent units in-
some extent matching the levels of demand for energy ser- stalled. In transposing capital cost data into USMM, the first
vices. The latter was necessary because the categories are value was applied to the five-year period the technology be-
not the same for the two models. For example, in the U.S. came available and the second value to all subsequent peri-
version of USMM, we express the energy service for auto- ods.
mobiles in vehicle-miles traveled rather than energy units. Since resource, manufacturing, construction and man-
We drew technology costs, efficiencies, heat rates, etc. from power capabilities impose high bounds to the increase of the
the Assumptions to the Annual Energy Outlook [2], and capacity of any technology, we use two methods in USMM
from the supplemental tables to the Annual Energy Out- to control technology growth. One is simply to put an upper
look 1999 [3]. We used these data as inputs to USMM bound on the technology by period. The model can choose
and compared our results with the results in AEO 1999. to invest in the technology from zero to the bound. An al-
For capital costs, the Assumptions to the Annual Energy ternative approach is to specify the maximum percent per
Outlook 1999 presents values for the first 5 units of a year allowable growth. The advantage of the direct bound
210 S.C. Morris et al. / A comparison of the electricity and carbon reduction projections

over the growth constraint is that it provides a “reduced cost” capacity, carbon emissions and price of electricity). We con-
for the technology if the technology hits an upper (or lower) centrate on the explanation of the differences, which is prob-
bound. This allows the analyst to immediately see if a bound ably more important than the differences themselves. We
has been hit and, if so, how much the model is willing to pay expected that greater variations would occur in new tech-
to increase the bound by one unit. We used both growth con- nologies with small capacities.
straints and bounds in this exercise.
Bounds were also set on environmental emissions to as- 4.1. Comparison of USMM with NEMS for the AEO1999
sure the model meets regulatory requirements for the Clean reference case
Air Act or, in the case of carbon emissions, to explore the
impact of different proposed emission rates. In this case the In this section we provide an overview of the core results
bound creates a shadow price on carbon emissions. from both models and hypotheses and explanations of the
differences are further explained.
3.2. Carbon emissions reduction For coal-fired steam plants, USMM shows a small but
steady decrease over time (−0.33%/y), while NEMS shows
The most recent projections are that U.S. carbon emis- a small but steady increase (0.44%/y) (table 4 and figure 1).
sions will be 1809 million metric tons in 2010 and 2041 mil- The difference in 2020 is 17%. USMM assigns a lifetime
lion metric tons in 2020 [6]. The seven percent below 1990 to each technology. At the end of that lifetime the technol-
levels means carbon emissions must be reduced to 1249 mil- ogy must be replaced. In reality, large electric power plants
lion metric tons/year in 2010 and remain at that level (or do not disappear after 30 or 40 years, they get upgraded or
below) through 2020. That is a 30% reduction from pro- re-powered. USMM includes an upgrading technology that
jected 2010 emissions and a 39% reduction over the pro- extends a plant’s life at a much lower cost than replacing it.
jected emissions in 2020. The model invested substantially in this technology, but not
Carbon reductions were produced in USMM by plac- enough to match the AEO projection. Since the predicted
ing a constraint (upper bound) on carbon emissions directly. GDP growth is similar in the two models, the most likely
The shadow price on the carbon emission bound represents explanation for the different results is the difference in the
the marginal cost of reducing carbon emissions to the con- electricity supply projections. USMM projects electricity
strained level. This value can also be interpreted as the car- to supply 13 to 14% of final energy to energy users, while
bon tax that would be needed to achieve the same level of NEMS projects 15 to 16%.
emissions. The reported emissions and emissions reduction For other fossil steam plants (oil and gas) both models
from both models are from all energy use, not just the elec- show a steady decrease (−2.6 for USMM and 3.0%/y for
tricity sector. NEMS) (table 4 and figure 2). Both models show an in-
crease in capacity for combined cycle plants and combus-
3.3. Endogenous technology learning
Table 4
Projected electricity generating capacity: MARKAL-MACRO base case
NEMS employs endogenous technology learning for
1999 Compared with AEO1999 (GW).
some technologies. Learning by doing is a recognized phe-
nomenon. As the installed capacity of new technologies in- 1995 2000 2005 2010 2015 2020 %/y growth
2000–2020
creases, learning results in decreased costs. NEMS does this
with a segmented learning curve. USMM uses a two-step MM Coal steam 305 295 293 285 282 276 −0.33
process: Endogenous technology learning was run in a linear AEO 306 305 305 309 316 333 0.43
model under mixed integer programming (MIP). The new MM OTH FOS steam 139 139 122 104 86 83 −2.55
capital costs developed in the learning process were then AEO 139 138 102 80 80 76 −2.97
transferred into USMM. MM Combined cycle 15 52 91 135 181 207 7.11
NEMS provides an initial optimism factor that reduced AEO 14 27 89 126 176 212 10.82
the contingency allowance for some technologies. This does MM CT & diesel 56 61 88 95 157 184 5.66
AEO 55 99 141 151 145 187 3.23
not exist in USMM. We represented this optimism factor in
USMM, by lowering the discount rate for the technologies MM Nuclear power 99 95 87 74 56 49 −3.26
AEO 95 95 87 74 56 49 −3.26
with optimism factors.
MM Pumped storage 17 18 19 19 20 20 0.58
AEO 20 22 22 22 22 22 0.00
4. Results and discussion MM Fuel cells 0 0 0 0 0 1
AEO 0 0 0 0 0 0
The results are not a single number. There are literally MM Renewables 87 87 88 88 90 95 0.40
thousands of results from each model. In this paper we de- AEO 89 90 91 92 94 97 0.38
scribe only the electric supply sector, although the models MM Total 718 748 787 801 873 915 1.01
also give predictions of the residential, commercial, indus- AEO 723 776 837 854 918 974 1.14
trial, and transportation sectors. We look first at key sum- Note: We have little insight on the future role of nuclear power and have
mary results (e.g., primary energy use, electric generating simply forced MARKAL-MACRO to use the AEO values.
S.C. Morris et al. / A comparison of the electricity and carbon reduction projections 211

