Artificial Neural Network Modeling of WPT Magnetic Fields in an EV Application

Significant recent work on wireless power transfer (WPT) for electric vehicles (EVs) has been conducted. With the rapid development of WPT solutions, and with many classes of EVs potentially leveraging the same infrastructure, there is a need to rapidly qualify and optimize WPT coil design. To that end, an artificial neural network (ANN) is trained to estimate 4,203 magnetic parameters of a wireless power transfer (WPT) system. The network is trained on 48,284 unique coil geometries. Appropriate cost functions to train the ANN are introduced. The trained ANN is shown to effectively reproduce data generated by a finite element method (FEM)