GE Energy
The Value of Forecast Accuracy
Richard Piwko
GE Energy Consulting
New York State Wind Study
Sponsored by NYISO and NYSERDA
Objective
To produce empirical information that will assist
the NYISO in evaluating the reliability
implications of increased wind generation
2
Wind MW by Zone 322
In Study Scenario
33,000 MW
Statewide
Total of 3,300 MW
Peak Load
Wind Generation
in 2008
Statewide
(10% of Peak Load)
358
400
261
105
570
684
600
3
Study Scenario; Wind and Load MW by Zone
( Total Wind Generation = 10% of Peak Load )
Total Potential 2008 Noncoincident Wind MW in Wind as % of
Wind Generation Peak Load Study Scenario Peak Load
Zone A 3,070 2,910 684.2 24%
Zone B 1,197 2,016 358.5 18%
Zone C 1,306 2,922 569.7 19%
Zone D 483 902 322.6 36%
Zone E 2,832 1,592 399.8 25%
Zone F 434 2,260 260.6 12%
Zone G 105 2,260 104.6 5%
Zone H 0 972 0.0 0%
Zone I 0 1,608 0.0 0%
Zone J 0 11,988 0.0 0%
Zone K 600 5,275 600.0 11%
sum 10,026 34,704 3300.0 10%
DPS Zn 1 8,887 10,342 2334.8 23%
DPS Zn 2 538 7,099 365.2 5%
DPS Zn 3 600 17,263 600.0 3%
sum 10,026 34,704 3300.0 10% 4
Analysis of Load and Wind Forecasting
Hourly load MW for 2001, 2002, 2003
• Forecast and actual
Hourly wind MW for 2001, 2002, 2003
• Forecast and actual
Production cost simulation for 3 years
• GE’s MAPS program
– Modeled NYISO, ISONE, and PJM regions
• Hourly load and wind profiles from 2001, 2002, 2003
• System load scaled for year 2008
5
Load Data and Wind Data Are
Time-Synchronized for All Analysis Performed
35000 3000
30000 NY Loads 2500
WIND
Wind Output (MW)
NYISO Load (MW)
25000
2000
20000
1500
15000
1000
10000
5000 500
0 0
1 25 49 73 97 121 145
Hour of Week
Source: One week in July 2002 6
NYISO Load: Day-Ahead Forecast vs Actual
25000
20000
MW
15000
10000
Load Forecast Actual Load
5000
1 49 97 145 193 241 289
Hour
Source: Two weeks in January 2001 7
Wind Generation: Day-Ahead Forecast vs Actual
25000
Forecast Wind Actual Wind
20000
15000
MW
10000
5000
0
1 49 97 145 193 241 289
Hour
Source: Two weeks in January 2001 8
Total Load and Wind: Forecast vs Actual
25000
20000
MW
15000
10000
Forecast Total (L - W) Actual Total (L - W)
5000
1 49 97 145 193 241 289
Hour
Source: Two weeks in January 2001 9
Load Forecast Error
3000
Load Error (F-A)
2000
MW Error
1000
0
-1000
1 49 97 145 193 241 289
Hour
Source: Two weeks in January 2001 10
Total Forecast Error (Load and Wind)
3000
Total Error (F - A)
2000
MW Error
1000
0
-1000
1 49 97 145 193 241 289
Hour
Source: Two weeks in January 2001 11
Standard Deviations of Day-Ahead Forecast Errors
950 MW
With Wind
800 MW
1000 Without Wind
800
Sigma (MW)
600
400
200
0
15000 20000 25000 30000 35000
Peak Load for Corresponding Month (MW)
Load Wind Load - Wind January 2001
12
Hourly Production Simulation Analysis
Examine the impact of 3300 MW wind generation
on overall operation of the NYISO grid
• Generation displacement
• Emission reductions
• Transmission congestion
• Economic impact
– Fuel costs
– Variable operation & maintenance costs
– Start-up costs
– Emission payments
13
Cases Analyzed
Base Case – Grid operation with no wind generation
• Basis of comparison for operation with wind
Cases with 3300 MW Wind generation (nameplate rating)
• Operation with no wind generation forecast
• Operation with state-of-art wind forecast
• Operation with perfect wind forecast
14
Operating Costs Without Wind Forecasts
• Wind generation not considered for day-ahead unit commitment
• Wind generation just “shows up” in the real time market
No Wind SOA Wind Perf
Forecast Forecast Fo
$ millions $ millions $m
Total variable cost reduction $ 335 $ 430 $
Wind generator revenue $ 305 $ 315 $
Non-wind generator revenue reduction $ 960 $ 795 $
Load payment reduction $ 720 $ 515 $
15
Operating Costs With
State-of-Art Wind Forecasts
• Day-ahead unit commitment considers forecasted wind generation
No Wind SOA Wind Perfect Wind
Forecast Forecast Forecast
$ millions $ millions $ millions
Total variable cost reduction $ 335 $ 430 $ 455
Wind generator revenue $ 305 $ 315 $ 325
Non-wind generator revenue reduction $ 960 $ 795 $ 755
Load payment reduction $ 720 $ 515 $ 485
16
Benefit of State-of-Art Forecast
System Variable Cost Reduction
Using SOA Wind Forecast: $430 M
No Wind Forecast: $335 M
Net Benefit : $ 95 M
Wind Generation : 8900 GWh
Value of SOA Forecast: $10.70 /MWh
17
Operating Costs With
Perfect Wind Forecasts
• Day-ahead unit commitment considers forecasted wind generation
No Wind SOA Wind Perfect Wind
Forecast Forecast Forecast
$ millions $ millions $ millions
Total variable cost reduction $ 335 $ 430 $ 455
Wind generator revenue $ 305 $ 315 $ 325
Non-wind generator revenue reduction $ 960 $ 795 $ 755
Load payment reduction $ 720 $ 515 $ 485
18
Benefit of Perfect Wind Forecast
System Variable Cost Reduction
Using Perfect Wind Forecast: $455 M
Using SOA Wind Forecast: $430 M
Net Benefit : $ 25 M
Wind Generation : 8900 GWh
Remaining Value of Improving Forecast: 2.80 /MWh
Total Potential Value of Forecast : $13.50 /MWh
19
Conclusions
Conclusions
State-of-the-art wind generation forecasts can
improve the value of wind energy by ~25%
• $335M vs $430M operating cost reduction with
3,300 MW wind generation in NY State
Improved wind forecasts could produce further
benefits, but with diminishing returns
21
Complete report is available at . . . . .
http://www.nyserda.org/publications/
wind_integration_report.pdf
Appendices are available at . . . . .
http://www.nyserda.org/publications/
wind_integration_apps.pdf
22
GE Energy
The Value of Forecast Accuracy
Richard Piwko
GE Energy Consulting