Deep reinforced learning-based inductively coupled DSTATCOM under load uncertainties

Abstract

Concerning the power quality issues in the power distribution network due to load uncertainties and improper impedance matching of the inductances, deep reinforced learning (DRL)-based inductively coupled DSTATCOM (IC-DSTATCOM) is proposed. First, by analyzing the impedance matching principle, the expression of source, load and filter current is derived with the help of inductive filtering transformer. And second, an individual DRL subnet structure is accumulated for each phase using mathematical equations to perform the better dynamic response. A 10-kVA, 230-V, 50-Hz prototype direct coupled distributed static compensator (DC-DSTATCOM) and IC-DSTATCOM experimental setup is buit to verify the experimental performance under uncertainties of loading. The IC-DSTATCOM is augmented better dynamic performance in terms of harmonics curtailment, improvement in power factor, load balancing, potential regulation, etc. The benchmark IEEE-519-2017, IEC-61727 and IEC-61000-1 grid code are used to evaluate the effectiveness of the simulation and experimental study.