Resource Adequacy Forecast Adjustments Allocation Methodology Miguel Cerrutti

  • Slides: 15
Download presentation
Resource Adequacy Forecast Adjustment(s) Allocation Methodology Miguel Cerrutti Demand Analysis Office Energy Assessments Division

Resource Adequacy Forecast Adjustment(s) Allocation Methodology Miguel Cerrutti Demand Analysis Office Energy Assessments Division R. 14 -10 -010 Workshop California Public Utility Commission San Francisco, February 9, 2015

Outline The challenges Year-ahead load forecast adjustments Coincident factor (CF) adopted CF adjustment methodology

Outline The challenges Year-ahead load forecast adjustments Coincident factor (CF) adopted CF adjustment methodology Weather normalization (WN) and short-term load forecasting (STLF) Improvements

Challenges Arrive at LSE-specific final year-ahead load forecasts for RA compliance Assign a value

Challenges Arrive at LSE-specific final year-ahead load forecasts for RA compliance Assign a value for each LSE’s contribution to CAISO peak loads Forecast weather normalized short-term peak loads for IEPR (summer) and RA (monthly) Ensure a transparent and repeatable process with well-supported and consistent key assumptions with RA and CEC

Year-ahead load forecast time line LSEs file historical load data LSEs file yearahead load

Year-ahead load forecast time line LSEs file historical load data LSEs file yearahead load forecast Final date to file yearahead load forecast changes LSEs receive initial year -ahead allocations March 20 th April 24 th Year-ahead compliance filings due July 31 st LSEs receive final year-ahead allocations August 19 th October 30 st September 18 th

Year-ahead forecast adjustments Coincident adjustment – LSE-specific peak load contribution at time of CAISO’s

Year-ahead forecast adjustments Coincident adjustment – LSE-specific peak load contribution at time of CAISO’s monthly peak load Plausibility adjustment – reconcile aggregate LSEs monthly peak load forecasts against CEC’s monthly WN STLF for IOU service areas Prorated adjustments to LSEs forecasts to account for demand side energy savings paid for through distribution charges Pro rata adjustment to match CEC forecast within 1%

Coincident factor (CF) adjustment - CPUC adopted D. 12 -06 -025 Coincident Factor O.

Coincident factor (CF) adjustment - CPUC adopted D. 12 -06 -025 Coincident Factor O. P. 4 “The resource adequacy program shall be modified so that the coincidence adjustment factor uses a load service entity-specific coincidence adjustment factor for annual resource adequacy requirements, and an energy service provider-composite coincidence factor for monthly resource adequacy requirements, as follows: *Annual Resource Adequacy Requirements – The California Energy Commission will calculate a Load Serving Entity-specific coincidence adjustment factor using Load Serving Entity hourly loads; and *Monthly Resource Adequacy Requirements – The California Energy Commission will calculate an Electric Service Provider-composite coincidence factor, which would be applied to each Electric Service Provider’s migrating load for the month; migrating load for community choice aggregators would be treated separately. ”

Coincident factor (CF) – the data CAISO’s EMS hourly load data (across 1 -3

Coincident factor (CF) – the data CAISO’s EMS hourly load data (across 1 -3 years) five highest monthly CAISO system peak hours LSE hourly load data (across 1 – 3 years) monthly non-coincident peaks Average hourly peak loads Weather data Weather normalized daily LSE and system peaks

Coincident factor (CF) - the process LSEs coincident peaks associated with the monthly five

Coincident factor (CF) - the process LSEs coincident peaks associated with the monthly five highest CAISO system peak hours Monthly CF as a median over the ratios of the five LSE’s coincident peaks to its non-coincident peak Include peak producing days – typical weather Monthly CF to develop LSEs peak forecasts coincident with the CAISO system peak hours

Coincident factor (CF) - the process … continuation LSE’s with stable load shapes and/or

Coincident factor (CF) - the process … continuation LSE’s with stable load shapes and/or correlated with system loads one year of current load data LSEs with unstable load shapes and/or not correlated with system loads at least three previous years of data average hourly peak loads LSEs with slightly higher load responses to more than normal weather patterns WN CF - daily time-series regressive model to normalize daily LSE and CAISO system peaks

Coincident factor (CF) - the process … continuation Review and validity assessment Small sample

Coincident factor (CF) - the process … continuation Review and validity assessment Small sample problems no days closer to one-in-two conditions Over time inconsistent loads so unstable coincidence patterns – meaningless statistics Monthly load migration CF for aggregate of ESPs

Coincident factor (CF) - the process … continuation LSE Moy CF 3 CP CF

Coincident factor (CF) - the process … continuation LSE Moy CF 3 CP CF 5 CP CF Avg WN CF CP / WN CP NCP / WN NCP LSE 1 10 . 528 . 937 . 841 LSE 2 11 . 920 . 868 . 789 LSE 3 8 . 605 . 718 . 802 LSE 4 6 . 720 . 719 . 842 LSE 5 12 . 674 . 859 ESP 7 . 695 . 937 All ESP 7 . 923 . 897 . 884 LSE 8 7 . 836 . 897 . 853 1. 166 1. 143 LSE 9 8 . 895 . 845 . 904 1. 277 1. 265 LSE 10 6 . 636 . 789 . 916 1. 438 . 998 LSE 11 8 . 978 . 914 . 782 . 842 1. 052 LSE 12 3 . 612 . 765 . 978 1. 341 . 839

Coincident factor (CF) - benefits Better information with well reasoned-analysis suggests a more appropriate

Coincident factor (CF) - benefits Better information with well reasoned-analysis suggests a more appropriate LSEs CF Accurate CF improves cost allocation Provides a realistic (as possible) LSE-specific CF without unfairly impacting the CFs of other LSEs Once a CF is assigned, it is considered fixed and is not changed CF is only corrected if it is found to be in error due to data filing or calculation errors

Weather normalized (WN) short-term load forecasting (STLF) WN STLF is used to reconcile the

Weather normalized (WN) short-term load forecasting (STLF) WN STLF is used to reconcile the aggregate LSEs year-ahead forecasts in each IOU area for RA compliance (plausibility adjustment) Inputs to WN STLF most current IEPR (e. g. , for 2016 RA, 2014 IEPR update) four years of CAISO hourly EMS data hourly demand response impacts 30 years weather conditions

Weather normalized (WN) short-term load forecasting (STLF) – the process First time-series regressive modeling

Weather normalized (WN) short-term load forecasting (STLF) – the process First time-series regressive modeling prior three years selecting functional form and explanatory effects using sample analysis (current year) Second time-series regressive modeling last three years estimating peak load sensitivities to selected effects Monte Carlo probabilistic simulation peak load sensitivities and 30 years weather one-in-two WN STLF for IEPR and one-inten (extreme weather) for CAISO’s LCR

Improvements Improving allocation of DR events and nonevents to hourly loads, LSE’s year-ahead forecasts,

Improvements Improving allocation of DR events and nonevents to hourly loads, LSE’s year-ahead forecasts, and CEC’s forecasts unclear whether or not DR impacts are embedded in LSE’s historic hourly loads and year-ahead forecasts LSEs need to provide additional information about the extent and type of DR embedded in the hourly and forecast data For transparency, there will be an attempt to post the monthly five highest CAISO system coincident peak load hours