Pattern Mining and Anomaly Detection based on Power System Synchrophasor Measurements

Abstract

Real-time monitoring of power system dynamics using phasor measurement units (PMUs) data improves situational awareness and system reliability, and helps prevent electric grid blackouts due to early anomaly detection. The study presented in this paper is based on real PMU measurements of the U.S. Western Interconnection system. Given the nonlinear and non-stationary PMU data, we developed a robust anomaly detection framework that uses wavelet-based multi-resolution analysis with moving-window-based outlier detection and anomaly scoring to identify potential PMU events. Candidate events were evaluated via spatiotemporal correlation analysis and classified for a better understanding of event types, resulting in successful anomaly detection and classification of the recorded events.