Research on Transmission Characteristics of Magnetic Couplers for Underwater Wireless Power Transfer Based on Prior Knowledge Input Neural Network

Underwater wireless power transfer (UWPT) operates under special conditions, where the conductivity of seawater introduces eddy current losses, thereby reducing system efficiency. Meanwhile, the design parameters of magnetic couplers significantly influence their transmission characteristics. This paper proposes a fast and accurate neural network prediction model for mutual inductance and losses of magnetic couplers based on mirror-method prior knowledge within a prior knowledge input (PKI) framework. The proposed model integrates a low-fidelity analytical model with data-driven learning to achieve high prediction accuracy while maintaining computational efficiency. Based on the developed model, the transmission characteristics of unipolar rectangular and bipolar DD magnetic couplers are systematically investigated. The results indicate that the rectangular couplers exhibit higher overall efficiency than the DD couplers, with a more monotonic variation in efficiency under design constraints. Owing to its structural characteristics, the DD couplers present an optimal current-carrying area ratio, which is approximately 0.85 within the parameter range. Experimental validation is conducted at a 1 kW power with outer dimensions of 200 mm × 250 mm. The optimal transfer efficiencies of the rectangular and DD couplers reach 97.33% and 96.19%, respectively. The experimental results show good agreement with both simulations and model predictions, demonstrating the reliability of the proposed method for UWPT magnetic coupler analysis.