Design and Evaluation of AI-Tuned Controllers for Three-Phase EV Charging Stations in Renewable Energy Grids

This paper presents a comprehensive study on artificial intelligence (AI)-based tuning of conventional PI, PID, and fractional-order PI/PID (FOPI/FOPID) controllers applied to a three-phase inverter supplying an electric vehicle (EV) charging station. The motivation arises from the increasing penetration of EVs and their dynamic charging profiles, which impose significant challenges on inverter control, including voltage regulation and harmonic suppression. While classical tuning methods such as Bode-based approaches are well studied, the application of AI methods such as Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Reinforcement Learning (RL) for fractional-order controllers in EV charging scenarios remains underexplored. The study uses a MATLAB/Simulink model representing a three-phase EV load (modelled after a 22 kW Renault Zoe with balanced 3-phase demand), and tests the controllers under three scenarios: step load (clustered EV arrivals), ramp load (gradual charging), and stochastic load (Poisson-distributed EV arrivals with normally distributed durations). Performance is evaluated in terms of voltage response, overshoot, settling time, total harmonic distortion (THD), and error metrics (ISE, R²). Results demonstrate that AI-tuned FOPID controllers achieve superior dynamic performance and harmonic suppression compared to AI-tuned PI and PID, with ANFIS providing the best accuracy and RL offering faster convergence. The findings highlight the potential of AI-enhanced control for next-generation EV charging stations integrated into smart grids.