1. The Economics of Commercial Demand Response for Spinning Reserve Michael Fisher, Jay Apt, Fallaw Sowell USAEE North American Conference October 26, 2015

2. What Does Reserve Do? 2 Returns frequency to nominal in the case of a contingency Source: Kirby 2003

3. What Drives the Cost of Reserve? Extra power plants kept online for idle generation Also: Part-load efficiency penalty Increased maintenance costs 3 Figures adapted from Hummon et al. (2013) Lower cost units backed down to allow higher cost peaker units to run part-loaded and provide reserve

4. CAISO Ancillary Services Spending 4 Source: CAISO 2013 Annual Report on Market Issues and Performance

5. Demand Response (DR) in California FERC Order 719 (2008) mandates that ISOs allow direct participation by retail customers in energy and ancillary service markets WECC still does not allow DR to participate in spinning reserve, only non-spinning reserve and regulation –ERCOT, PJM, NYISO, MISO, ISO-NE (2017) all allow it Commercial load ~50% of CAISO load 5

6. But, Can The Business Case Be Made? StudyCustomer Segment ServicesEnd- Uses Consider Actual Costs? Matched Time- Varying Resource with Market Prices? Kirby (2003)Resregulation reserve A/CYesNo Mathieu/ Callaway et al. (2009-2014) Resenergy regulation A/CNo PG&E Pilot (2009) C&IreserveHVAC Process Yes (#Participants = 3) Rubinstein et al. (2010) C&Iregulation reserve LightingNo MacDonald et al. (2014) C&Ireserve regulation HVAC Lighting No Hummon et al. (2013) VariousAllVariousNoYes Hao et al. (2014)C&IregulationHVACNo 6

7. Research Questions In the absence of incentives, can commercial DR participants make money in spinning reserve in California if regulatory barriers are removed? Are certain end-uses, building segments, or geographical regions better positioned to make money? Should California subsidize the equipment necessary for DR to participate in spinning reserve? 7

8. Automated DR Cost Data 8 Telemetry cost ~$50/kW for large commercial (Kiliccote 2014) Customer incentives discussed later… Note: AutoDR incentive data includes industrial customers

9. Resource Availability Assume DR resource availability is proportional to load California Commercial End-Use Survey (CEUS) provides load profiles 9

10. 10 Zones (12) Building Segments (12) End-Uses (13) 1,872 unique hourly profiles Disaggregation

11. End-Use Profile Adjustments Removed end-uses inappropriate for reserve –e.g. exterior lighting (code issues) Converted standardized load profiles to 2011-2013 profiles to match market prices –Weather-dependent end-uses  regression model –Non-weather dependent end-uses  day matching 11

12. Steady End-Uses are Best 12

13. The Business Case is Marginal 13 Incentives for participation would push average paybacks past business thresholds, but… South CA, Lodging, Int. Lighting

14. Niche Applications Can Still Make Money 14 0.15 0.10

15. Policy Implications Market prices do not include damages from emissions  Does the dispatch of reserve increase emissions?  If so, do the associated damages justify incentives to enable spinning reserve capability for DR? 15

16. First-Order Damage Estimate Only consider carbon damages Conservatively assume natural gas serves the margin and base load 16 Figure adapted from Katzenstein 2010 % of annual hours that plants are held online for reserve # power plants shutdown savings Combined-Cycle Emissions Profile Unload marginal reserve generator Re-load base load generator

17. Carbon Damages May Justify Telemetry Incentives 17 SCC = Social Cost of Carbon AB32 refers to California’s cap-and-trade program Telemetry cost calculated at $50/kW Annual CO 2 Savings ≈ 1 million tonnes (0.2M – 2.8M)

18. Telemetry Incentive Improves Economics 18

19. Conclusions WECC should fully integrate DR into ancillary services Absent IOU incentives, an aggregator can still make money implementing AutoDR for spinning reserve in niche applications Constant loads are better for spinning reserve than seasonal loads (e.g., lighting vs cooling) Carbon emission damages may justify additional incentives for telemetry 19

20. Acknowledgments We thank the Carnegie Mellon Electricity Industry Center for its support 20