Mutual Inductance Estimation of SS-IPT Systems with Physics-informed Neural Network

Inductive power transfer (IPT) systems heavily rely on accurate values of mutual inductance for efficient and optimal performance. This paper proposes a parameter estimation method for series-series IPT (SS-IPT) systems with a physics-informed neural network (PINN). Firstly, the physics model of SS-IPT systems is established by deducing its ordinary differential equations. Then, the physics model is merged into a neural network to establish PINN. With the PINN, the mutual inductance of the IPT system is identified by using the input current and input voltage of the transmitter-side only. The proposed PINN leverages the power of data-driven technology while incorporating known physics-based principles, leading to more accurate and generalizable identification results. The proposed method is verified in several groups of simulation data. The results show that the estimated error of the proposed method is less than 2% under different coupling coefficients.