1. Climate and Energy in California David W. Pierce Tim P. Barnett Eric Alfaro Alexander Gershunov Climate Research Division Scripps Institution of Oceanography La Jolla, CA

2. Determine the economic value of climate and weather forecasts to the energy sector Scripps Inst. Oceanography University of Washington Georgia Inst. Tech California Energy Commission California ISO PacifiCorp San Diego Gas & Elec. SAIC Academia State Partners Industrial Partners

3. Climate & weather affect energy demand Source: www.caiso.com/docs/0900ea6080/22/c9/09003a608022c993.pdf

4. …and also energy supply Green et al., COAPS Report 97-1 Typical effects of El Nino: CA hydro

5. What we’re going to show you Four case studies, each done with a different utility What were the successes? What were the issues? –Policy, regulatory, business issues From your perspective, how to address these issues?

6. 1. California "Delta Breeze" An important source of forecast load error (CalISO) Big events can change load by 500 MW (>1% of total) Direct cost of this power: $250K/breeze day (~40 days/year: ~$10M/year) Indirect costs: pushing stressed system past capacity when forecast is missed!

7. NO delta Breeze Sep 25, 2002: No delta breeze; winds carrying hot air down California Central valley. Power consumption high.

8. Delta Breeze Sep 26, 2002: Delta breeze starts up; power consumption drops >500 MW compared to the day before!

9. Weather forecasts of Delta Breeze 1-day ahead prediction of delta breeze wind speed from ensemble average of NCEP MRF, vs observed.

10. Statistical Forecast Shown as percent of cases. RED are errors, BLUE are correct forecasts.

11. Delta Breeze summary Using climate information can do better than dynamic weather forecasts Possible savings of 10 to 20% in costs due to weather forecast error. Depending on size of utility, could be in range of high 100,000s to low millions of dollars/year.

12. Issues Had a motivated, technically savvy stakeholder – what to do when we don’t? Releasing cost data is a very sensitive issue – yet we can’t be specific about the forecast value without it!

13. 2. Load demand management Induce customers to reduce electrical load on peak electrical load days Prediction challenge: call those 12 days, 3 days in advance Amounts to calling weekdays with greatest "heat index" (temperature/humidity)

14. Why shave peak days? http://www.energy.ca.gov/electricity/wepr/2000-07/index.html

15. Price vs. Demand http://www.energy.ca.gov/electricity/wepr/1999-08/index.html

16. What can climate analysis say?

17. Peak day electrical load savings If knew electrical loads in advance: 16% With event constraints: 14% If knew temperature in advance: 11% Super simple scheme (24C, 0.5): 6% (Load is relative to an average summer afternoon)

18. Peak day summary Might (?) ultimately be a real-time program –Driven by "smart" electric meters –Main benefit would be avoided cost of peaker generation plants ~$12M/yr. Until then, climate prediction: –Far less deployment cost –Cost of avoided procurement ~$1.3M/yr

19. Issues With this new information, stakeholder (now more informed) found they needed something different – yet the program ended SDG&E were open to change – but not everyone is Probabilistic forecasts means you would be changing to something that will not always be right – and yet is better than previous

20. 3. Irrigation pump loads Electricity use in Pacific Northwest strongly driven by irrigation pumps When will the pumps start?

21. Irrigation pump electrical use

22. Pump start date Eric Alfaro, SIO

23. Total load affected by soil moisture Eric Alfaro, SIO

24. Irrigation load summary Buying power contracts 2 months ahead of a high-load summer saves $25/MWh (over spot market price) Use: about 100,000 MWh Benefit of 2 month lead time summer load forecast: $2.5 M

25. Issues Product they asked for wasn’t what they really wanted – start date vs. total summer usage. How to transfer technology to operational users?

26. 4. NPO and winter heating

28. Issues Couldn’t find a willing stakeholder for this one! Regulations constrain utilities’ actions; how can this information be used within that framework? Again, the probabilistic nature of forecasts. You will be wrong sometimes, but better in the long run. What time horizon is acceptable for “long run”?

29. Economic value of climate forecasts to the energy sector Just the few case studies here indicate $10’s M/year. But, need the following for these practices to be adopted: –Stakeholders who are willing to change –Business models that can make use of probabilistic information – it will be wrong sometimes, but better in the long run –Time and relationship with stakeholder that allows iterating –Suggestions that work within the regulatory framework –Whose role is it to produce continuing, operational forecasts? –How to get quality energy data when it’s considered proprietary? (Example: FERC energy data set problems)