Zonal Supply Curve Estimation in Transmission

Download Report

Transcript Zonal Supply Curve Estimation in Transmission

Zonal Electricity Supply Curve Estimation
with Fuzzy Fuel Switching Thresholds
North American power grid is “the largest and most complex machine in the world” Amin, (2004)
Mostafa Sahraei-Ardakani
Seth Blumsack
Andrew Kleit
Department of Energy and Mineral Engineering
Penn State University
[email protected]
MOTIVATION
How to analyze supply and demand policies
considering the transmission constraints ?
– Pennsylvania’s Act 129: Energy conservation and
peak demand reduction in Pennsylvania.
• What would happen to the prices in PA?
• What would happen to the prices in other states?
• What would happen to the emissions?
– Carbon tax:
• What would happen to the prices?
• What would happen to the emissions?
7/6/2015
2
DISPATCH CURVE MODEL
•
What would happen to electricity prices if a CO2
price was imposed?
•
•
Very complex model
Data may
not be publicly available
Each point represents
a single power plant
– Policy analysts
•
•
•
•
7/6/2015
Newcomer et al., 2008
– Engineers
Collect marginal cost data from power plants
Collect fuel price data
Form a supply curve by sorting generators from cheap to
expensive
Ignore transmission network
3
DIFFERENT MODELS
•
Engineering models
–
–
–
•
Econometric models
–
–
•
Estimate prices well
Do not do a good job in estimating
fuel mix and emission impacts of policies.
Dispatch curve
–
•
Too complex
Data may not be available
Takes a long time to converge
Ignores transmission system and how congestion makes prices
different.
Our model
–
–
7/6/2015
Needs no more data than a dispatch curve
Implicitly accounts for transmission constraints
4
OTHER APPROACHES
•
Econometric models
– Predict prices well.
– Do not do a good job on estimating fuel utilization.
•
Engineering models: Power Transfer Distribution
Factor (PTDF)
– Need detailed data which is not publicly available.
– They are complex and take a lot of time to converge
for large power systems.
7/6/2015
5
OUR APPROACH
For each zone we want to identify:
1. Thresholds where the marginal fuel
changes (Coal, Gas, Oil)  CMA-ES
Fixed and variable thresholds
1. The slope of each portion of the
overall dispatch curve.  OLS
7/6/2015
6
FUZZY THRESHOLDS
Fuzzy Gap
qT
Variables to be estimated:
1. Relative fuel price
Summer for having the
threshold
2011gap
fuzzy
2. Fuzzy gap width coefficient
Fuzzy Thresholds
Summer 2008
Deterministic
Thresholds
qT,C/G
GAS
COAL
ΔC/G
7/6/2015
qi
qi,G/O
7
IMPLEMENTATION IN PJM
•
•
Seventeen PJM
utility zones
Data: (2006-2009)
–
–
–
•
•
Hourly zonal load
Hourly zonal prices
Fuel prices
Insufficient data for
nodal level modeling
Robustness Check:
–
–
7/6/2015
Linear and quadratic
curves
Fixed and Variable
Thresholds
8
RESULTS: THRESHOLDS
Price
($/MWh)
PSEG= Public Service Electric and Gas Company
Marginal Fuels in PSEG
PSEG demand=
5.8 GW
$/MWh
PJM demand= 118 GW
APS price=80 $/MWh
Gas
Total Load in PJM (GW)
Oil
Gas-Oil
Fuzzy Region
Coal – Gas Fuzzy
Region
Coal
7/6/2015
Load in PSEG (GW)
9
RESULTS: SUPPLY CURVE PROJECTION
Central Pennsylvania and West Virginia
Philadelphia