Table 5
Projected renewable electricity generating capacity: MARKAL-MACRO
base compared with AEO1999 (GW).

1995 2000 2005 2010 2015 2020 %/y growth


2000–2020

MM Hydroelectric 78.48 78.03 78.00 78.00 78.50 79.00 0.06


AEO 78.48 78.14 78.45 78.51 78.51 78.51 0.02
MM Geothermal 3.02 3.04 2.44 1.83 1.68 4.50 1.98
AEO 3.02 3.06 3.08 3.16 3.25 3.52 0.70
MM MSW 2.91 3.56 3.76 4.01 4.17 4.27 0.91
AEO 2.87 3.56 3.76 4.01 4.17 4.27 0.91
MM Wood OTH BIO 1.91 2.40 3.20 4.10 5.40 6.00 4.69 Figure 3. Combined cycle plants and combustion turbines, both models
AEO 1.91 1.76 1.97 2.33 3.61 5.61 5.97 show a strong increase in capacity (base case).
MM Solar thermal 0.72 0.67 0.62 0.56 0.51 0.55 −0.98
AEO 0.36 0.37 0.42 0.44 0.48 0.52 1.72
MM PV 0.02 0.14 0.25 0.33 0.43 0.52 6.78
AEO 0.01 0.04 0.14 0.30 0.46 0.64 14.87
MM Wind 2.84 3.07 3.30 3.54 3.77 4.00 1.33
AEO 1.89 2.80 3.24 3.39 3.41 3.61 1.28
MM Total 89.90 90.91 91.57 92.37 94.46 98.84 0.42
AEO 88.49 89.73 91.05 92.14 93.89 96.68 0.37
Note: Since municipal solid waste (MSW) was not treated as part of the
optimization in NEMS, we have simply set MARKAL-MACRO to use the
AEO values for MSW.
Figure 4. For cogeneration and diesel, the two models predict similar results
(base case).

Figure 1. For coal-fired steam plants, USMM shows a small but steady
decrease over time, while AEO shows a steady increase (base case).
Figure 5. For pumped storage, fuel cells, and renewables (figure 5) the two
models predict similar results (base case).

