Component Outage Data Analysis Methods. Volume 1. Overview and Special Applications. Final Report

摘要

This project addresses methods for improving the quality of outage statistics for power equipment. Statistical techniques were evaluated and applied to utility outage data to illustrate their use. Two major difficulties confront the data analyst: imprecision and inaccurate models. Imprecision stems from the limited data samples typical of generation and transmission work. To improve precision, one must first measure it; routine use of interval estimates is advocated and explained in detail. Fitting distributions is discussed for interval estimation and verifying models of equipment failure and repair. Precision can be gained by pooling data across similar experience/equipment categories; methods are demonstrated for testing data samples to be pooled. Similarly, use of regression analysis to uncover relationships and integrate information from dissimilar samples is illustrated. Errors in modeling equipment performance, e.g., neglecting dependences, are discussed and model verification by simulation is shown to be a powerful tool. Emphasis is given to infrequent events and small data samples; Bayesian analysis and model simplification (reducing the number of parameters estimated) are examined as remedies to data sparsity. Limitations to the development of improved power systems outage statistics were identified by reviewing the physical processes, existing data collection activities and the end-uses of outage statistics. A list of suggested industry research activities is provided for further evaluation. (ERA citation 06:031949)