• Zonal price differences are captured.
•50 $/ton carbon tax
7/6/2015
10
RESULTS: MARGINAL FUEL SHARES
•DUQ in western
PA is a coal
dominated zone.
•RECO in northern
NJ is a natural gas
dominated zone.
•Natural gas often
sets the prices in
PJM.
• Another robustness check
• Natural Gas often sets the prices.
7/6/2015
11
RESULTS: PRICES
•BGE is in eastern
PJM (Baltimore).
•DUQ is in western
PJM (Pittsburgh).
•Our model
captures zonal price
differences.
•50 $/ton carbon
tax would increase
prices by about
70%.
7/6/2015
12
APPLICATION: PENNSYLVANIA ACT 129
• Act 129 is a wide-reaching energy policy initiative in
Pennsylvania. Among other things, Act 129 requires all
Pennsylvania utilities to:
1. Reduce annual electricity demand by 1%
2. Reduce “peak” demand (highest 100 hours) by 4.5%
• We will estimate the impacts of Act 129 on total electricity
costs, fuels utilization and greenhouse gas emissions in the
PJM territory, using our model and the “dispatch curve”
model that I discussed earlier. We use 2009 as our “base”
year.
7/6/2015
13
APPLICATION: PENNSYLVANIA ACT 129
Electricity Cost Savings ($ million):
Zone
APS
AEP
AECO
BGE
COMED
DPL
DUQ
JCPL
METED
PECO
PPL
PENELEC
PEPCO
PSEG
RECO
DAY
DOM
PJM 7/6/2015
Total
Without Act 129
2,280
5,471
638
2,026
4,222
1,072
550
1,311
838
2,253
2,189
809
1,972
2,546
83
688
5,669
34,617
With Act 129
2,220
5,465
635
2,026
4,209
1,069
535
1,304
822
2,194
2,144
793
1,973
2,533
83
687
5,674
34,366
Savings
60
6
3
0
13
3
15
7
16
59
45
16
-1
12
0
1
-5
250
Savings
(%)
2.64
0.11
0.51
0
0.3
0.29
2.78
0.51
1.97
2.62
2.06
1.98
-0.03
0.48
0.47
0.12
-0.08
0.01
• Savings: 333 million
dollars
•253 million dollars in PA
•Dispatch Curve: 150
million dollars
14
APPLICATION: PENNSYLVANIA ACT 129
Shifts in Marginal Fuel (% Increase with Act 129):
Zone
APS
AEP
AECO
BGE
COMED
DPL
DUQ
JCPL
METED
PECO
PPL
PENELEC
PEPCO
PSEG
RECO
DAY
DOM
7/6/2015
Coal
1.4
2.3
1.39
0.95
1.6
0.86
3.13
0.89
1.21
0.98
1.04
1.9
0.98
0.93
1.11
2.63
0.94
Gas
-1.13
-2.1
-1.33
-0.94
-1.6
-0.82
-3.1
-0.89
-0.98
-0.98
-0.78
-1.89
-0.96
-0.92
-1.11
-2.63
-0.94
Oil
-0.28
-0.21
-0.07
0
0
-0.03
-0.02
0
-0.23
0
-0.25
-0.01
-0.02
0
0
0
0
Emission decreases by
4 million metric tons.
Dispatch Curve: 2.3
million metric tons.
15
CONCLUSIONS
•
•
We have developed an approach to estimating zonal supply
curves in transmission constrained electricity markets:
-
Requires no proprietary data
-
Can be implemented by analysts without requiring
complex engineering calculations
Our approach captures regional effects of policies that
“transmission-less” dispatch models do not. Regional impact
differences may be important in policy evaluation.
⁻
Zonal fuel utilization shift
⁻
Zonal price differences
7/6/2015
16
Thanks!
Mostafa Sahraei-Ardakani
Department of Energy and Mineral Engineering
Penn State University
[email protected]
Comprehensive Exam
PRICE INCREASE IN DC
MC1=MC2
P1+P2=10+50 (MW)
MC1=P1
($/MWh)
λ1=MC1=35
λ1=MC1=20 ($/MWh)
10 MW 25$/MWh
20MW
35
25MW
1
Virginia and Washington, DC
Rest of PJM
10MW
25/3
MW
MW
25
MW
20
50/3
5
40$/MWh
($/MWh)
λ2=MC2=35 ($/MWh)
λ2=MC1=50
MW
25
MW
3050/3
25MW
MC2=10+P2
MW
MW
40
25
7/6/2015 MW
30
Thermal Capacity
=20 MW
2
3
50MW
45MW
18