USMM makes a similar jump from 2010 to 2015. We made


no attempt to guess the fate of nuclear power plants and sim-
ply used the AEO values. For pumped storage, fuel cells,
and renewables (figure 5) the two models predict similar re-
sults. Overall, as shown in table 4 NEMS projects a some-
what stronger growth in electricity capacity than USMM.
Table 5 provides a closer look at the renewable tech-
nologies. The two models give similar results for conven-
tional hydropower (figure 6). NEMS shows a steady increase
Figure 2. For other fossil steam plants (oil and gas) both models show a in geothermal capacity (0.7%/y), while USMM shows a
steady decrease (base case).
decline, followed by an upward turn in 2020, with an annual-
ized rate of 2.0%/y (figure 7). Municipal solid waste (MSW)
tion turbines (figure 3). USMM projects a growth of 7.2%/y is an unusual power producer and difficult to project due to
for combined cycle and AEO projects 10.9%/y growth. For varying policies and public attitudes toward incinerators. As
combustion turbines, USMM projects 5.7%/y, while AEO with nuclear, we chose to simply follow the 1999 AEO.
projects 3.2%/y. There are, however, some differences in USMM projects stronger increases in biomass energy
timing. NEMS makes a big jump from 2000 to 2005, while then NEMS early on, but they both provide an upward trend
212 S.C. Morris et al. / A comparison of the electricity and carbon reduction projections

Figure 6. Conventional hydropower (base case).

Figure 9. Solar thermal; the two models start at different points about
0.4 GW apart, but they converge to the same point in 2020.

Figure 7. NEMS shows a steady increase in geothermal capacity while


USMM shows a decline, followed by an upward turn in 2020 (base case).

Figure 10. Photovoltaics; both models show a comparable and reasonably


steady increase (base case).

Figure 8. USMM projects stronger increases in biomass energy then NEMS


early on, but they both provide an upward trend (base case).
Figure 11. Wind; both models show a reasonably close and steadily upward
growth (base case).
(figure 8). USMM projects a 4.7%/y growth while NEMS
projects 6.0%/y. The two models take different tracts, but increase in both models (figure 10). USMM projects 6.8%/y
end up in 2020 very close. Their overall 20-year growth growth, while AEO projects 15%/y. The growth curves look
rates differ only by 1.3 percentage points. This kind of dif- very close; although NEMS growth rate is more than twice
ference may be as a result of USMM’s global perspective as USMM’s, the difference in 2020 is only 0.12 GW. Wind also
compared to NEMS’ year-by-year approach. follows a reasonably close and steadily upward growth in
The two models start at different points for solar thermal both models (figure 11). Both models have the same growth
(figure 9) about 0.4 GW apart. Nonetheless, they converge rate (1.3%/y) and are very close throughout the projection.
to basically the same point in 2020. USMM projects growth Overall, there are only minor differences between the two
of 0.98%/y, while AEO projects 1.7%/y. This may be an in- models in their projections for renewables.
dication that both models see about 0.5 GW an appropriate For geothermal energy, the results of the two models dif-
level. Although one shows negative growth and one posi- fer considerably. It is unusual to see a technology decline
tive, their growth rates are only 2.7 percentage points apart and then grow. As in the case with coal-fired power plants,
(2000–2020 growth rate USMM −0.98%/year compared to without upgrading, the capacity declines as individual plants
NEMS at 1.7%/year). This shows how a relatively small en- drop out. USMM found no need for geothermal until 2020
ergy source can appear to be quite different in a graph that when capacity increases in a 22 percent per year burst
includes only a tiny fraction of the total energy picture. for one 5-year period. USMM shows USMM projects
Photovoltaics show a comparable and reasonably steady 2.0%/year growth, while AEO projects 0.7%/y. Again, the
S.C. Morris et al. / A comparison of the electricity and carbon reduction projections 213

Table 6 Table 9
Projected electricity price: MARKAL-MACRO base case compared with Projected technology capacity MARKAL-MACRO compared with NEMS
AEO1999 (cents/kWh). −7% case (GW).
2000 2005 2010 2015 2020 %/y growth 1995 2000 2005 2010 2015 2020 %/y growth
2000–2020 2000–2020

