SCE Smart Energy Program: 2018 Load Impact Evaluation DRMEC

SCE Smart Energy Program: 2018 Load Impact Evaluation DRMEC Workshop on the 2019 Load Impact Protocol (LIP) Report Annual Filing April 26, 2019

Agenda Smart Energy Program overview Ex post methodology Ex post results Ex ante methodology Ex ante results Conclusions and recommendations 2

Smart Energy Program (SEP) Description Load impacts are achieved through SCE-approved but customerinstalled programmable communicating thermostats (PCTs) SEP events are called by SCE and dispatched by the PCT manufacturers by remotely raising participants’ cooling setpoints Event length: 1-4 hours in duration between 11:00 AM and 8:00 PM Availability: Non-holiday weekdays, year-round Restrictions: No more than four hours of load control per day Notification: PCT manufacturers are now given 20 minutes day-of notice, rather than day-ahead – precooling no longer feasible Dual-enrollment: Customers are no longer permitted to dually enroll with SDP Customers receive sign-up and capacity payments as incentives to participate Incentives are paid as bill credits on June through September bills Energy payments are eliminated in 2019 but were in place through 2018 3

Summary of 2018 SEP Events Date Day of Week Dispatch Protocol Event Start Event Stop 6/8/2018 8/7/2018 9/13/2018 9/14/2018 9/20/2018 10/2/2018 Friday Tuesday Thursday Friday Thursday Tuesday M&E/Test Economic Economic Economic Economic Economic 2:00 PM 2:00 PM 2:00 PM 2:00 PM 2:00 PM 2:00 PM 4:00 PM 6:00 PM 6:00 PM 6:00 PM 6:00 PM 6:00 PM Heat Average Buildup Event Event Number of Avg. 12 AM Duration Temp. Customers to 5 PM ( F) ( F) 2:00 71.8 70.9 44,848 4:00 83.5 82.9 48,748 4:00 74.2 73.2 51,288 4:00 76.5 76.0 51,344 4:00 71.1 71.2 51,565 4:00 73.3 74.4 52,498 SEP Event Reference Loads Reference Load (kW) 3.5 3 08/07/18 - Tues. 09/14/18 - Fri. 09/13/18 - Thurs. 09/20/18 - Thurs. 06/08/18 - Fri. 10/02/18 - Tues. 2.5 2 1.5 1 0.5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hour 4

Ex Post Methodology Matched control group selected based on 15 proxy days close in temperature to actual event days Each customer was matched for hot, moderate, and cold event days Propensity score matching was performed to select control groups Each participant was matched to a non-participant in the same Abank and with the same CARE status Used difference-in-differences fixed effects regression model to estimate impacts Cool Moderate High 5

2018 Ex Post Impacts Heat Buildup Number of Avg. Avg. Avg. 12 Event Date Customer Reference Load w/ AM to 5 s Load (kW) DR (kW) PM ( F) Avg. Load Impact (kW) Aggregate % Load Snapback Load Impact (kW) Impact (MW) *6/8/2018 9/20/2018 9/13/2018 10/2/2018 9/14/2018 8/7/2018 44,848 51,565 51,288 52,498 51,344 48,748 70.9 71.2 73.2 74.4 76.0 82.9 0.58 1.13 1.32 1.04 1.56 2.54 0.53 0.80 0.92 0.77 1.18 1.80 0.06 0.34 0.40 0.27 0.37 0.74 9.6% 29.7% 30.1% 26.1% 24.0% 29.3% -0.03 -0.09 -0.13 -0.10 -0.11 -0.19 2.5 17.3 20.3 14.2 19.2 36.3 Average Event 51,089 75.4 1.50 1.08 0.42 27.9% -0.12 21.5 Average Event Day *6/8 TEST EVENT NOT INCLUDED IN AVERAGE EVENT CALCULATON 6

2018 Ex Post Summary Event Date 2018 vs. 2017 Ex Post Impacts Heat Buildup Avg. Avg. Load Number of Avg. Load Avg. 12 AM Reference Impact Customers w/ DR (kW) to 5 PM Load (kW) (kW) ( F) % Load Impact Snapback (kW) Aggregate Load Impact (MW) *6/8/2018 44,848 70.9 0.58 0.53 0.06 9.6% -0.03 2.5 9/20/2018 51,565 71.2 1.13 0.80 0.34 29.7% -0.09 17.3 9/13/2018 51,288 73.2 1.32 0.92 0.40 30.1% -0.13 20.3 10/2/2018 52,498 74.4 1.04 0.77 0.27 26.1% -0.10 14.2 9/14/2018 51,344 76.0 1.56 1.18 0.37 24.0% -0.11 19.2 8/7/2018 48,748 82.9 2.54 1.80 0.74 29.3% -0.19 36.3 Average Event 51,089 75.4 1.50 1.08 0.42 27.9% -0.12 21.5 % Load Impact Snapback (kW) Aggregate Load Impact (MW) *6/8 EVENT NOT INCLUDED IN AVERAGE EVENT CALCULATON 2017 Ex Post Summary Event Date Heat Buildup Avg. Avg. Load Number of Avg. Load Avg. 12 AM Reference Impact Customers w/ DR (kW) to 5 PM Load (kW) (kW) ( F) 7/31/2017 36,066 76.5 1.83 1.26 0.57 31.1% -0.15 20.6 6/19/2017 31,036 76.9 1.86 1.23 0.63 33.7% -0.16 19.4 6/21/2017 25,965 77.5 1.94 1.33 0.60 31.1% -0.13 15.6 9/5/2017 39,141 77.7 1.90 1.32 0.58 30.4% -0.16 22.7 6/20/2017 31,231 78.4 2.02 1.39 0.64 31.5% -0.19 19.9 7/6/2017 32,792 79.3 2.07 1.40 0.67 32.3% -0.13 21.9 SEP performed nearly identically in 2018 vs. 2017 Same % load reduction Same MW However, that is despite a large increase in the number of enrolled customers 2018 SEP events were much cooler on average than 2017 Mean17 of 75.4 vs 81.4 ( F) Cooler temperatures lead to lower reference loads and lower load impacts 8/1/2017 30,333 79.7 2.18 1.49 0.69 31.8% -0.19 21.0 8/28/2017 32,515 80.3 2.46 1.78 0.68 27.6% -0.13 22.1 8/2/2017 35,408 82.2 2.08 1.64 0.44 21.0% -0.21 15.4 Ref. load 1.50 kW vs. 2.31 kW Load imp. 0.42 kW vs. 0.64 kW 8/3/2017 36,460 83.1 2.33 1.63 0.70 30.0% -0.28 25.5 8/29/2017 38,953 83.7 2.71 1.98 0.73 26.8% -0.29 28.3 7/7/2017 32,916 84.1 2.53 1.85 0.68 26.7% -0.27 22.2 8/30/2017 39,031 85.1 2.77 2.04 0.73 26.3% -0.24 28.4 8/31/2017 37,160 86.2 2.76 2.09 0.67 24.2% -0.18 24.8 9/1/2017 32,799 87.8 2.94 2.31 0.63 21.4% -0.20 20.7 Avg. 2017 Event 34,120 81.4 2.31 1.66 0.64 27.8% -0.21 21.9 7

