A robust diagnosis scheme for voltage sensor fault in lithium-ion battery system using maximum correntropy optimized Kalman filter

Reliable measurement information including terminal voltage is critical to ensure effective lithium-ion battery energy management. In response to the limitations of model and residual-based battery fault diagnostic techniques concerning external noise suppression and constraints imposed by fault threshold settings, this paper proposes a robust method for voltage sensor fault diagnosis. This approach aims to simultaneously mitigate measurement noise interference and reduce the dependence of state of charge (SOC) residual on the predefined fault threshold. First, the maximum correntropy criterion (MCC) is utilized to correct the variance of the measurement noise within the extended Kalman filter (EKF), and the measurement update parameters are reconstructed to generate the MCC-adaptive EKF (MCCAEKF) method. Then, robust SOC estimation is obtained iteratively by the MCCAEKF based on the higher-order moment information of measurement data, and the corresponding SOC residual is generated from the SOC estimation with a voltage sensor fault. Finally, the fault diagnosis effectiveness is evaluated based on the change rate of the SOC residual curve and the time point at which the curve triggers the multiple fault thresholds. The feasibility of the presented diagnostic scheme is verified by several experimental results under different conditions, temperatures and fault scenarios.