MM 6.99 6.70 6.83 7.09 7.26 0.19 MM −7% Coal steam 305 285 282 221 160 99 −5.14
AEO 6.60 6.40 6.10 5.80 5.60 −0.82 NEMS 297 288 165 84 −6.14
MM −7% OTH FOS Steam 139 145 151 147 141 158 0.45
NEMS 130 120 111 70 −3.08
Table 7 MM −7% Combined cycle 15 52 89 130 174 201 6.95
Projected gross domestic product: MARKAL-MACRO base case compared NEMS 70 125 217 319 7.88
with AEO1999 (trillions of 1998$). MM −7% CT & diesel 15 52 89 130 174 201 6.95
NEMS 47 79 89 113 4.47
1995 2000 2005 2010 2015 2020 %/y growth
MM −7% Nuclear power 95 95 87 74 56 49 −3.26
2000–2020
NEMS 95 96 94 90 −0.28
MM 6.76 7.84 8.60 9.40 10.40 11.61 1.98 MM −7% Pumped storage 17 22 22 22 22 22 0.00
AEO99 6.76 7.83 8.77 9.90 10.80 11.68 2.02 NEMS 19 20 20 20 0.34
MM −7% Fuel cells 0 0 1 1 1 1 –
NEMS 0 0 0 0 –
Table 8 MM −7% Renewables 87 89 91 95 101 108 1.00
Projected carbon emissions: MARKAL-MACRO base case compared with NEMS 97 107 138 212 3.99
AEO1999 (million metric tons/year). MM −7% Total 718 744 808 771 808 821 0.49
NEMS 776 834 907 0.78
1995 2000 2005 2010 2015 2020 %/y growth
2000–2020 Note: We have little insight on the future role of nuclear power and have
MM 1443 1513 1596 1670 1872 2023 1.46 simply forced MARKAL-MACRO to use the AEO values. 2000 values not
AEO 1443 1585 1678 1790 1890 1975 1.11 available from NEMS were estimated as the mean of 1995 and 2005.

tion with the economy, although USMM makes a jump over


NEMS in 2020.
In overall, USMM’s results, for the most part, are not sig-
nificantly different from those of NEMS. The most impor-
tant difference is the decrease in coal-fired power plants.

4.2. Comparison of USMM with NEMS in the −7% case


(1999) [2]

Table 9 shows comparisons of the USMM base case


with the USMM case constrained to 7% below 1990 car-
Figure 12. Carbon emissions (base case). bon emissions and the NEMS 1990-7% early start case (EIA,
1999) [2].
overall growth rates throughout the period are only two per- There are significant differences in the way the two mod-
centage points apart. els handle the seven-percent reduction below 1990. NEMS
Table 6 shows the electricity prices. USMM shows a very projects twice the amount of renewables than USMM and
slightly increasing price (0.19%/y), while NEMS shows a more than twice the renewables in AEO1999, while USMM
very slightly decreasing prices (−0.82%/y). The difference provides only a moderate increase in renewables. More-
in growth rate is only one percentage point. The range be- over, USMM chooses a different set of renewables for in-
tween the two is only 1.7 cents/kWh and is within the range vestment. NEMS reduces other fossil steam electric (oil and
of U.S. electricity prices in 1999. One important difference gas) while USMM increases its investment in these tech-
here is that NEMS may be transitioning from a “cost of ser- nologies. USMM projects an increased capacity of combus-
vice” mode to a marginal cost mode, whereas USMM de- tion turbines, nearly doubling the NEMS projected capacity
rives the electricity price from the shadow price on electric- in 2020. An important reason for the difference between the
ity, a pure marginal cost approach, calibrated to the actual two models is the fact that the USMM base case projected
1995 price with the use of a mark-up factor. substantially less coal use, and thus found it easier to reduce
Table 7 indicates the models’ projection of the economy. carbon emissions.
Although the initial growth was derived from, USMM can For coal-fired steam plants, both NEMS and USMM de-
modify the rate of GDP growth endogenously. AEO projects crease capacity, but NEMS reduced coal use much more
GDP growth at 2.19%/y while USMM projects 2.16%/y. Ta- severely. NEMS also reduced other fossil fuels, while
ble 8 and figure 12 show predicted total carbon emissions. USMM holds them fairly steady. Both USMM and NEMS
These are unconstrained. AEO grows carbon emissions at provide a considerable increase in combined cycle plants,
1.11%/y while USMM projects growth of 1.46%/y. The but USMM maintains about the same level of increase as in
slightly lower increase in carbon emissions mimics the situa- the base case, while NEMS projects a 50% increase over the
214 S.C. Morris et al. / A comparison of the electricity and carbon reduction projections

Table 10
Projected renewable energy generating capacity: MARKAL-MACRO com-
pared with NEMS −7% case (GW).

