Capacitive Coupler Dimensions Optimization Using Deep Learning-Based Genetic Algorithm

A capacitive coupler optimization method using a deep learning-based genetic algorithm (DLGA) is introduced in capacitive power transfer. After training the neural network using simulation data, a genetic algorithm is employed for multi-objective optimization of the coupler, providing a three-dimensional Pareto front as a series of optimal solutions. Compared with the optimization method only using finite element analysis, DLGA reduces the computation time with an average error of only $0.27 \%$. Better results could be obtained based on the proposed optimization method, not limited to the discrete points in the dataset.