A Noise-Resilient Observer for Enhanced Surface Temperature Estimation of Li-ion Battery Cells

Reliable temperature monitoring techniques for lithium-ion (Li-ion) battery cells is of utmost importance in electric vehicles (EVs) as EVs' battery cells face extreme thermal dynamics. This paper presents a new approach for estimating the surface temperature of Li-ion battery cells using the variational Bayesian-based maximum correntropy cubature Kalman filter (VBMCCKF). The proposed method combines the advantages of the Variational Bayesian technique, with its adaptation to variation in the measurement noise, and the maximum correntropy criterion, with its robustness to measurement outliers. Nonlinear models are derived through polynomial regression to exploit the dynamics and measurements models describing the evolution of the battery surface temperature. The proposed method is evaluated and validated through publicly available experimental datasets. Results indicate even with measurement outliers, comparative analysis confirms VBMCCKF's superior accuracy in estimating Li-ion battery surface temperature, achieving a remarkable 77% average performance improvement over the Extended Kalman Filter (EKF), Cubature Kalman Filter (CKF), and Variational Bayesian-based Cubature Kalman Filter (VBCKF).