Ex Ante Assumptions Enrollment Enrollment expected to increase from 52,795 in October 2018 to 192,970 in 2029 This year forecast 250,000 200,000 150,000 100,000 50,000 0 t, Oc 1 20 8 ( Last year forecast al) 019 020 021 022 023 024 025 026 027 028 029 tu 2 2 2 2 2 2 2 2 2 2 2 c A Year The evaluation specifically projects load impacts expected to be observed during the resource adequacy (RA) window The summer RA window has shifted from 1-6 PM to 4-9 PM The winter RA window is remains unchanged at 4-9 PM RA Window Hour Ending 14 15 16 17 18 19 20 21 Previous RA Window (Apr - Oct) Previous RA Window (Nov - Mar) Current RA Window 8

Ex Post Load Impacts vs. Weather We estimate ex ante load impacts for each hour as a function of heat buildup We note a behavioral component unique to SEP that we don’t see in SCE’s direct load control program (SDP): Downward trend in magnitude across the four event hours Downward trend in weather responsiveness across the four event hours Event fatigue, potentially occupancy? SDP does not show this strong an effect and they would be likely be subject to the same occupancy patterns. 9

Per customer load impacts – SCE 1-in-2 and 1-in-10 Weather Conditions (kW) Month SCE Weather Year 1-in-2 1-in-10 May June July 0.45 0.69 0.50 0.64 0.53 0.66 August September October 0.66 0.73 0.60 0.79 0.59 0.65 Aggregate load impacts – SCE 1-in-2 and 1-in-10 Weather Conditions (MW) – 2019 Enrollment Month SCE Weather Year 1-in-2 1-in-10 May June July 28.54 43.63 31.58 40.55 33.94 42.15 August September October 42.13 46.42 37.81 50.22 37.21 41.38 10

Typical Event Day Aggregate Impacts for SCE 1-in-10 Weather Conditions (MW) Avg. Aggregate Impact (MW) Forecasted Enrollment 140 250000 120 MW Impact 100 80 150000 60 100000 40 50000 20 0 # of Customers 200000 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 0 Forecast Year 11

Comparison to PY 2017 ex ante load impacts – per customer impacts for SCE 1-in-10 Weather Conditions (kW) The PY 2017 evaluation estimated SEP as capable of delivering 163 MW on a SCE 1-in-10 September peak day The PY 2018 evaluation estimated 153 MW for the same scenario The difference reflects the net effect of three changes: RA window change Enrollment forecast New PY 2018 ex post data 12

RA Window Change By moving the summer RA window later in the day, there’s more load to shed (example here is SCE 1-in-10) : Current RA Window Previous Previous RA Window Current Ref. Load 3.5 Hour 3 Load (kW) 2.5 2 1.5 1 0.5 0 1 2 3 4 5 Hour Ending 14 15 16 17 18 19 20 21 Avg. Hour in RA Window Avg. Customer kW Impact Current Previous RA RA Window Window 0.8 1.0 0.7 0.5 1.3 0.4 0.9 0.6 0.4 0.4 0.70 0.73 13

Enrollment and New Information Changes The PY 2017 evaluation projected 229,000 enrolled customers in 2028; the PY 2018 evaluation projects 192,970 enrolled customers in 2029. The PY 2018 events added impacts to the ex ante estimation dataset that were much cooler than the PY 2017 events – the new relationship was steeper (lower impacts at cooler temperatures and higher impacts at hotter temperatures) 14

Recommendations and Conclusions SCE should conduct market research to understand the characteristics, sentiments, and behaviors of new SEP customers: How are the characteristics, sentiments and behaviors of SEP participants changing over time? Call events under varying circumstances: Variety of time periods: help establish relationship between load impacts and time of day, particularly during the new RA window. Variety of weather conditions: calling events on cooler summer days will provide valuable data points for ex ante estimation Vendors should provide override information: Help identify conditions under which customers opt out of events 15

Contact Us Candice Potter Principal, Utility Services [email protected] (415) 640-2955