Improved Capacity Estimation Method for Li-Ion Battery Cells Using a Modified Kalman Filter and a Cell Thermal Model

This study introduces a novel model for estimating Li-ion battery cell capacity by leveraging the battery’s temperature rise dynamics. The proposed model employs the Variational Bayesian Maximum Correntropy Cubature Kalman Filter (VBMCCKF) to achieve high-accuracy capacity estimation. By using a variational Bayesian approach, the filter effectively adapts to uncertainties in measurement noise, while the maximum correntropy criterion (MCC) handles outliers. Additionally, a new model correlating capacity with internal cell resistance is introduced to further enhance estimation. The VBMCCKF’s performance was compared against the Extended Kalman Filter (EKF), Cubature Kalman Filter (CKF), and Variational Bayesian CKF (VBCKF). Using the NASA dataset, which includes four batteries with varying initial conditions (ICs), the VBMCCKF consistently outperformed these algorithms. Notably, the filter achieved an average performance improvements ranging from 35% to over 200%, depending on the presence of outliers and the battery ICs. The VBMCCKF shows significant promise in improving the reliability of battery management systems (BMSs), optimizing battery usage, and enhancing the safety and durability of battery systems.