Imagination at work
Forecasting wind energy GE Energy Consulting Group March 30, 2016
Gene Hinkle Manager Director, Power Economics
GE Energy Consulting Group
518-385-5447
2
GE’s integration of renewables experience 2004 New York
3 GW Wind
10% Peak Load
4% Energy
2005 Ontario 15 GW Wind
50% Peak Load
30% Energy
2006 California 13 GW Wind
3 GW Solar
26% Peak Load
15% Energy
2007 Texas 15 GW Wind
25% Peak Load
17% Energy
2009 Western U.S. 72 GW Wind
15 GW Solar
50% Peak Load
27% Energy
2010 New England 12 GW Wind
39% Peak Load
24% Energy
2012 Nova Scotia ~1500MW Wind
40% Energy
2013 PJM
96GW Wind
22GW Solar
30% Energy
2014 Minnesota 8 GW Wind
4.5 GW Solar
50% Energy
Underway Pan-Canadian
~72GW Wind
30% Energy
Studies commissioned by utilities, commissions, ISOs...
• Examine feasibility of 100+ GW of new renewables
• Consider operability, costs, emissions, transmission
Gradients indicate systems subject to individual studies and also included in larger regional studies
GLOBAL RENEWABLE INTEGRATION STUDIES
• Barbados Wind & Solar Integration Study (2015)
• Vietnam Wind Grid Code Development and Renewable Integration Study (2014)
• REserviceS Project Economic Grid Support from Variable Renewables (Europe)
GE Proprietary Information
3
Introducing Variability & Uncertainty
Uncertainty
• Wind generation are not always available when called upon
• Are not dispatchable … output is predicted by a forecast
• Actual power output is different than forecast output
0
5000
10000
15000
20000
25000
0 6 12 18 24 30 36 42 48
Hours
MW
Wind Actual
Forecasted Wind
0
500
1000
1500
2000
2500
0 6 12 18 24 30 36 42 48
Hours
MW
CSP Actual
CSP Forecasted
A perfect forecast eliminates uncertainty, but there is still variability
Variability
• Wind and solar generation vary as the intensity of their energy sources
• Several timescales … minute (regulation), hour (ramping), daily, seasonal
Solar Actual
Solar Forecasted
GE Proprietary Information
4
For grid operations, wind is “similar” to load
0
5
10
15
20
25
30
35
40
45
50
0 6 12 18 24Hour
GW
Load
Wind
Load - Wind
Net Load
= Load Minus Wind (This is what must be served
by other types of generation)
• Like load, wind can be forecasted
accurately for planning purposes
• Grid operators can plan day-ahead (or
shorter) operations based on a load
forecast and a wind generation forecast
• Dispatchable generation is allocated to
serve the net of the forecast load minus
the forecast wind
• Uncertainty in the wind forecast adds to
the uncertainty in the load forecast
• Adjustments are made using hour-
ahead forecasts and real-time data
Dispatchable Generation Serves “Net Load”
GE Proprietary Information
5
• Basic options are to increase reserves, demand response, curtail, rely on neighbors, storage
• Increasing reserves
– Commit additional generation so that load will never be interrupted
– Need to do it 100% of the time, because you do not know when the reserves will be required
– Potential to increase system cost, additional capacity online may not be needed and runs the system less efficiently
• Demand response
– Interrupt or reduce load occasionally, as need arises
– A paid ancillary service
• Curtail when under forecasted
Dealing with Uncertainty
GE Proprietary Information
6
Forecasting increases economic value of renewable power
Wide-spread extreme events are predictable (e.g. widely publicized Texas events were predicted)
Forecasting can Help to Reduce Uncertainty
Texas February 24, 2007 event
Arrival of such fronts is generally forecastable, several hours ahead within a 30-minute window
Thirty-Minute Extreme Wind Drops
0
10
20
30
40
50
60
70
80
90
100
-2600 -2200 -1800 -1400 -1000 -600 -200
Wind Delta (MW)
Nu
mb
er
of
30
-Min
ute
Pe
rio
ds
5000 MW
10000 MW(1)
10000 MW(2)
15000 MW
Extreme Thirty-Minute Wind Drops
~1.5hours
~1600 MW
GE Proprietary Information
7
Timing is Everything
Conceptual Timeline for Day-Ahead Unit Commitment
15
20
25
30
35
-24 -18 -12 -6 0 6 12 18 24
12 am 12 am 12 am 12 pm 12 pm
GFS
Updates (6-hr period)
5 am
Load + Unit
Data Received
Wind
Forecast
11 am
Unit Commitment
Completed
SCUC
5 am
Morning
Load Rise
8 pm
Peak Load
29 hrs
44 hrs
Load
Hours
GW
Day of Operation
Start Times (warm): http://www.nrel.gov/docs/fy12osti/55433.pdf
GE Proprietary Information
8
Timing is Everything
Conceptual Timeline for Day-Ahead Unit Commitment
15
20
25
30
35
-24 -18 -12 -6 0 6 12 18 24
12 am 12 am 12 am 12 pm 12 pm
GFS
Updates (6-hr period)
5 am
Load + Unit
Data Received
Wind
Forecast
11 am
Unit Commitment
Completed
SCUC
5 am
Morning
Load Rise
8 pm
Peak Load
29 hrs
44 hrs
Load
Hours
GW
Day of Operation
Start Times (warm): http://www.nrel.gov/docs/fy12osti/55433.pdf
Type Minimum Up-Time
Minimum Down-Time
Hours to Start
Combined Cycle 8 8 4-8
Combustion Turbine 1 2 0.6- 2
Steam Coal 24 – 48 24 – 48 12 - 48
Steam Oil/Gas 24 – 48 24 – 48 4 - 48
Nuclear Weeks – months 168 24 - 48
GE Proprietary Information
9
Short-Term Forecasting
Using a 4-Hour forecast resulted in a $70M reduction in production cost.
