Machine-Learning Based Optimal Design of a Wireless Power Transfer Coil for Battery-Powered Tram

This paper proposes an optimal design method for a wireless power transfer (WPT) coil using a machine learning regression model. The proposed method was applied to design a WPT coil for a tram wireless power charging system. Designing a wireless power transmission coil with ferrite is challenging since it is difficult to calculate self-inductance, coupling coefficient, and winding resistance with an analytical model. This paper proposes a method for optimal design by inferring coil parameters using a machine learning regression model based on finite element analysis (FEA) simulation data. The proposed design methodology achieves optimal efficiency for coupling coefficients and shapes under given design conditions by using parameter estimation with a regression model and NSGA-II. This methodology searches not only for optimal shapes but also for optimal coupling coefficients, resulting in high coil-coil efficiency. The proposed WPT coil design methodology was used to design a coil for a 200kW 60kHz WPT system for a battery-powered tram, achieving a coil-coil efficiency of up to 97%.