1995 2000 2005 2010 2015 2020 %/y growth


2000–2020

MM −7% Hydro- 78.48 78.03 78.00 78.00 78.50 79.00 0.06


electric
NEMS 79.60 80.72 80.78 81.62 0.13
MM −7% Geo- 3.02 3.04 2.44 1.83 1.68 4.50 1.98
thermal
NEMS 3.43 3.83 5.50 7.51 4.00
MM −7% MSW 2.91 3.56 3.76 4.01 4.17 4.27 0.91 Figure 13. PV (7% case).
NEMS 3.27 3.63 3.99 4.43 1.53
MM −7% Wood 1.91 2.21 2.57 2.98 3.45 4.00 3.01
OTH BIO
NEMS 2.08 2.25 7.14 52.65 17.53
MM −7% Solar 0.72 1.16 1.86 3.00 4.84 7.80 10.00
thermal
NEMS 0.55 0.38 0.44 0.54 −0.09
MM −7% PV 0.02 0.06 0.17 0.26 0.43 0.69 12.99
NEMS 0.05 0.08 0.39 0.91 15.61
MM −7% Wind 2.94 4.12 6.03 9.11 11.67 12.08 5.53
NEMS 9.34 15.73 39.60 64.07 10.11
MM −7% Total 90.00 92.2 94.83 99.19 104.74 112.34 0.99
NEMS 98.3 106.61 137.84 211.74 3.91
Note: Since municipal solid waste (MSW) was not treated as part of the
optimization in NEMS, we have simply set MARKAL-MACRO to use the Figure 14. Wind (7% case).
AEO values for MSW. 2000 values not available from NEMS were esti-
mated as the mean of 1995 and 2005.

reference case. As discussed further below, NEMS intro-


duces renewables at a much higher rate than USMM.
Table 10 takes a closer look at renewables under carbon
constraint. The two models choose different technologies
to achieve the carbon constraint. Both models have similar
projections of hydroelectric.
For Geothermal, USMM exhibits the same form as in the
base case, decreasing, then increasing markedly. USMM
does not increase its base case projection of a growth rate
of 2%/y. NEMS, however increases the growth rate for Figure 15. Biomass (7% case).
Geothermal from its reference case of 0.7%/y to 15%/y, with
a total increase over its base case of 4 GW.
For PV, figure 13 shows that USMM and NEMS are rea-
sonably close. The difference in growth rate is only 3 per-
centage points and the total difference in 2020 is 0.2 GW.
USMM projects an increase over the base case PV of only
about 0.17 GW while NEMS increases by 0.27 GW in this
constrained case compared to the reference case.
Wind is a different story. NEMS grows wind at 10%/year
to 64 GW in 2020 from practically nothing. A sustained 20
year growth at 10%/year is something very hard to achieve.
In comparison, AEO2001 projects only 19 GW in 2020 Figure 16. Solar thermal (7% case).
for its “High Renewable Energy Case”. USMM projects a
growth rate of only 5.55%/y, yielding a 2020 value of 12 able Energy Case for 2020 was 3.2 GW. USMM projects a
GW compared with 4 GW in the base case (figure 14). 3.3% growth with a 2020 projected capacity of 6 GW com-
The differences in the biomass predictions are similar to pared to 4 GW in the base case.
those of the wind predictions (figure 15). NEMS grows bio- For solar thermal, NEMS projects low growth (1.9%/y),
mass from 2 GW to over 50 GW. Moreover, the rapid growth whereas USMM grows solar thermal at 10%/y, a difficult
does not begin until 2005. This is a 23%/y growth main- growth rate to maintain, but still reaches only 8 GW by the
tained over 15 years. This is an extraordinary growth rate to year 2020 compared to 0.6 GW in the base case (figure 16).
sustain over that length of time. The AEO2001 High Renew- USMM prefers solar thermal to biomass or wind, but it still
S.C. Morris et al. / A comparison of the electricity and carbon reduction projections 215

