Outage data collection and parameter estimation for an improved probabilistic contigency analysis

Probabilistic risk assessment (PRA) techniques are increasingly being used in electric power industry applications for better coping with uncertainties over deterministic approaches. One application where PRA techniques can add value is data analysis for parameters such as outage frequency. Focusing on a probabilistic contingency analysis (PCA), this study examines the issue of obtaining a point estimate of outage statistics by lumping or pooling outage data records together from different sources. A Pearson Chi-square test is adopted to determine the poolability of data, and a lognormal distribution is used to model the data source variability and capture variations of operation and maintenance practices among different utilities. The distribution parameters representing outage frequencies and durations are calculated from the raw outage data. An improved PCA scheme based on the outcomes of this study is proposed and being implemented.