The catastrophic day identification problem in distribution reliability, and the robust estimation approach

Abstract

The Major Event Day identification process used in distribution reliability calculation described in IEEE Standard P1366 is based on computing a threshold from five years of data. Massive, infrequent reliability events, such as a one in one hundred year ice storm, dubbed Catastrophic Days, can skew the threshold computation and thus affect the identification of Major Event Days and the calculation of normal reliability levels until the Catastrophic Day rolls out of the five year window. The Catastrophic Day Task Force of the Distribution Reliability Working Group has explored methods to identify Catastrophic Days and exclude them from the Major Event Day computations. One such method, Robust Estimation, is described in this work. While theoretically valid, Robust Estimation has not worked well in practice.