Robust battery state of charge estimation incorporating modified correntropy Kalman filter with adaptive kernel width and weighted multi-innovation compensation

Efficient implementation of state of charge (SOC) estimation is an essential core function of the energy storage system in electric vehicles. The uncertainty under complex operating conditions of lithium-ion batteries (LIBs) can easily lead to measurement data being interfered with by various types of noise. To enhance the adaptability of SOC estimation method in complex environments, the paper proposes an advanced estimation algorithm that integrates adaptive kernel width and weighted multi-innovation compensation on the basis of modified correntropy criterion-based extended Kalman filter (MCCEKF). Firstly, error covariance and observation noise are added to the MCC objective function for optimization, and the iterative calculation of the MCCEKF is derived using the weighted least squares (WLS) method. Secondly, the Pseudo-Huber weight function is introduced to adjust the kernel width of MCC, utilizing observation information to mitigate the impact of abnormal data on the estimation outcomes. Finally, considering the characteristic differences in temporal and numerical aspects among innovations, multiple weight factors are assigned to them to strengthen the correction effect of various innovation data on SOC estimation during the measurement update stage. The validity of the optimization method is verified using multiple conditional data, and the effects of kernel width, multiple innovation, and initial error on SOC estimation are analyzed. The results of the experiment demonstrate that the method can reduce the sensitivity of the estimator to multiply types of noise originating from voltage and current. Compared with the baseline methodology, the proposed method can achieve better estimation performance across various evaluation metrics.