EV Wireless Power Transfer Core Design for High Magnetic Coupling designed by Reinforcement Learning

A novel approach to designing the optimal core structure of an electric vehicle (EV) wireless charger using a reinforcement learning (RL) algorithm based on Deep Q-Learning Network (DQN) is proposed in this paper. The conventional inductive power transfer (IPT) cannot be theoretically designed due to the nonlinearity of magnetic fields, which makes it challenging to optimize the core structure of the EV wireless charger. The proposed RL algorithm uses an agent based on the ε-greedy algorithm for the core selection process, which is evaluated in this paper. The agent learns from the reward (mutual inductance) and the state of the environment (finite-element-method simulation) to obtain the optimal core design for high mutual inductance. The proposed RL algorithm provides the optimal core structure based on WPT3/Z3 standard coil from J2954, which is superior to the conventional RL algorithm in terms of computation time and higher mutual inductance. The use of RL algorithms in designing optimal core structures for EV wireless chargers is an innovative approach that can significantly improve the efficiency of wireless charging systems. The proposed RL algorithm has the potential to reduce the design time and cost and improve the overall performance of wireless charging systems.