GNN-RL APF Switching Controller for Power Quality Enhancement in Renewable Integrated Grids

Abstract

The integration of renewable or green energy sources in the grids introduce several problems like power quality issues such as harmonic distortion, voltage fluctuations and unbalanced reactive power flow etc. Active Power Filters (APF) can effectively work on these issues. However, the key performance of APFs depend on the optimal switching control strategy. In this paper, a novel Graph Neural Network Reinforcement Learning (GNN-RL), a machine learning based optimal switching controller is applied for APFs in the renewable integrated grid. The GNN’s spatial relationship modelling is used to integrate the complex network configurations of power distribution network and adaptive control is achieved under dynamic operating conditions using RL optimization. The proposed work is carried out in a modified IEEE 13 bus distribution feeder with the integration of PV and Wind sources. The results show that the notable reduction in Total Harmonic Distortion (THD) as compared to the traditional PI controlled based switching in different loading conditions