Double “D” and “O” Coupler Optimization Using Reinforcement Learning for Wireless Power Transfer Electric Vehicle Charging System

A wireless resonant inductive power transfer (WRIPT)-based electric vehicle (EV) charging system needs an effective geometry coil with high misalignment tolerances and a lightweight coupler. This article proposes a double “D” and O (DDO) geometry coil and an optimal reinforcement learning (RL) algorithm for the DDO coupler optimization. The DDO geometry inculcates D, inverted D, and partially overlapped O-shaped coils. It combines the features of “DD” and “Circular” coil geometries and introduces a large interoperable coupling flux pipe region over the coupler surface. This large interoperable flux pipe region enhances the surface magnetic field density (B), addresses the coupling misalignment, improves the effectiveness of coupling (K), and also eliminates the power null phenomenon (PNP) effect. Due to the large number of ferrite cores in the coupler, optimizing with theoretical analysis is complex and requires significant computational time. For this reason, an improved RL algorithm is proposed to optimize the nonlinear behavior of the cores in a coupler. This coupler optimization applied to the significant parameters are; core positioning, number of cores, core layers, core loss, and core thickness. By optimizing the cores, archives are lightweight couplers, the cost is reduced, core saturations are eliminated, core loss is reduced, and the coupling is enhanced with high coupling coefficients under misalignment conditions. An experimental setup is developed for a 3.3-kW power capacity with a proposed coupler, achieving 93% power transfer efficiency (PTE). Hence, this machine learning (ML)-based optimization finds its benefits by reducing the core, and the power losses to enhance the PTE performance for a WRIPT EV charging system.