Loss of Load Probability (LOLP) Project October 16, 2009 Resource Adequacy Technical Committee Meeting
BPA LOLP Project Joint BPA / Council Effort: Genesys (generation evaluation system) was developed by the Council in the mid to late 1990’s Genesys is a Monte Carlo (stochastic) model of the PNW power system; with a hourly dispatch function; one game = 1 fiscal year BPA has been working with the Council over the past 18 months to make improvements to the Genesys model Federal Genesys - uses the same platform as the Council Genesys BPA specific resource/demand issues (i.e. slice, regulating reserves, Federal loads and resources etc.) are handled either through a series of switches or input data Monte Carlo Events: Hydro Generation 70 Water Years – for Federal Hydroelectric Projects both regulated and hydro independent plants Wind Generation: 40 capacity factor tables (each 8,760 hours) – for 253 MW of wind projects serving load based on historical wind data Temperature Sensitive Load: 58 temperature traced loads 1949 to 2006 (each 8,760 hours) – of the load following customers Forced Outages (Thermal Units): CGS and Wauna Cogeneration (expires July 2016) – must run resources with random forced outages (FOR) All Events Assumed Independent *162,400 possible events (excluding FOR)* BPA’s Loss of Load Probability (LOLP) Project2
3 BPA LOLP Project (Continued) Joint BPA / Council Effort: (continued) Data completed for 2013 – Regional Dialogue contracts implemented Slice logic switched ‘on’ – 27% of generation is sent to slice “account”; does not model slice customer behavior to take more or less capacity 40 capacity factor tables for wind based on 6 years of historical wind data from 5 farms – considered independent from temperature Forced Outages (Thermal Units): (continued) Federal sustained peaking table (i.e. 2, 4, 10 hour max capacity table and min generation) developed from HOSS Sustained peaking table includes the affect of holding ‘DEC’ for regulating reserves Federal ‘INC’ regulating reserves incorporated into reserve requirement (also includes the contingency reserves) BPA’s Loss of Load Probability (LOLP) Project
4 Genesys Output – How to Build Metrics and Thresholds? Reliability record: Shows hourly curtailment and reserve violations for all games. LOLP metrics of interest – the number, magnitude and duration of events, event timing type (HLH, LLH, or both), game percent, time percent, expected unserved energy, and the 95 th & 99 th percentiles of hourly unserved demand are obtained through post-processing this record. Dispatch records: Show the hourly and monthly loads and resources details for a single user- selected game. The hourly dispatch record includes demand, contracts, source production, overgen, transmission, reserves, unserved demand, and slice accounting. The monthly dispatch record, in addition to summarizing most of the hourly segment data, also shows source availability and thermal source displacement. Hydro-regulator output : Similar to BPA Hydsim standard output, for up to 70 games. Includes plant-by-plant summaries of flows, spill, generation, draft, ending storage and reservoir elevations, and overall summaries of generation for each period. Additional files based on the regulator output are useful for model validation and finding duration curves for reservoir refill. Miscellaneous output files : Assist with debugging and validation, log game parameters and input filenames, and are used with the Visual Basic shell (not currently active). Some of these output files include contract detail, hydro capacity estimation and hourly loads for a single game, water/temp/wind year selections and overgen detail for all games, and monthly hydro block sizing and flex accounting for up to 10 games. BPA’s Loss of Load Probability (LOLP) Project
What is the Industry Standard? Metric and threshold (or a derivative of) 1 day (or event) in 10 years Used across the United States and in Canada The only exception found was Hawaii (1 event every 4.5 years) Modeling done for capacity limited systems – explicitly don’t model fuel supply and load deviation (and/or have limited load deviation) Typically driver of energy-not-served (ENS) events: Forced Outages First Energy quipped, “The criteria (metric) creates a good target, but a lousy planning/compliance measure” BPA’s Loss of Load Probability (LOLP) Project5
Conversion of the Industry Standard Metric into a Probability Distribution Industry Standard Metric: “One day in 10 years” taken to mean on average one ENS event every 10 years In running many Monte-Carlo games, there is a distribution in the number of ENS events; some games have more than 1 while others have fewer Assume Industry Standard Metric has a distribution of ENS events shaped like the Poisson Distribution, which requires – Events occur randomly (Forced Outage being typically the #1 driver) – Different events are independent of each other Converting hourly Genesys LOLP results into compatible data for comparison with Industry Standard Metric: – Count ENS events only during peak hours – ENS lasting through multiple peak periods counted as multiple events – On average “1 ENS event every 10 years” becomes on average “0.1 ENS event every year” BPA’s Loss of Load Probability (LOLP) Project6
7 Industry Standard Metric Poisson Distribution, with an average of 0.1 ENS Event per year, as Industry Standard Metric BPA’s Loss of Load Probability (LOLP) Project
BPA Comparison to the Industry Standard Distribution Study Assumptions 1,000 games ran for FY 2013 Current FCRPS, Regional Dialogue Contracts, Resource Program market assumptions, and approximately 600 aMW of additional HLH resources available (with a seasonal shape) Added resources until the number of games with events < 1% Load following customers; historical conservation embedded in load forecast No additional conservation assumed Load following customers Tier 2 placed on BPA Results BPA LOLP ENS events do not fit the Industry Standard Metric Exceed the industry standard metric for games with zero event (99% versus 90%) Yet, there are also games with multiple events outside the Industry Standard Metric These events are primarily the result of bad water years (energy limited system); and secondarily the result of high load forecasts The single event lasting 14 days has an average ENS (over the event period) greater than 1 aGW BPA’s Loss of Load Probability (LOLP) Project8
9 BPA Compared to the Industry Standard Distribution Industry Standard Metric (90% of games with no event) BPA LOLP (99% of games with no event) BPA LOLP (1% of games have more than 3 events) BPA’s Loss of Load Probability (LOLP) Project
Initial BPA Metrics LOLP (time%) LOLP would measure the percent of total hours simulated that had ENS occurrences A threshold of reliability could be established Tail Event Analysis Designed to capture the frequency, duration, and magnitude of highly infrequent but significant events Building resources of sufficient size to cover these events that would operate so infrequently may not be the desired outcome What additional resources are available (but are not explicitly modeled in Genesys)? BPA’s Loss of Load Probability (LOLP) Project10 PSRI contract - major task items are to review these initial metrics and to give BPA a view to countries that have developed metrics for energy limited systems
The Oregon Experience PGE August 2009 IRP Metrics: LOLP = hours with ENS/Total Hours Expected Unserved Energy (EUE) = Unserved Energy/Number of hours with Unserved Energy Tail Var90 of Unserved Load No Thresholds for LOLP metrics Rather metrics are given a weight in portfolio selection (15 total portfolios) LOLP no weight given EUE given a 15% weight TailVar90 given a 10% weight PAC May 2008 IRP Metrics: Average annual ENS July probability of > 25,000 GWhr ENS Average annual ENS July probability of > 25,000 GWhr ENS No thresholds for LOLP metrics Rather metrics are given a weight in portfolio selection (21 total portfolios) Average annual ENS given a 5% weight July Probability of >25,000 GWhr ENS given a 5% weight BPA’s Loss of Load Probability (LOLP) Project11 In January 2007, OPUC adopted IRP guidelines – including Guideline 11 Reliability “Electric utilities should analyze reliability within the risk modeling of the actual portfolios being considered. Loss of load probability, expected planning reserve margin, and expected and worst-case unserved energy (ENS) should be determined by year for top- performing portfolios.”