Research on Data-Driven Fault Scenario Simulation Model of Distributed Network

Abstract

Distribution lines are critical components of today's power system, and can directly impact on power supply security and stability. An effective power system protection system should detect any faults as soon as they occur. There are two steps in fault diagnosis. One is defect classification, which has already achieved excellent accuracy rates. As a result, this study concentrates on the opposite objective, which is fault location. This research introduces a War Strategy Water Wave Optimization (WSWWO) based Deep Q Network for radial distribution networks using synchronous generators for Distributed Generation (DG). The algorithm estimates the fault site by analyzing voltage and current samples at the main feeder head, and scheduled active and reactive power injections by network synchronous generators. A full-order synchronous machine model was utilized to analyze the dynamic behavior of DG plants during fault transients. Here, the Deep Q Network is trained by WSWWO, and hybridized by two optimization algorithms, War Strategy Optimization (WSO) and Water Wave Optimization (WWO). Furthermore, the experimental results revealed that the WSWWO-Deep Q Network outperformed state-of-the-art models in terms of Accuracy and Error, with values of 0.997 and 0.141, respectively.