Machine Learning-Assisted Hierarchical Electromagnetic Co-Design of Double-D Wireless Charging Pads for Electric Vehicles

This paper presents a machine learning-assisted hierarchical optimization framework for the electromagnetic co-design of Double-D wireless charging pads for stationary electric vehicle (EV) applications. The proposed methodology integrates Random Forest-based surrogate modeling with Particle Swarm Optimization (PSO) and a Genetic Algorithm (GA) to enable computationally efficient and material-aware exploration of the electromagnetic design space. In the first stage, the coil geometry is optimized using continuous design variables, including turn count and conductor spacing, to achieve the required inductance with reduced copper utilization. In the second stage, a binary GA is employed to optimize the segmented ferrite arrangement for improved magnetic flux guidance and reduced ferrite dependency. The hierarchical optimization strategy follows the physical interaction between magnetic field generation and flux shaping, enabling stable convergence while minimizing repeated full-wave simulations. The optimized charging pad achieved the target inductance requirement using a compact 35-turn coil structure and reduced ferrite utilization by nearly 80% compared with a fully ferrite-backed configuration. FEA analysis confirmed stable electromagnetic performance under misalignment conditions up to ±80 mm longitudinally and ±75 mm laterally. Experimental validation further demonstrated a peak end-to-end DC–DC efficiency of 93% under rated aligned operation together with stable high-efficiency performance under moderate misalignment conditions. Overall, the proposed framework provides a computationally efficient and material-aware electromagnetic co-design methodology for practical stationary EV wireless charging systems.