Accurate Parameter Identification and Robust State of Charge Estimation for Li-Ion Batteries Under Measurement Outlier Mitigation

Accurate state-of-charge (SOC) estimation in Li-ion batteries is crucial for optimizing control strategies and ensuring system reliability. Among existing estimation methods, Kalman filter-based techniques, often implemented with equivalent circuit models, are commonly used due to their high timeliness and precision. However, the presence of abnormal measurement noise may lead to bias in parameter identification (PI), which in turn undermines the reliability of SOC estimation. To overcome this issue, a generalized correntropy-based framework is proposed for joint PI and SOC estimation. Specifically, the total error induced by input–output noise is incorporated into the generalized maximum total correntropy (GMTC) criterion, which serves as the basis for a forgetting-factor recursive GMTC algorithm designed to achieve robust PI. This algorithm efficiently accommodates diverse noise distributions, significantly improving both convergence speed and PI accuracy. Furthermore, within the square-root cubature Kalman filter framework, an outlier detection and gain adaptation mechanism is integrated based on the generalized maximum correntropy criterion. This mechanism enables real-time anomaly detection through dynamic error tracking and adaptively adjusts the Kalman gain, thereby enhancing both the accuracy and responsiveness of SOC estimation. Finally, to reduce the sensitivity of the kernel width (Kw) for estimation performance, two adaptive Kw update strategies are independently developed for PI and SOC estimation methods, improving information utilization and overall robustness. Experimental results demonstrate the effectiveness of the proposed framework, showing that it achieves high estimation accuracy and computational efficiency under various operating conditions.