A Hybrid Model Based on Transformer-LSTM for Battery Health Prediction

With the widespread application of lithium-ion batteries in electric vehicles and energy storage systems, accurate prediction of the State of Health (SOH) and Remaining Useful Life (RUL) of batteries is crucial for improving system reliability and safety. This paper proposes a hybrid deep learning model based on Digital Twin, combining Transformer and Long Short-Term Memory (LSTM) architectures, trained and validated using the public CALCE battery dataset. The model captures the temporal dependencies in battery degradation through LSTM and enhances the extraction of key features by introducing the self-attention mechanism of Transformer, thereby improving the robustness of long-term predictions. The Digital Twin framework enables dynamic interaction between the physical characteristics of the battery and the data-driven model, making the prediction results more closely aligned with real-world aging behavior. Experimental results demonstrate that the proposed hybrid model Transformer-LSTM outperforms traditional methods in battery prediction, reducing the MAE by over 15% while exhibiting strong generalization capabilities under different charge-discharge conditions. This study not only provides a high-precision prediction method for battery health management but also validates the application potential of Digital Twin technology in the field of battery prognostics, offering theoretical support for the optimized design of intelligent battery management systems.