New Structure Design of Ferrite Cores for Wireless EV Charging by Machine Learning

Abstract

In this article, a machine learning algorithm is applied for the first time in inductive power transfer (IPT) to find a ferrite core structure with high magnetic coupling between transmitting (Tx) and receiving (Rx) coils for an electric vehicle (EV) wireless charging system. Since formula-based theoretical design is not available due to the nonlinear magnetic field distortion stemming from the presence of the ferrite core in an IPT system, the proposed core structure design has been achieved through finite-element analysis simulation-based data learning. The proposed design methods are so general that they can be applied to any conventional IPT coil design. Furthermore, it can optimize the core structures for high coupling coefficient, mutual inductance, desired magnetic flux density in the specific area, etc. By training only 0.011% data out of the total possible cases, it is verified by simulation and experiment that the ferrite core structure obtained by the proposed method has a mutual inductance that is 0.6% higher than that of the conventional design level in the case of 15 cm distance between the Tx and Rx coils, even though the volume of the ferrite cores are reduced to 90%. Also, a prototype 3.0 kW stationary EV wireless charging system was implemented and showed fairly better performance than a conventional case.