The reduction was a result of an improved commitment, shifting from CT to CC’s.
CCGT increased by 2.0 TWh Coal increased by 4.7 TWh SCGT decreased by 0.2 TWh Imports decreased by 6.7 TWh
$70M
$250M
PJM Renewable Integration Study
GE Proprietary Information
10
Reduction in Spilled Energy over SOA, I30 Scenario
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
10%
2004 2005 2006 Average
Sp
ille
d E
nerg
y R
ed
ucti
on
(%
)
10% Forecast Improvement
20% Forecast Improvement
Unserved Energy, I30 Scenario
0
10
20
30
40
50
60
2004 2005 2006 Average
Reserv
e S
ho
rtfa
lls (
GW
h)
SOA Forecast
10% Improvement
20% Improvement
Avg Annual Savings over SOA
0.0
25.0
50.0
75.0
100.0
125.0
150.0
175.0
200.0
225.0
250.0
0% 5% 10% 15% 20% 25% 30%
Forecast Improvement (%)
Avg
An
nu
al O
per
atin
g C
ost
Sav
ing
s
($M
)
Pre
I10
I20
I30
Improved SOA Wind Forecast
Western Wind & Solar Integration Study
GE Proprietary Information
11
11 3/28/2016
GE Energy Consulting
Just Forecast?
Available Information
Available Resources/
Options Requirements T
i
me
H
o
r
iz
o
n
Load
Forecast
Renewable
Forecast
Forced Outages
Accuracy
Unit
Commitment
Operation
Quick Start
DSM
Energy
Spin Regulation
Schedule
.
.
Frequency
Response
Short Circuit
Ratio
Others
How should Markets factor these into decisions?
Long Term Start
Mid Term Start
Options
Em
erg
ing
T
od
ay
GE Proprietary Information
12 © 2014 GE Energy Consulting
Conclusions
• The forecast schedule used depends on the system…
o Isolated or Interconnected
o Flexible generation or inflexible base load generation
o Hydro availability and flexibility (environmental constraints)
• Forecasting improvements result in operating cost savings to the utility. These savings increase with increased wind penetration and increased forecast accuracy. o The savings is not proportional to the penetration level of wind energy
o Diminishing returns with increased penetration
• A more accurate forecast, in general, can reduce operating reserve carried by a system for uncertainty
• Forecasting improvements reduce wind curtailment and reduce reserve shortfalls, increasing the efficiency of power system operations
• Other changes in operating practices are also needed to improve operating cost savings
GE Proprietary Information
14
All power grids can accommodate substantial levels of wind and solar power… There is never a hard limit
Key lessons learned …
15 GE Energy Consulting
GE Proprietary Information
Variability and Uncertainty… Layperson’s terms
For Example… Generator Owner… “I can guarantee 1000MW of hydro all day tomorrow.”
System Operator… “OK, I will turn off 1000MW of other generation.
Variability: Generator Owner...“I can guarantee 1000MW of hydro from 2PM to 4PM tomorrow.”
System Operator… “OK, I may turn down 1000MW of other generation, rather then shutting it off.”
Uncertainty: Generator Owner… “I think I will have 1000MW of hydro sometime tomorrow.”
System Operator… “OK, I may turn off only 600MW of other generation and I will keep 400MW spinning and have quick start capacity ready to fire.”