Battery-Insight-PSO: A machine learning model for accurate prediction of state of health and remaining useful life in lithium-ion batteries

Condition based monitoring (CBM) of the lithium-ion (Li-ion) battery has become very popular in recent years because of its wide usage as an energy storage for smart grids, power sources in various industrial equipment, electric vehicles (EVs), etc. As a result, predicting the state of health (SOH) and the remaining useful life (RUL) of Li-ion batteries with high accuracy ensures optimal performance and safe utilization, preventing non-scheduled failures and saving maintenance costs. This paper illustrates the significance of highly accurate SOH and RUL prediction for Li-ion batteries. This paper proposes a model called Battery-Insight-PSO, which employs the Extreme Gradient Boosting Regression (XGBoost) machine learning algorithm to forecast SOH and RUL. In this study, the Particle Swarm Optimization Algorithm (PSO) is used to optimize different parameters of XGBoost for ensuring precise and reliable predictions of SOH and RUL for Li-ion batteries. In this study, the National Aeronautics and Space Administration (NASA) Li-ion Battery Aging Datasets and the NMC LCO 18650 battery dataset from the Hawaii Natural Energy Institute (HNEI) were analyzed. Additionally, the performance of Battery-Insight-PSO was compared with other machine learning algorithms. Machine learning models were evaluated using various performance metrics. The estimation errors of Battery-Insight-PSO are very low, which means that this model can be highly accurate in predicting SOH and RUL. Moreover, the R 2 scores for the training and testing sets of this model also show high consistency with 0.9998 for each dataset, demonstrating high accuracy and reliable performance. • Battery-Insight-PSO predicts SOH and RUL for NASA and HNEI datasets. • The model is compared with different approaches and related literature. • PSO tunes XGBoost model’s hyperparameters to improve model performance. • Battery-Insight-PSO achieves high R² and low MAE, MSE, RMSE, MAPE, NRMSE. • A comparative evaluation confirms its strong SOH and RUL estimation ability.