Reinforcement-Learning-Based Intelligent Control of Multi-Source Power Converters for Dynamic Renewable-Energy-Assisted Electric Vehicle Charging
Abstract
In this work, an intelligent control strategy based on reinforcement learning (RL) is designed for renewable-energyassisted electric vehicle (EV) charging power converters when the source and load vary during operation. The charging architecture combines photovoltaic (PV) generation, wind energy, battery storage, alternating current (AC) grid support, multi-port power converter and PWM-based control feedback. The RL controller takes inputs of direct current (DC)-link voltage error, renewable power availability, battery state of charge (SoC), EV charging current, grid-current components and harmonic distortion index to generate adaptive duty-ratio and modulation control actions. System performance is evaluated using DC-link voltage regulation, EV chargingcurrent tracking, renewable power sharing, RL training convergence, SoC variation, and converter duty-ratio adaptation. The results confirm that the DC-link voltage can be effectively regulated near the reference value with less transient deviation, and the EV charging current follows the reference profile with better dynamic performance. The renewable power-sharing response demonstrates that PV, wind, battery, and grid inputs are coordinated. In the RL model, the response shows greater reward improvement and a decrease in control error as the number of episodes increases. The proposed framework enables smart energy infrastructure with renewable integration to efficiently and adaptively support EV charging with power quality assurance.