CNN-Transformer-LSTM-based lithium battery health state prediction

Abstract Accurate prediction of the state of health (SOH) of lithium batteries is a core challenge to ensure the safe operation and lifetime optimization of batteries. To address the problems of insufficient modeling capability of a single model and weak multi-feature synergy of a hybrid model, this study proposes a hybrid deep learning model based on CNN-Transformer-LSTM. Incremental Capacity (IC) curves are extracted using capacity increment analysis, while constant current charging time, average charging voltage, and average charging current are extracted from the charging curves as healthy features and highly correlated features are screened by correlation analysis. CNN is first used to extract local features from the data while reducing redundant information; then Transformer is used to capture global dependencies and identify long-term healthy trends; and finally, long-term dependencies and trends in time are captured by LSTM to ensure smooth and stable predictions. The experimental results demonstrate that the model achieves positive results on the NASA lithium battery dataset, improving the prediction accuracy and generalization ability compared to existing state-of-the-art models.