Hybrid Adaptive Control of Bidirectional Wireless EV Charging Using Deep Learning and Model Predictive Control

Abstract

The wireless electric vehicle (EV) charging system needs the control in order to ensure the efficiency of power transfer, the bidirectional process, and the nonlinear characteristics of the system behavior under the dynamic conditions. The paper is a proposal of a Hybrid Adaptive Control Framework with some potential to integrate a Deep Neural Network (DNN)-based predictive model with a Model Predictive Control (MPC) concerning bidirectional wireless EV charging. The DNN is trained using large data sets of simulation and hardware in the loop of coil misalignment, coupling variation, load transitions and switching dynamics. It forecasts the reactions of the system- e.g. the secondary-side voltage, current and power of the system under different operating conditions. MPC uses these predictions to generate optimal control measures and impose voltage, current and bidirectional G2V / V2G power flow constraints. The hybrid technique adds dynamism stability, minimizes temporary overshoot, and increases voltage regulation in contrast to the traditional PID and standalone ANN controllers. The outcome of the simulation outcomes in the presence of random misalignment and disturbances in the load data also indicate better tracking and convergence rates. The suggested hybrid DNN-MPC, therefore, offers a powerful, intelligent, and constraint-sensitive control framework of the next-generation wireless EV charging systems.