Answers are grounded in this paper and cite their sources.
Machine learning methods improve state of charge estimation in lithium-ion batteries, suggesting enhanced battery management systems.
Accurate estimation of the State of Charge (SoC) of batteries is critical for the optimal performance and safety of electric vehicles and portable electronic devices. In this study, several machine learning regression models were investigated for state-of-charge (SoC) estimation, including Linear Regression, Support Vector Machines (SVM), Regression Trees, and Neural Networks. The publicly available LG 18650HG2 dataset [1] was utilized, comprising over 670,000 data points collected during charging and discharging cycles. Input features included voltage, current, temperature, average voltage, and average current. The models were implemented using MATLAB, and their performance was evaluated using standard metrics: Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²). The best-performing model—a wide neural network—achieved an RMSE of 0.0211, MAE of 0.0109, and an R² of 0.9959. These results indicate that Neural Networks outperform traditional regression techniques in terms of prediction accuracy and robustness.
No takes yet. Share an insight, caveat, or question.
Nossier et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: