Physics Guided Data-Driven Characterization of Anomalies in Power Electronic Systems

Abstract

The transition of conventional power system onto power electronics dominated grid (PEDG) has lead to amplified complexity in system-level control schemes to maintain reliability and operational stability. Considering the abundance of data in PEDG, machine learning (ML) schemes have emerged as a promising alternative. In this article, a physical guided data-driven approach using pattern recognition neural network (PRNN) is employed with semi-supervised learning. To distinguish between the faults and cyber-attacks without relying historical data scenarios. Finally, the results of proposed approach are discussed by utilizing ML tools.