Anomaly Identification for Active Distribution Networks Using Random Matrix Theory

Power distribution networks have been undergoing a transition towards active distribution networks (ADNs) that feature more complicated operating conditions. Accordingly, accurate and efficient anomaly identification can significantly improve the operational reliability of ADNs. This paper employs random matrix theory (RMT) to develop an edge computing method that only relies on local measurements from smart meters to perform data-driven anomaly identification. First, a general data acquisition and processing framework is proposed for ADNs, where edge computing plays the fundamental role of online data analysis. Then, the autoregressive moving average (ARMA) model and empirical spectrum density (ESD) calculation are integrated to check the existence of anomalies by considering time-lag correlation in the set of measurements. Besides, two indices are proposed to evaluate and examine the benefits of edge computing-based anomaly identification. The accuracy and efficiency of the proposed method are finally validated by conducting case studies on the IEEE European Low Voltage Test Feeder.