1. Analysis of Demand Response Modeling in GridView Andy Satchwell and Sarah Smith Modeling Work Group December 21, 2015 The work described in this presentation was funded by the National Electricity Delivery Division of the U.S. Department of Energy’s Office of Electricity Delivery and Energy Reliability under Lawrence Berkeley National Laboratory Contract No. DE-AC02-05CH11231.

2. Purpose of today’s presentation Present analysis of GridView’s DR modeling capabilities Comparison of GridView and LBNL modeling approaches Provide recommendation for DR modeling in TEPPC studies 2

3. DSM Inputs to Western Regional Planning LBNL has worked with WECC staff and the State and Provincial Steering Committee (SPSC) over the past four years to develop DSM-related assumptions and modeling inputs for WECC’s regional transmission planning studies Two types of DR modeling assumptions required for each study case: –DR resource quantities: How much DR is available to be dispatched in any given hour for each load zone? –DR dispatch mechanics: When is the DR dispatched and how does it affect hourly loads and peak demand? DR resource quantities are based on non-firm load forecasts reported by balancing authorities to WECC 3

4. 2024 Common Case DR Capacity by BA and Program Type DR resource is growing in size and importance in the West for maintaining system reliability with generating plant retirements and for renewables integration

5. Context for analysis and recommendation LBNL previously implemented a modeling approach using its Demand Response Dispatch Tool (DRDT) premised on PROMOD’s then limitations for realistically modeling DR WECC transitioned to a new production cost modeling tool (GridView) and we need to revisit endogenous modeling capabilities to select most realistic and appropriate modeling approach DRDT requires iteration with modeling runs and may not be most efficient from process standpoint –Question is whether gain in efficiency from endogenous approach comes at too much of a cost in terms of reduced accuracy of how DR is modeled 5

6. What do we mean by “realistic” dispatch? DR programs are used for reliability and economic purposes –Several program types are utilized during high price periods (e.g., critical peak pricing) DR program dispatch is limited by tariff provisions specifying maximum number of events per month or year DR program tariffs also specify multiple, sequential blocks (e.g., 4 to 6 hours) for events 6

7. Brief background on LBNL’s DRDT 7 Inputs Hourly Load Hourly LMPs Maximum Available Monthly DR Program constraints Resource Availability Calculate “hourly shaping factors” to scale maximum available DR to hourly load Simulated Dispatch Identify top- LMP hours to act as dispatch trigger Dispatch DR over top-LMP hours, subject to program constraints Output 8760 load- modifying profile of DR used in production cost model as static profile

8. Overview of GridView’s DR modeling approaches Discussions with GridView revealed there was no standard approach to modeling DR Most common approach was to use an energy limited hydro proxy with a pre- established trigger price (e.g., $150/MWh) 8

9. Summary of approach Compared LBNL and GridView modeling approaches based on GridView’s pre- existing capabilities Worked with ABB staff to improve DR modeling in GridView Conducted a separate comparison based on changes Developed a recommended approach 9

10. Summary of approach Compared LBNL and GridView modeling approaches based on GridView’s pre- existing capabilities Worked with ABB staff to improve DR modeling in GridView Conducted a separate comparison based on changes Developed a recommended approach 10

11. Comparative analysis of GridView Hydro approach versus LBNL DRDT approach 11 How was the hydro proxy constructed? Used LBNL DRDT profiles to set monthly energy amounts and used 2024 Common Case maximum DR capacities Approach did not distinguish between DR program types and did not use program constraints (e.g., expected hours of dispatch) Dispatch was triggered in high load periods

12. Comparative analysis of GridView Hydro approach versus LBNL DRDT approach 12 How similar are the two approaches? Both approaches dispatched same maximum amount of DR In some cases, LBNL approach had higher average load reductions, and opposite was true in other cases GridView hydro approach minimum dispatched amounts were much lower than LBNL DRDT approach in several cases

13. Comparative analysis of GridView Hydro approach versus LBNL DRDT approach 13 How realistic is the GridView hydro approach? Almost half the dispatch events (196 out of 401) were for 1 to 2 hour blocks which differs from typical DR program utilization of blocks greater than 2 hours GridView hydro approach had more, smaller dispatch events than LBNL approach, and more dispatch amounts <1 MW

14. Summary of approach Compared LBNL and GridView modeling approaches based on GridView’s pre- existing capabilities Worked with ABB staff to improve DR modeling in GridView Conducted a separate comparison based on changes Developed a recommended approach 14

15. Building DRDT in GridView GridView’s hydro approach did not appear to be an improvement over LBNL DRDT approach With LBNL’s direction, GridView modelers developed a new algorithm that built DRDT approach into GridView and reflecting typical program dispatch behavior We also enhanced the core dispatch approach in order to allow programs to be dispatched based on load or price, depending on whether they are typically used for system reliability or economic purposes 15

16. Summary of approach Compared LBNL and GridView modeling approaches based on GridView’s pre- existing capabilities Worked with ABB staff to improve DR modeling in GridView Conducted a separate comparison based on changes Developed a recommended approach 16

17. DRDT Test Cases 17 CaseDR CapacitiesProgram Dispatch ConstraintsDispatch Trigger LBNL DRDT 2024 Common Case Interruptible: 2-hour blocks, 10 times per year Direct load control (DLC): 4- hour blocks, 10 times per year Pricing: 5-hour blocks, 10 times per year Load as a capacity resource (LCR): 6-hour blocks, 10 times per year All programs dispatched on highest average price periods GridView DRDT #1 Same Same as LBNL DRDT and minimum dispatch set at 1MW All programs dispatched on highest average load periods GridView DRDT #2 Same Same as GridView DRDT #1 Interruptible and DLC programs – highest average load periods and load is “resorted” after each dispatch event Pricing and LCR programs – highest average price periods

18. Comparative analysis of DRDT Test Cases (1) All three test cases showed similar distributions of event durations GridView DRDT test cases had larger DR amounts and somewhat more frequent dispatch events – likely due to high load hour dispatch trigger and coincidence with hourly shaping factors 18 Represent back-to-back dispatches

19. Comparative analysis of DRDT Test Cases (2) GridView test cases have hourly shapes that better match seasonal system load due to load dispatch trigger and, thus, increased shaping factor GridView case #1 dispatch was most coincident with system load but does not reflect economic dispatch of some DR program types (e.g., pricing and LCR) 19 LBNL DRDT CaseGridView DRDT #1 CaseGridView DRDT #2 Case

20. Summary of approach Compared LBNL and GridView modeling approaches based on GridView’s pre- existing capabilities Worked with ABB staff to improve DR modeling in GridView Conducted a separate comparison based on changes Developed a recommended approach 20

21. Recommendation We recommend implementing the DRDT approach used in GridView Case #2 for the 2026 TEPPC study cases Use same program constraints as in 2024 TEPPC study cases and use highest average load periods as dispatch trigger for Interruptible and DLC program types 21

22. Questions? Project Team: Andy Satchwell | asatchwell@lbl.gov | 510-486-6544 Sarah Smith| sjsmith@lbl.gov | 510-486-6142 Publications: https://emp.lbl.gov/reports/resp 22