Rapid Design Optimization of Planar Magnetic Coupler for Undersea IPT Utilizing Electromagnetics-Embedded Neural Networks

Abstract

Magnetic couplers are a crucial component of undersea inductive power transfer (IPT) systems. Existing research on undersea IPT has focused on magnetic structure design. The impact of design parameters on the performance of undersea magnetic coupler mechanisms is mainly evaluated utilizing finite element method (FEM), which is time-consuming for design optimization. A systematic design optimization methodology for magnetic couplers to achieve high power density and efficiency is lacking. In this paper, a model embedding electromagnetics and backpropagation neural network (BPNN) is developed to calculate two essential electromagnetic parameters: mutual inductance and power loss of magnetic couplers based on design parameters. Compared to analytical methods, the proposed model demonstrates superior accuracy and can model more complicated eddy current problems. The proposed model also features a simpler network structure and requires a smaller dataset (less than 50% ) than pure-data-driven approaches. Moreover, this methodology exhibits greater design flexibility over FEM with a significant reduction in optimization time by at least six orders of magnitude. The objective functions and constraints are established for multi-objective optimization. FEM and a 1.5 kW prototype verify the proposed method. The optimization objectives calculated using the proposed model are highly consistent with FEM results and experimental results, with an error of less than 5.1%.