Similarity measures in Small World Stratification for distribution fault diagnosis

Abstract

Small World Stratification (SWS) is a sampling strategy aims to solve the problem of insufficient historical data for fault diagnosis in a small local region. SWS involves sampling relevant fault events by Geographic Aggregation (GA) and Feature Space Clustering (FSC), and identifying the group of fault events that should be investigated together. In order to apply FSC, proper measures of similarity among regions are needed. In this paper, we propose four types of regional feature vectors (RFV): normalized regional feature vectors (NRFV), relative regional feature vectors (RRFV), likelihood regional feature vectors (LRFV) and generalized regional feature vectors (GRFV), derived from the measures used to analyze distribution faults. Similarity measures based on the distance between RFVs are evaluated using fault events simulated by the Distribution Fault Simulator. Experimental results suggest that GRFV is the best among the four.