A New Design Optimization Method for Dynamic Inductive Power Transfer Systems utilizing a Neural Network

Abstract

Ordinary optimization methods for inductive power transfer (IPT) coils rely heavily on FEM simulations due to the large air gap between the primary and secondary coils. This problem is further compounded for dynamic inductive power transfer (DIPT) systems since DIPT requires more calculation points than IPT to know the average output power across the primary coil. This paper proposes a new design optimization method for DIPT systems utilizing a combination of FEM simulations and a neural network. A neural network trained by FEM results can generate a significantly larger number of design points than conventional methods in the same simulation time. To demonstrate the effectiveness of the proposed method, designs using the proposed and conventional methods are compared. The results achieve 20% smaller core volume and 26% higher average output power for the proposed method over conventional one with the same simulation time and design criteria. Furthermore, to verify the accuracy of the proposed method, the final design generated by the proposed method is compared with FEM results.