Delta and Inverse Delta Coupler Optimization Using Machine Learning for Wireless Power Transfer Electric Vehicle Charging Application

Abstract

A wireless resonant inductive power transfer (WRIPT)-based electric vehicle charging system requires an efficient lightweight inductive coupler with high misalignment tolerance. This article proposes a new delta and inverse delta (Δ–∇) coil geometry coupler and a machine learning (ML)-based reinforcement algorithm for Δ–∇ coupler optimization. Δ–∇ is a combination of Δ and ∇ geometry coils, which introduce a diagonal flux pipe compared with the conventional geometry coil. This diagonal flux pipe region enhances the surface magnetic field ( B ) over the coil surface and improves the coupling coefficient to address misalignment. Also, this new geometry eliminates the power null phenomenon effect and limits power fluctuations in the WRIPT coupling architecture. The ferrite core in the coupler has a nonlinear magnetic field ( B ) on the surface, which is not easy to express and optimize using a conventional formula-based approach. Hence, the proposed ML-based ferrite core optimization of Δ–∇ pad finds its benefit in improving power transfer efficiency (PTE) by reducing power losses. This optimization method is applied to significant parameters (ferrite core position, number of cores, core layers, and core thickness) of the Δ–∇ geometry coil, by training 2.5% datasets out of the total possible cases. Also, the developed system was experimentally verified successfully and ensures that a Δ–∇ coupler achieves a higher PTE than the conventional geometry during various coupling conditions.