Data-Driven Fault-Tolerant Control Framework for EV Dynamic Wireless Power Transfer System Based on Self-Learning Predictor

This paper aims to develop a constant output voltage controller for dynamic wireless power transfer systems(DWPTSs) incorporating sensor noise filtering and fault tolerance. DWPTSs are designed to alleviate range anxiety in electric vehicles(EVs); however, the output voltage fluctuations are their significant drawback compared to static charging mode. Additionally, DWPTSs also face sensor measurement noise and potential faults that exacerbate system instability. To mitigate above challenges, a data-driven fault-tolerant control framework is designed for DWPTS based on a self-learning predictor, which implements constant voltage regulation with enhanced noise and fault immunity. Specifically, a self-learning predictor is integrated into the feedforward loop of a high-gain extended state observer (ESO) to filter sensor noise. Then, a data memory stack is constructed to store predicted states and estimated disturbances, and a concurrent learning algorithm is introduced to recover control gains online. Finally, a composite anti-disturbance control law is implemented to generate the required control signals for the charging circuit. A notable advantage of this scheme is its ability to simultaneously address both sensor noise and faults, ensuring a constant output voltage during EV driving. Experimental results validate that the designed control framework effectively eliminates output voltage fluctuations and measurement noise, even in the presence of sensor faults.