Multi-Objective Optimization of Magnetic Parameters in Wireless Charging Coupler Based on Deep Belief Network
Yu Lan,Jinhai Jiang,Kai Song,Fengshuo Yang,Lin Sang,Ke Wang,Zhenjun Zhang,Xuan Zhang,Xuling Li
DOI: https://doi.org/10.1109/icems59686.2023.10344432
2023-01-01
Abstract:The magnetic coupler is the core component of the wireless charging system. Its magnetic parameters have a significant effect on transmission performance. However, the current analytical models used for multi-objective optimization design are inadequate with poor calculation accuracy, leading to challenges in achieving comprehensive multi-objective optimal. This paper proposes an efficient fitness function modeling method for multi-objective optimization based on the deep belief network (DBN), enhancing both the design accuracy and optimization speed. Firstly, a three-dimensional magnetic circuit model is developed to consider the coil, magnetic core, and aluminum plate. Secondly, a loss correlation model of the magnetic coupler is established. Thirdly, the deep belief network (DBN) regression is utilized to obtain an accurate fitness characterization function. Finally, the fractional particle swarm optimization (FPSO) is applied to achieve the Pareto optimal solution of the magnetic coupler’s design parameters. In this work, the WPTI rectangular coil is taken as an illustrative example. An optimized design with a coupling coefficient of 0.23, transmission efficiency of 97.5%, and a weight of 9.6 kg is selected for experimental verification. The results demonstrate that the predicted values match the measurements with an error less than 5%.