Machine learning-based parameter identification method for wireless power transfer systems

Parameter identification is an effective way to obtain uncertain parameters of wireless power transfer (WPT) systems, which is essential to achieving robust control and efficiency improvement. The traditional method relies on the phase lock of the primary impedance angle or lengthy algorithm iterations, and the identification depends on a high sampling accuracy and is time-consuming. In this study, a flexible parameter identification method based on the fusion of a machine learning model and a circuit model is proposed. Taking the primary voltage and current as input characteristic factors, support vector regression (SVR) is used to establish a machine learning model for coupling coefficient identification. In addition, the optimal model parameters are sought based on the grid search method. On the basis of coupling coefficient identification, the circuit model is used to realize the identification of the load resistance. Finally, the effectiveness of the proposed parameter identification method for a WPT system is verified by experimental results.