Electricity Sector Disaggregation

Global Trade Analysis Project
Capacity utilization and expansion in
the dynamic energy landscape
Jeffrey C. Peters
PhD Candidate (Dec 2015), Center for Global Trade Analysis , Purdue University
Thomas W. Hertel
Executive Director, Center for Global Trade Analysis, Purdue University
33rd USAEE/IAEE North American Conference (2015)
Center for Global Trade Analysis
Department of Agricultural Economics, Purdue University
403 West State Street, West Lafayette, IN 47907-2056 USA
[email protected]
http://www.gtap.agecon.purdue.edu
Examples of the dynamic energy
landscape
• US shale oil and gas boom and fall in gas prices
• Decreasingly relative price for electricity generation from gas
• Opportunity for oil exports
• Opportunity for LNG exports
•
•
•
•
Increasing efficiency of renewable technologies
Increasing efficiency of end-use electricity
Plug-in electric vehicles
Clean Power Plan and other environmental policies
• Economy-wide factors may have important consequences on the
electricity sector
2
Electric power and economy-wide
modeling
• “Bottom-up” models
• Partial equilibrium or simulation-based
• Can be technologically-rich
• Exogenous projections of input costs and electricity demand drive
endogenous outcomes in the electricity sector
• Typically, capacity factors for technologies and fuel prices are
fixed
• “Top-down” models – computational equilibrium
• Economy-wide equilibrium captures inter-industry and interregional linkages
• Endogenous input prices and electricity demand – “feedbacks”
• Limited sector-level detail
• Rarely validated against observations
Electricity
sector
Rest of
economy
Electricity
sector
Rest of
economy
3
Increasing technological detail
• Computational equilibrium models (e.g. CGE) are well-suited for
the economy-wide linkages in the dynamic energy landscape
• How can we overcome aforementioned limitations?
• Advances in economically-consistent databases
• GTAP-Power expands “electricity” to T&D and 11 generating technologies (Peters, 2015)
• Matrix balancing specific to electric power (Peters and Hertel, forthcoming)
• Balancing methodology shown to influence modeling results (Peters and Hertel, in review)
• Advances in representing electric power
• Capacity factor utilization – i.e. adjustments to economic conditions with existing
capacity
• Capacity expansion – i.e. additional and retiring capacity
• Validated against observations
4
Capacity utilization and expansion
• Explicitly and endogenously determine capacity utilization,
expansion, and their interdependency
• Increased utilization drives up returns to capital, drives expansion
• Increased expansion can crowd-out utilization
𝑔
𝑔
𝑐
• 𝑄𝑡 = 𝑐𝑡 ∙ 𝛼 ∙ 𝑄𝑡
→ 𝑞𝑡 = 𝑐𝑡 + 𝑞𝑡𝑐 (percent change)
5
Utilization: flexible vs inflexible
• Flexible technologies
substitute O&M for capital
• Increase labor
• Increase regularly scheduled
maintenance
• Increasingly costly
• Inflexible technologies
cannot substitute (fixed
short-run capacity)
6
Utilization: substitution
• Substitution of flexible
technologies
•
•
•
•
Imperfect substitution
Represent base and peak load
Impacts returns to capital
Decrease in gas prices leads to:
• substitution to gas power,
increasing returns
• substitution away from coal power,
decreasing returns
• decreasing returns for inflexible
technologies due to lower overall
cost
7
Utilization: validation
• Exogenous shocks
• Input prices
•
•
•
•
O&M
Gas
Oil
Coal
• Income
• Population
• Total electricity
demand
• Capacity expansion
8
Expansion: a multinomial logit model
• The MNL is a choice model where
• Utility of the choice is given by
• 𝑈 𝑄𝑡𝑐 = 𝑎𝑡𝑐 ⋅ 𝑃𝑘,𝑡 −
𝑖 𝑇𝑖𝑡
• With probability of choosing
• 𝑃𝑡𝑐 =
𝑐)
𝑈(𝑄
𝑡
𝑒
𝑈(𝑄𝑐
𝑠)
𝑒
𝑠
• Which is also the share of new capacity allocated to a certain
technology
• 𝑄𝑎𝑐 = 𝑄𝑞 -𝐶 + 𝑄𝑟𝑐
• Need to validate
• Total capacity 𝑄 𝑎𝑐 − 𝑄𝑟𝑐
• Contributions from each technology 𝑃𝑡𝑐
9
Expansion: controlling for total capacity
• Control for total
capacity
• “Perfect foresight”
of service year fuel
prices
10
Expansion: controlling for total capacity
• Service year prices
Foresight of decline in gas prices
• Planning year prices
• Reality somewhere
in between
• Model fails in an
expected way
11
Expansion: projecting total capacity
• Exogenous projections of generation needs from rolling
average of EIA Annual Energy Outlooks
Not all 2017 and 2018 planned yet
AEO overestimated actual
generation needs
• Again, fails in expected way
• Highlights the importance of economic linkages
12
Overcoming the limitations
• Limited sector-level detail
• Capacity factor utilization
• Capacity expansion
• Their interdependency
• Rarely validated against observations
• Capacity factor utilization is highly correlated with observations 2002--2012
• Total capacity expansion is highly correlated using EIA AEO demand
projections
• Contributions to expansion for each technology are also highly correlated
• The validation exercises fail in expected ways
13
Carbon tax versus investment subsidy
• US Clean Power Plan
• Improved plant-level efficiency (exogenous)
• Switching from coal to gas power with existing plants (utilization)
• Constructing more renewable power (expansion)
• Two strategies
• Carbon tax (economically efficient)
• Investment subsidy for wind and solar (a more tractable policy?)
