Battery SOH Prediction Under Different Conditions via MBLSTM and iTransformer With Anomaly Detection and Explainability

The prediction of the State of Health (SOH) of lithium-ion batteries is essential for applications such as electric vehicles, renewable energy storage, and portable electronic devices. Existing methods struggle with accuracy and interpretability under mixed operating modes and nonlinear degradation. To address this challenge, this paper proposes a hybrid deep learning model based on parallel multi-channel bidirectional Long short-term memory (MBLSTM) and inverted transformer (iTransformer). The model extracts local temporal features through MBLSTM, and employs iTransformer, which leverages an attention mechanism, to identify global dependencies among variables. This integration enables the model to combine local dynamics with global interactions, thereby improving predictive performance. Additionally, an anomaly detection mechanism and the LIME (Local Interpretable Model-agnostic Explanations) method are introduced to analyze key influencing factors at anomalous prediction points.Comparative experiments on four real-world battery datasets with different charging/discharging strategies show that the proposed model outperforms existing methods and provides interpretable insights into anomalous predictions.