60 likes | 64 Vues
ERCOT Staff Analysis of Weather Sensitivity Code Changes. February 8, 2005. Analysis Background. Analysis based on currently active IDRs Includes both NOIE and IOU ESIIDs Includes prospective changes based on ERCOT’s 2004 assignment review Analysis period is market open to present.
E N D
ERCOT Staff Analysis of Weather Sensitivity Code Changes February 8, 2005
Analysis Background • Analysis based on currently active IDRs • Includes both NOIE and IOU ESIIDs • Includes prospective changes based on ERCOT’s 2004 assignment review • Analysis period is market open to present
Current Weather Sensitivity Code Assignmentsand Change History • About 28% of current ESIIDs classified as weather sensitive, 23% for IOUs and 63% for NOIEs • About 26% of current ESIIDs have had one or more code changes since market open
Weather Sensitivity Code Changes • 80% of IOU, 70% of NOIE ESIIDs with no assignment changes • About 14% of current ESIIDs are changing as a result of 2004 review
Year-by-Year Weather Sensitivity Code Changes • 2002 WS Review was not Performed by ERCOT • 2003 had a significant migration (88% of changes) to WS • 2004 had a significant migration (69% of changes) to NWS • 57% of ESIIDs with 2004 changes were reverting to pre-2003 code
Potential Action Items • Continue using current methodology … current data loading rates lower the impact of applying proxy days • Assign all IDRs to WS … WS proxy day routine okay for most ESIIDs and falls back to NWS proxy day if no suitable days are found • Assign all IDRs to NWS … reduced processing for proxy day routine, may not introduce that much inaccuracy • However … if Initial Settlement is moved from 17 to 10 days, the proxy day routine will be invoked more often … a less accurate routine will increase profiling error • Conduct analysis to improve WS assignment methodology • Look at R-square distributions • Use multiple-year window for the analysis • Incorporate additional variable(s) e.g. summer ratio, kW/kWh minimums • Outlier analysis e.g. some ESIIDs have large year-to-year changes in R-square • Apply dead-bands