• How does the US electric power sector evolve in the response
to these two strategies?
14
Preliminary results: shocks to 2030
Baseline
Carbon Tax
Wind and solar subs.
2014 fuel prices
Baseline shocks
Baseline shocks
Population
Income
Labor cost
Swap total generation
with TFP
Swap total generation
with TFP
Total generation with
endogenous TFP
Carbon tax of
$34/metric ton CO2
Capital subsidy for
wind and solar -70%
-13.6% total CO2
-23.6% total CO2
-23.6% total CO2
15
Results: utilization and returns
Percentage change in
capacity utilization
80
1 Declining capacity factor
60
40
Baseline
3 "Loses" with renewable subsidy
20
Carbon Tax
0
-20
Nuclear
Coal
GasBL
Wind
Hydro
Other
GasP
Oil
Solar
W+S Subsidy
-40
2 "Hurt" more under carbon tax
-60
Percentage change in
returns to capacity
60
40
4 Returns hit mainly by tax
Baseline
5 High relative rates of return
20
Carbon Tax
0
-20
-40
-60
Nuclear
Coal
GasBL
Wind
Hydro
Other
GasP
Oil
Solar
W+S Subsidy
6 Other tech loses big by picking winners
16
Results: generation
5,000
4,500
127
TWh
3,000
214
1,356
1,623
817
4,000
3,500
192
573
971
2,500
507
492
734
516
800
900
2,000
1,210
1,500
1,000
79
230
109
243
62
228
500
841
863
843
Baseline
Carbon Tax
W+S Subsidy
-
Solar
Wind
Oil
GasP
GasBL
Coal
Other
Hydro
Nuclear
17
Conclusions and future work
• Important economic and operational insight can be captured
• Linkage between capacity utilization and returns to capacity
• Investment subsidies picks winners (and losers)
• The computational equilibrium here overcomes methodological
limitations
• Detailed representation of electricity
• Validated against observations
• The next step is to incorporate stronger inter-industry and interregional linkages in CGE framework
• Welfare impacts – total and distributional
• Trade – LNG, coal opportunities and impacts domestically and abroad
18
Global Trade Analysis Project
Thank you
Jeffrey C. Peters and Thomas W. Hertel
[email protected]
Center for Global Trade Analysis
Department of Agricultural Economics, Purdue University
403 West State Street, West Lafayette, IN 47907-2056 USA
[email protected]
http://www.gtap.agecon.purdue.edu
Electricity disaggregation
• Many researchers have independently disaggregated the
electricity sector into specific technologies
• Technology-specific policies (renewable subsidies)
• Refined operational considerations (generation mixes)
Tech 1
GTAP ‘ely’
Capital
O&M
Coal
Gas
Oil
Tech … Tech T
Capital
O&M
Coal
Gas
Oil
20
The disaggregation problem
• Termed the matrix-balancing problem:
• “Given a rectangular matrix Z0, determine a matrix Z that is
close to Z0 and satisfies a given set of linear restrictions on its
entries.” (Schneider and Zenios, 1990)
Tech 1
Tech …
Tech T
Tech 1
Tech …
‘ely’
Tech T
Capital
Capital
Capital
O&M
O&M
O&M
Coal
Z0
Coal
Z0
Coal
Gas
Gas
Gas
Oil
Oil
Oil
21
Constructing the target matrix, Z0
• The “bottom-up” data to create Z0 :
• total input employment in aggregate sector (e.g. GTAP ‘ely’)
• total generation (GWh) by each new technology
• levelized/annual costs of capital, O&M, and fuels
• Many researchers use the same or similar data
• However, the matrix-balancing methodologies to convert Z0 to
Z differ
• Resulting in fundamentally different baselines for modeling
• Remain largely undocumented
22
Share preserving cross entropy
min
𝑧𝑖𝑡
𝑖
𝑡 𝒛𝒊𝒕
∙ ln
∗
𝑡 𝑧𝑖𝑡 = 𝑜𝑖
𝑐𝑖𝑡 ∙𝑟𝑖𝑡
0 0
𝑐𝑖𝑡
∙𝑟𝑖𝑡
(16)
(17)
23
Correlations
24
Utilization: validation
• Exogenous shocks
• Input prices
•
•
•
•
O&M
Gas
Oil
Coal
• Income
• Population
• Total electricity
demand
• Capacity expansion
25
Utilization: policy-adjusted validation
• Includes non-economic
considerations
• EPA mercury regulations
• Increased base load
substitution due to
shortening of coal contracts
• Gains in correlation
• Illustrates the joint
importance of qualitative
information
26