A Tuning-free Fault Detection Method for Distribution Network Based on PMU Data and Random Matrix Theory

Abstract

The increasing penetration of renewable energy sources has significantly altered the fault characteristics of distribution networks with distributed generators (DGs). The fault current is limited as the emerging solar plants normally connect to the grid via inverters. And the operating patterns of the system have become more diverse and complex. The traditional pre-tuned protection and fault detection methods are facing substantial challenges. This paper proposes a centralized fault detection scheme based on the random matrix theory (RMT) that utilizes synchronous phasor data from the distribution-level phasor measurement unit (PMU). Firstly, data normalization and resampling techniques construct the operating state matrix (OSM). Then the mean spectral radius (MSR) is calculated and used for fault detection without the complex tuning according to the operation state of the system. Finally, the faulty section is determined by comparing the features from different sub-group of the PMUs according to the topology partition. A multi-feeder distribution network is modeled in PSCAD and used to verify the proposed method.