A NOVEL REINFORCEMENT LEARNING - BASED CONTROL APPROACH FOR GRID-INTEGRATED PV-FESS SYSTEMS USING INDUCTION MOTORS

Tóm tắt

This study introduces a novel reinforcement learning-based control strategy for a grid-connected photovoltaic system integrated with a flywheel energy storage system. The proposed method replaces the conventional dual-loop current control of the Field-Oriented Control scheme for the induction motor  within the FESS with a single-agent controller based on the Deep Deterministic Policy Gradient algorithm. This intelligent controller leverages the strengths of Reinforcement Learning to handle the nonlinearities and parameter uncertainties inherent in the Flywheel Energy Storage System. Simulations in MATLAB/Simulink evaluate the performance of the proposed control system under various operating conditions. Results demonstrate that the Deep Deterministic Policy Gradient -based controller outperforms traditional Proportional-Integral  controllers in ensuring stable power output to the grid, even under significant fluctuations in photovoltaic generation. The proposed control method enhances system stability, optimizes energy storage dynamics, and maintains power quality, contributing to the broader adoption of intelligent energy storage solutions in renewable energy integration.