Analysis of catastrophic events using statistical outlier methods

摘要

Use of IEEE Standard 1366–2003 for determining major events has provided substantial consistency when comparing year-on-year performance, however certain events (such as hurricanes or ice storms) that destroy substantial portions of a company's power system may result in extremely large daily SAIDI results which could skew the calculation of the major event threshold over the subsequent five years (according to the standard). This change in threshold could result in certain days in subsequent years not being considered as Major Events. The Catastrophic Event Task Force, convened under the Distribution Reliability Working Group, undertook substantial investigations to explore the impact of these events and the ability for an outlier method to be developed which transferred well across a variety of companies. This paper will explore an outlier identification method used by statisticians, notably the use of box and whiskers plots, to determine its application in segregating outlier data. It will also summarize the approaches considered and the final conclusions arrived at by the Catastrophic Event Task Force.