Fuzzy Partitioning of a Real Power System for Dynamic Vulnerability Assessment

Abstract

Recently, the authors proposed a clustering approach based on the fuzzy C-medoid algorithm (FCMdd), for segregating large power systems into coherent electric areas centered around a representative so-called medoid-bus. This bus was shown to be a natural location for PMU in the context of wide-area measurement system (WAMS) configuration for of dynamic vulnerability assessment (DVA). The method was demonstrated on two test systems. The goal of this companion paper is to extend the approach to an actual grid (Hydro-Quebec) with more realistic characteristics in terms of geography and system dynamics. We start by developing a formulation of the coherency matrix that is recursive in time to enable online grid partitioning. The FCMdd is then implemented and compared with other statistical learning techniques. It is observed that only FCMdd is able to provide an intuitively appealing 7-clusters solution for 429-bus system. It is further demonstrated that medoids-based system-wise indices can forecast the contingencies severity under varying network configurations and loadings.