Table 11
only brings the USMM total to 24 GW. A possible expla-
Projected electricity price MARKAL-MACRO and NEMS −7% case
nation is that there were differences in the solar availability (cents/kWh).
estimates between the two models. In the 2001 AEO “High
2010 2015 2020 %/y growth
Renewable Energy Case”, the projection for solar thermal in
2010–2020
2020 was 0.48 GW.
For Geothermal, USMM exhibits the same form as in the MM −7% 13.54 13.64 14.31 0.55
NEMS −7% 10.40 9.65 8.90 −1.55
base case, decreasing, then increasing markedly. USMM
does not increase it’s base case projection of a growth rate of Note: 2000 values not available from NEMS were estimated as the mean of
0.02%/y, while NEMS increases the growth rate for Geother- 1995 and 2005.
mal from 0.7 to 4%/y with a 2020 increase over the base case Table 12
of 4 GW. Projected gross domestic product: MARKAL-MACRO compared with
Overall, NEMS projects nearly twice the capacity in re- NEMS −7% case (trillions of 1998$).
newables than USMM. 1995 2000 2005 2010 2015 2020 %/y growth
Table 11 shows projections of electricity prices under the 2000–2020
specified carbon constraint. The price shows a similar pat-
MM −7% 6.760 7.840 8.590 9.290 10.190 11.320 1.85%
tern to the base case, although the divergence of the two NEMS −7% 7.52 8.27 9.31 10.81 1.83
models was more pronounced. USMM projects an increas-
ing price growing at a rate of 0.55%/y, while NEMS projects Note: 2000 values not available from NEMS were estimated as the mean of
1995 and 2005.
prices decreasing by 1.5%/y. The absolute difference in
2020 is 5.41 cents/kWh, a substantial difference. Table 13
NEMS calculates prices in two ways [7]. First, NEMS Projected carbon emissions: MARKAL-MACRO −7% compared with
uses an approach that simulates the cost-of-service method NEMS −7% case (million metric tons/year).
used by state regulators. The cost of service prices repre- 1995 2000 2005 2010 2015 2020 %/y growth
sents average costs. Second, NEMS determines “compet- 2000–2020
itive” prices for electricity generation, based on marginal MM −7% 1421 1454 1517 1253 1253 1253 −0.74
costs. Marginal costs are primarily the operating costs of the NEMS −7% 1362 1302 1250 1249 −0.43
most expensive plant required to meet the demand. Prices
Note: 2000 values not available from NEMS were estimated as the mean of
for transmission and distribution are still based on average 1995 and 2005.
cost. These costs are then added to the generation costs. It
is not clear which method was used in the AIA report; re- Table 14
sults show that likely the cost-of-service method was used. Carbon price: MARKAL-MACRO compared with NEMS −7% case
($/ton).
USMM determines the cost of electricity as the shadow price
of summer day electricity. The shadow price is the pure mar- 1995 2000 2005 2010 2015 2020
ginal cost of delivering electric to the customer. This is the −7% $254 $281 $235
cost of the most expensive plant required to meet the demand NEMS $344 $290
(this is the last plant to come on line). The decrease in elec-
Note: NEMS electricity price only available for 2010 and 2020. NEMS
tricity cost is due to the greater efficiency of combined cycle GDP and carbon emissions only available for 2005, 2010, and 2020. NEMS
plants, which weigh heavily in their base case and their −7% carbon prices were presented in 1996 dollars and converted to 1998 dollars,
case projections and due to manufacturers ability to reduce using a multiplier of 1.087 [2–4].
capital costs of wind and biomass technologies as output in-
creases [7]. ble 12). This results in a 2.6% decrease in the GDP in 2020
One reason why the NEMS electricity price decreases is compared to the base case for USMM and an 8% decrease
the projected shift to a market with marginal cost pricing, in 2020 compared to the AEO reference case for NEMS.
which is expected to result in lower prices. USMM consis- Reducing carbon emissions involves significant costs that
tently uses marginal cost pricing with a “mark-up” for ad- would affect the economy. USMM projects slightly lower
ministrative and other costs not otherwise fully captured in economic growth in the base case (2.16%/year compared to
the model. It is possible that the learning curve in NEMS, 2.19%/year for AEO1999). USMM, however, maintains a
being somewhat more complex than that in USMM, may re- stronger growth rate in the −7% case (2.06%/year compared
duce costs faster. The NEMS projection is also premised to NEMS 1.83%/year). This difference between the two
on the assumption that after 2001, the price of natural gas models is again due to the lower use of coal in the USMM
(the fuel for combined cycle plants) will drop again to 1999 base case. Both models show lower carbon prices (tax) in
levels and then only gradually begin increasing again. 2020 than in 2010, based on the ability to bring on board
Both models project a decrease in GDP under carbon con- more efficient technologies and more non-carbon emitting
straint compared to the base case. USMM projects a growth technologies.
rate of 1.98%/y for the base and 1.85%/y under carbon con- Table 13 shows the carbon emissions. Table 14 indicates
straint. NEMS projects a growth rate of 2.02%/y in the ref- the carbon emission prices. The prices of both models are
erence case but 1.83%/y in the carbon constrained case (ta- in the same range. NEMS, however, projects a decrease in
216 S.C. Morris et al. / A comparison of the electricity and carbon reduction projections

