1. Toward optimization of a wind/ compressed air energy storage (CAES) power system Jeffery B. Greenblatt Samir Succar David C. Denkenberger Robert H. Williams Princeton University, Princeton, NJ 08544 Guyot Hall, (609) 258-7442 / 7715 FAX, jgreenbl@princeton.edu Electric Power Conference, Baltimore, MD, 30 March – 1 April 2004 Session 11D (Wind Power II), 1 April 2004 Foote Creek Rim, Wyoming

2. Does wind power need storage? Three contexts: 1.Make wind dispatchable (price arbitrage; potential at small market share) 2.Boost wind capacity factor at large market penetration (offsets fuel cost only) 3.Exploit high-quality but remote wind resources (by reducing transmission costs) Time Power Time Market share Value Few markets currently exist

3. Electric storage options Technology Compressed Air Energy Storage (CAES) (350 MW) Pumped hydroelectric Advanced battery (10 MW) Flywheel (100 MW) Superconductor (100 MW) 370 1100 2100 6200 6100 Cost of 20 hrs. storage ($/kW) Capacity ($/kW) Storage ($/kWh) 350 900 120 150 120 1 10 100 300 Source: Schainker, 1997 (reproduced in PCAST, 1999) CAES is clear choice for: Several hours (or more) of storage Large capacity (> ~100 MW)

4. Compressor trainExpander/generator train Fuel (e.g. natural gas, distillate) CAES system Intercoolers Heat recuperator PCPC PGPG Air Exhaust Air Storage Aquifer, salt cavern, or hard mine h S = Hours of Storage (at P C ) P C = Compressor power in P G = Generator power out

5. A wind/CAES model Wind farm Transmission CAES plant Underground air storage For this application CAES is needed to provide baseload power P WF = Wind Farm max. power out (rated power) P T = Transmission line max. power P WF PTPT

6. Research objectives What is optimal wind/CAES system for baseload power transmission? What is optimal capacity factor (CF) of that transmission line? How much will such a system cost, and can it compete against other baseload systems (nuclear, coal, natural gas)? Note: Costs of system components were not available in time for the Feb. 2 deadline. If component costs can be obtained, a cost optimization will be presented at the conference.

7. Key parameters Size of CAES generation relative to transmission line (P G /P T ) CAES compression/generation ratio (P C /P G ) Relative size of wind farm (P WF /P T ) CAES storage time relative to wind autocorrelation time (h S /h A ) Ratio of turbine speed rating to resource wind speed (v rate /v avg ) Comp Gen v rate v avg. hShS hAhA

8. Secondary parameters CAES electricity output/input ratio (E o /E i ) Wind turbine array spacing (xD 2 ) Weibull shape parameter (k) and wind power density (P wind ) EiEi EoEo

9. Wind farm simulation Weibull dist. Wind speed Probability Wind speed time series Autocorrelation time (h A ) Time Wind speed Power curve Wind speed Power Losses P WF Time Wind speed Wind power time series Rated power (k 2 > k 1 ) } Power “lost”

10. CAES model Compressor Generator Spilled power (if storage full) Fuel P WF Transmission line capacity CAES capacity Spilled power Air X PGPG Total system output (≤ P T ) Direct output (≤ P T ) Losses CO 2 Transmission losses PCPC Air storage hShS

11. Base case configuration Wind farm: P WF = 2 P T (4000 MW) Spacing = 50 D 2 v rated = 1.4 v avg Transmission: P T = 2000 MW Comp Gen P C = 0.85 P T (1700 MW) CAES system Wind resource: k = 3, v avg = 9.6 m/s, P wind = 550 W/m 2 (Class 5) h A = 5 hrs. System CF = 0.80 E o /E i = 1.30 P G = 0.50 P T (1000 MW) h S = 10 hrs. (at P C )

12. Compressor and generator sizes P C /P T 1 0 1 P G /P T CF = 81% CF = 76% CF = 68% CF = 72% Cut along constant P G /P T : 0.5 1.5 CF P C /P T Base case CF improves (with diminishing returns) as either P C /P T or P G /P T increases Base case

13. Compressor/generator ratio P C /P T 1 0 1 P G /P T CF = 81% CF = 76% CF = 68% CF = 72% Base case Slope ~ 1.7 For given CF, least cost configuration appears to lie along slope line Minimal increase in CF for P G /P T = 0.5  1 Slope expected to be controlled by P WF /P T and turbine rating 0.5 1.5 Max. CF = 85%

14. Wind farm parameters Some improvement at large P WF /P T, but most improvement at P WF /P T ≤ 2 Small change in CF with array spacing Array spacing (D 2 ) Base case CF P WF /P T (oversizing) Base case P WF = P T case

15. Storage vs. autocorrelation time 100 10 0.1 1 Storage time (h S ) (hrs. log scale) Autocorrelation time (h A ) (hrs. log scale) 0.1110100 Base case CF = 70% CF = 79% CF = 74% CF = 65% No improvement in CF if h S >> h A or vice-versa h A (hrs. log scale) CF Cut along constant h S : Base case h S = h A case

16. Power derating Wind speed Power v rate = 1.8v avg Wind speed Probability- weighted power Wind turbine power curve v rate = 1.4v avg 7% above rated speed v rate = 1.0v avg Wind speed 36% Probability- weighted power Wind speed 72% Probability- weighted power As v rate decreases, turbines run at rated (maximum) power more of the time CF increases, but rated power decreases, so more turbines needed for same P WF

17. 0.6 0 P G /P T 11.52 v rate /v avg 0.4 0.3 0.2 0.1 0.5 CF = 80% CF = 60% CF = 40% CAES generation vs. turbine rating Base case (“large CAES”) Large v rate /v avg Alternative case (“small CAES”): Small v rate /v avg Small CAES case may be more economical if  (COST WT N WT ) +  COST CAES < 0 Alternatively, P WF /P T could be increased (may be more expensive)

18. Dependence on E o /E i CF Base case E o /E i Little change in CF with CAES efficiency

19. Wind resource parameters CF P wind (W/m 2 )Weibull k Base case Virtually no change in CF over P wind = 200-1000 W/m 2 (classes 2-7+) CF trend with k depends strongly on v rate /v avg v rate /v avg 1.0 1.4 1.8

20. Conclusions Capacity factor (CF) of 80% is achievable for our base case: P WF /P T = 2P G /P T = 0.5P C /P G = 1.7 h S = 10 hspacing = 50 D 2 v rate /v avg = 1.4 Base case is somewhat improved by increasing P WF /P T, P G /P T or array spacing, but all likely to be expensive Optimal storage time (h S ) should be somewhat larger than the wind autocorrelation time (h A ) Gen hShS hAhA > Base case CF = 80%

21. Conclusions (cont’d) Comparable CF is achieved by reducing CAES system size and rating turbines lower (alternatively, P WF /P T could be increased but this is probably more expensive). Dependence of CF on k is coupled to turbine rating, with CF increasing with k for lower v rate /v avg, and decreasing for higher v rate /v avg. Changing E o /E i, P wind has little effect on CF. EiEi EoEo + CAES size

22. Acknowledgments Dennis Elliott, Michael Milligan, Marc Schwarz, and Yih-Wei Wan, NREL Al Dutcher, HPRCC Marc Kapner, Austin Energy Nisha Desai, Ridge Energy Storage Bob Haug, Iowa Municipal Utilities District Paul Denholm, University of Wisconsin, Madison Joseph DeCarolis, Carnegie Mellon University Al Cavallo, NIST