Clustering-Based Dynamic Event Location Using Wide-Area Phasor Measurements

Abstract

This paper presents an online dynamic event location method that uses wide-area measurements. The method is based on an offline hierarchical clustering of N generators into coherent groups. The clustering method can be used to produce any number of groups between 1 and N. One measurement location is selected during the offline phase to represent each cluster in the online phase. Event location is formulated as finding the most likely group from which an event originates. The rotor frequencies of the representative generators are used to identify the cluster with the largest initial swing. The method is demonstrated for locating events caused by faults and generator trips on a 176-bus model. The online location accuracy is around 80% when the offline clustering is good.