A Robust Misalignment Recognition Algorithm Using Multi-Domain Network Fusion Model for Wireless EV Charger

In this paper, the multi-types misalignment recognition (MR) problem of electric vehicles (EVs) is studied, with a particular focus on the adverse impact of height fluctuations on the recognition algorithm. First of all, novel detection coils are used to collect the induced voltage signal between the energy transmission coils, and a multi-type misalignment coordinate system is established according to SAE J2954. Then, to enhance the robustness of the multi-types MR algorithm in height fluctuation of the energy transfer coil, a multi-domain network model based on transfer learning and model fusion is proposed. Furthermore, a large number of simulation datasets are used to optimize the model parameters, thereby improving the transfer learning effect and increasing the generalization capability of the multi-domain network model. Finally, several samples are selected uniformly within the height fluctuation of $160 \pm 20$ mm for testing. The results show that the proposed multi-domain network model can effectively solve the multi-types MR problem even under height fluctuation.