Improved real-time Lithium-ion battery parameter extraction and prediction analysis using ML framework

Abstract Lithium-ion batteries (LIBs) are vital energy sources in power systems and daily life, and have relatively high power and energy densities. The expanded use of LIBs has paved the way for advancements in the automobile and energy-storage sectors. Hence, this is a vital transition element for green energy generation in power scenarios. This study investigates the safety aspects in real time and the effects of internal and external parameters on battery performance characteristics. MATLAB/Simulink was used to create an ideal battery model and investigate the effects of several parameters on a standard 18650 Panasonic LIB. In addition, the state of charge (SoC) and State of Health (SoH) were calculated by employing a variety of techniques, including MATLAB/Simulink with an Adaptive Kalman Filter (AKF), in the Real-Time Target Machine (RTTM), and the values were analysed using predictive Machine Learning (ML) algorithms. Temperature and C-rate were the external characteristics examined in this study. The circuit was tested for temperature vectors ranging from −50 °C to 100 °C, and the Urban Dynamometer Driving Schedule (UDDS) drive cycle was set as the load for the emulator and HIL simulations. The percentage errors for the estimated values of the SoC and SoH were also measured. The optimal machine learning algorithm for these parameter estimations was identified in comparison with the seven selected ML models. Evaluation metrics such as the Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) were used to compare the models.