Kalman Filter — Machine learning fusion for core temperature estimation in Li-ion batteries

Conventional vehicles use fossil fuels to drive internal combustion engines, causing air pollution. Conversely, electric vehicles use various energy storage elements including batteries and fuel cells to power motor drives, offering minimal pollution. One critical aspect of battery operation is monitoring core temperature (T_c), which is used to estimate the internal temperature of a cell/battery and help prevent thermal run-away. This study estimates T_c by applying a Kalman Filter to the fundamental governing equations. Herein, T_c estimation is performed for two different battery chemistries: lithium polymer (LiPo) and lithium iron phosphate (LiFePO₄), and modeled using MATLAB/Simscape with appropriate step size and solver. Further, various regression models are applied to evaluate the prediction accuracy. Linear regression is found to have a good balance between model complexity and accuracy due to the linearity in T_c with respect to the charge/discharge currents. Thus, a detailed analysis of the linear regression model is performed to verify the predictions over trained and tested datasets. The prediction for LiFePO₄ batteries shows that the obtained prediction curve fits approximately 98–99 %, while that for LiPo batteries fits approximately 96–99 % during charging/discharging open circuit voltage–state of charge variations. Additionally, a comprehensive review of commercially available datasets for lithium-ion batteries is conducted to facilitate researchers in the field of battery management systems.