AI-Enabled Wireless EV Charging with Battery Health Estimation and Energy Management

Abstract

This study presents an AI-enabled symmetric wireless electric vehicle (EV) charging framework that integrates artificial neural networks (ANN) for real-time battery state estimation and reinforcement learning (RL) for adaptive power control. The proposed system addresses efficiency losses and battery degradation typically observed in conventional wireless power transfer (WPT) systems under coil misalignment. The ANN, trained on the publicly available EVBattery dataset (1.2 million records from 464 EVs), predicts the state of charge (SOC) and state of health (SOH) with RMSE values of 1.8 ± 0.2% and 2.5 ± 0.3%, respectively. A Proximal Policy Optimization (PPO)-based RL controller dynamically tunes compensation and frequency parameters to sustain optimal coupling. Laboratory and simulation tests, repeated five times per condition, confirm an average charging efficiency of 85 ± 2% across 5–20 cm air gaps, battery degradation reduction from 1.2% to 0.8 ± 0.1%, and renewable-energy utilization of 72 ± 3%. Statistical analysis (p < 0.05) validates significant performance gains over baseline and literature benchmarks. The integrated ANN–RL framework thus provides a reproducible, health-aware, and grid-coordinated solution for sustainable next-generation wireless EV charging.