2024 | Vahid Safavi, Arash Mohammadi Vaniar, Najmeh Bazmohammadi, Juan C. Vasquez, Josep M. Guerrero
This paper presents a comparative study of multiple machine learning (ML) models for predicting the remaining useful life (RUL) and capacity fade of lithium-ion (Li-ion) batteries. The study aims to enhance the accuracy and reliability of RUL predictions, which are crucial for preventing system failures and optimizing battery management in safety-critical applications. The research involves three case studies using two distinct datasets, considering various operating conditions such as temperature, C-rate, state of charge (SOC), and depth of discharge (DOD). The ML models evaluated include Random Forest (RF), extreme gradient boosting (XGBoost), light gradient-boosting machine (LightGBM), multi-layer perceptron (MLP), long short-term memory (LSTM), and attention-LSTM. Hyperparameter tuning is applied to improve the performance of XGBoost and LightGBM models. The results demonstrate that the XGBoost with hyperparameter tuning (XGBoost-HT) model outperforms other models in terms of root-mean-squared error (RMSE) and mean absolute percentage error (MAPE) for both capacity fade and RUL prediction. The XGBoost-HT model also handles multi-feature multi-target (MFMT) feature mapping effectively, achieving lower RMSE and MAPE values compared to other models. The study highlights the superior performance of XGBoost-HT in predicting RUL and capacity fade, making it a reliable choice for battery management systems.This paper presents a comparative study of multiple machine learning (ML) models for predicting the remaining useful life (RUL) and capacity fade of lithium-ion (Li-ion) batteries. The study aims to enhance the accuracy and reliability of RUL predictions, which are crucial for preventing system failures and optimizing battery management in safety-critical applications. The research involves three case studies using two distinct datasets, considering various operating conditions such as temperature, C-rate, state of charge (SOC), and depth of discharge (DOD). The ML models evaluated include Random Forest (RF), extreme gradient boosting (XGBoost), light gradient-boosting machine (LightGBM), multi-layer perceptron (MLP), long short-term memory (LSTM), and attention-LSTM. Hyperparameter tuning is applied to improve the performance of XGBoost and LightGBM models. The results demonstrate that the XGBoost with hyperparameter tuning (XGBoost-HT) model outperforms other models in terms of root-mean-squared error (RMSE) and mean absolute percentage error (MAPE) for both capacity fade and RUL prediction. The XGBoost-HT model also handles multi-feature multi-target (MFMT) feature mapping effectively, achieving lower RMSE and MAPE values compared to other models. The study highlights the superior performance of XGBoost-HT in predicting RUL and capacity fade, making it a reliable choice for battery management systems.