A load and self/mutual inductance identification method based on PyTorch neural network for LCC-S WPT systems is proposed to address the low accuracy problem of parameter identification in magnetically coupled resonant wireless power transfer (WPT) systems. The method turned the load and self/mutual inductance identification problem of the WPT system into a deep learning nonlinear fitting problem. The nonlinear fitting problem is solved by training a neural network model. In this paper, the training method of the model is given and validated in simulation, and the identification result is good with an accuracy of 97.34%.