1. 1 Pravin Varaiya Electrical Engineering & Computer Sciences, UC Berkeley Institute for Advanced Study, Hong Kong UST Selling Wind Randomly Research supported by EPRI. Thanks to Ram Rajagopal, Felix Wu, Christ Dent,Janusz Bialek, Robert Entriken, Kameshwar Poolla, Eilyan Bitar, Pramod Khargonekar.

2. 2 CA mandate: 30% clean energy in 2020?

3. 3 Outline  Some CA energy numbers  Variability of wind (and solar)  Integrating wind into current operations is very costly  Sell wind randomly  What does this get us?

4. 4 California energy numbers  Daily peak power 60GW  Bulk energy cost $52/MWh; retail price $130-$400/MWh  Wind power –CA wind purchased at ~$200/MWh –MA retail av $90/MWh; wind purchased at $200/MWh+3.5%/yr –RI retail av $130/MWh; wind purchased at $244/MWh+3.5%/yr  Ancillary services (reserves) costs in CA: –$15-$18 per MW capacity per hour for regulation –$6-$9 ($2-$3) per MW capacity per hour for spinning (non-spinning) reserve –$80-$120 MWh for real time energy  Wind integration cost EWITS estimates: $6-$8 per MWh (seems low)  Carbon tax can make fossil fuel power more expensive

5. 5 CA hopes: 20% RPS by 2012 Source: CAISO 20% RPS Total 5484 12600

6. 6 Outline  Some CA energy numbers  Variability of wind (and solar)  Integrating wind into current operations is very costly  Sell wind randomly  What does this get us?

7. 7 Normal variability in wind, solar power  Variability in single wind farm and solar PV plant

8. 8 Hourly wind power variation in CA  CA total hourly wind power varies from maximum to zero

9. 9 Not much spatial diversity  Variability not decreased by diversity over 1000 miles

10. 10 Correlation with load  Wind power and demand not positively correlated

11. 11 Variability in BPA wind power (5-min)  Wind power ramps up and down quickly and unpredictably No wind

12. 12 ERCOT wind power ramps  ERCOT 3,039MW up-ramp in 1 hour (left); 2,847MW down-ramp in 1 hour (right)

13. 13 EWITS 1-hour load and net load changes 8GW wind rated capacity 37GW peak load  Wind ramp up exceeds 2700 MW, ramp down exceeds 2700 MW, frequency > 70/8760 = 0.8% > 2.5σ; wind power has ‘fat tails’

14. 14 EWITS 10-min load and net load changes  Wind power distribution has ‘fat’ tails. Not Gaussian!

15. 15 Outline  Some CA energy numbers  Variability of wind (and solar)  Integrating wind into current operations is costly  Sell wind randomly  What does this get us?

16. 16 EWITS 20% wind: Production (ex. cap) cost  Wind (with 0 cost) displaces fuel costs from reference case 6% 20% 30%

17. 17 EWITS 20%: Total cost  Integration cost may be underestimated. New generation is to maintain system adequacy with wind 6%20%30%

18. 18 EWITS 20% wind: Integration cost ~ $5/MWh

19. 19 What is wind integration cost?  It is the extra cost of capacity and energy required to make a wind power source behave like firm power source: –reserve capacity to counter unpredictable shortfall –following and regulation capacity to counter variability –energy costs associated with following and regulation

20. 20 Bounding integration cost (1/4) + A/S ( Reserves )

21. 21 Bounding integration cost (2/4): Full info

22. 22 Bounding integration cost (3/4): No info

23. 23 Partial information

24. 24 Bounding integration cost (4/4): Reg. cost

25. 25 Summary  Cost of integrating wind may be very high, especially if prediction of wind is poor  Why not abandon attempt to make wind power look like gas plant and sell wind power as random power?

26. 26 Outline  Some CA energy numbers  Variability of wind (and solar)  Integrating wind into current operations is very costly  Sell wind randomly  What does this get us?

27. 27 Interruptible power contracts suppliercustomer tnaturesupplier

28. 28 How to design contracts, rationing

29. 29 Wind generator example s1s1 s2s2 s3s3 s4s4 generation availability

30. 30 Optimal control formulation

31. 31 Structure of optimal design s1s1 s2s2 s3s3 s4s4 1 1 2 2 3 3 4 4 w1w1 w2w2 w3w3 w4w4 generation availability

32. 32 What has this gotten us?  Integrating wind into current operations means making wind look like firm power  This may be too expensive and will require subsidies  Selling random wind requires no subsidy  May permit innovation to mitigate randomness  Can be incrementally deployed

33. 33 Open problems

34. 34 Incremental deployment 34 Permit interruptible power in distribution system and reliable power in bulk power system