carbon price over time, while USMM projects a continuing References


increase.
EIA (1998) lists seven factors that result in differences in
projected carbon emission prices among models. Pertinent [1] A. Brooke, D. Kendrick and A. Meeraus, GAMS, a User’s Guide
to this comparison is the amount of knowledge decision- (GAMS Development Corp., Washington, DC, 1996).
makers are assumed to have about future events, amount of [2] EIA, Annual Energy Outlook 1999 (EIA-0383), Energy Information
Administration, U.S. Department of Energy, Washington, DC (1999).
lead time decision-makers have to adjust; the level of aggre- [3] EIA, Assumptions to the Annual Energy Outlook 1999. (DOE/EIA-
gation for technologies and goods; the focus on costs associ- 0554), Energy Information Administration, U.S. Department of En-
ated with transition to lower carbon emissions, and assump- ergy, Washington, DC (1999).
tions about the speed and extent of changes consumers can [4] EIA, Annual Energy Review 1999. (DOE/EIA-0384-(99)), Energy In-
make. formation Administration, U.S. Department of Energy, Washington,
DC (1999).
Three important differences between the two models [5] EIA, Analysis of the impacts of an early start with the Kyoto protocol
clearly contribute to the differences in results. First, the (SR/OIAF/99-02), Energy Information Administration, U.S. Depart-
USMM assumption of marginal cost pricing corresponds ment of Energy, Washington, DC (2000).
to a completely deregulated energy market. NEMS uses a [6] EIA, Annual Energy Outlook 2001 DOE/EIA-0383 (2001) Energy In-
mixed market that more closely matches the actual US en- formation Administration, U.S. Department of Energy, Washington,
DC (2000).
ergy market. Second, the two models differ substantially
[7] EIA, National energy modeling system an overview 2000 (EIA-
in the technology change process. USMM applies a global 0581), Energy Information Administration, U.S. Department of En-
optimization over the entire study period, effecting perfect ergy, Washington, DC (2000).
foresight when investing in technology. NEMS, on the other [8] L.G. Fishbone and H. Abilock, MARKAL, a linear-programming
hand, makes decisions year-by-year. Third, the two models model for energy systems analysis; technical description of the BNL
differ considerably in the handling of tradeoffs between en- version, Energy Research 5 (1981) 353–375.
[9] L.G. Fishbone, G. Giesen, G. Goldstein et al., User’s guide for
ergy, capital, labor and consumption. USMM uses a simpler MARKAL (BNL/KFA version 2.0) a multi-period linear program-
macroeconomic model in comparison with NEMS. ming model for energy systems analysis (BNL 51701), Brookhaven
National Laboratory, Upton, N.Y. and Kernforschungsanlage Jülich,
D5170, Germany (1983).
5. Conclusion [10] G.A. Goldstein, L.A. Greening and partners in the International
Energy Agency, Energy Technology Systems Analysis Programme
NEMS and USMM predict similar results for the base (www.ecn.nl/unit_bs/etsap/).
[11] L.D. Hamilton, G.A. Goldstein, J. Lee, A.S. Manne, W. Marcuse,
1999 US annual energy outlook case, but their results in a S.C. Morris and C.-O. Wene, USMM: an overview (BNL 48377),
case with carbon constraints are different, especially in the Brookhaven National Laboratory, Upton, NY 11973 (1992).
small capacity (e.g., renewable) technologies. The differ- [12] A.S. Manne and C.-O. Wene, USMM: a linked model for energy-
ences of the models and those of their predictions are ex- economic analysis (BNL 47161), Brookhaven National Laboratory,
plained by the different assumptions and attributes of the Upton, NY (1992).
[13] A. Seebregts, T. Kram, J. Gerrit and A. Boz, Endogenous learning and
two models. These differences being transparent, both mod-
technology clustering: analysis with MARKAL model of the West-
els are equally useful to the energy policy analyst. USMM ern European energy system, Int. J. Global Energy Issues 14 (2000)
can be used to provide an alternative and complementary ap- 289–319.
proach to projections of renewable technologies penetration
and their potential in reducing carbon dioxide emissions in
the USA.

You might also like