Hybrid XGBoost–LSTM model for state-of-health prediction of lithium-ion batteries under different thermal conditions

Accurate monitoring of the State of Health (SoH) of lithium-ion batteries is critical for the reliable and safe operation of energy storage systems, particularly in grid-level applications where battery performance directly impacts the stability and efficiency of the grid. Traditional SoH estimation methods face challenges due to the complex and dynamic behavior of batteries, especially under varying thermal conditions. This paper proposes a hybrid machine learning approach that combines XGBoost with Long Short-Term Memory (LSTM) networks to predict the SoH of lithium-ion batteries. The XGBoost component models the relationship between engineered features, while the LSTM network captures the temporal degradation patterns in time-series data. The model is validated across a broad temperature range of 5°C–35°C to account for thermal variability. To rigorously assess generalization, the model is evaluated using ten-fold cross-validation and leave-one-temperature-out (LOTO) analysis, ensuring robustness across unseen conditions. Experimental results demonstrate that the proposed framework outperforms traditional methods, including electrochemical models and support vector machines, in terms of accuracy and robustness, particularly in fluctuating thermal conditions. The model's interpretability is enhanced through TreeSHAP analysis, providing actionable insights into battery degradation mechanisms. This approach offers a reliable and scalable solution for real-time SoH monitoring, predictive maintenance, and optimal management of energy storage systems.