Enhancing Battery State of Health Prediction via Transfer Learning Method

To address the challenge of limited full-life cycle data availability for specific battery types, this study proposes a transfer learning framework to improve the prediction accuracy of battery State of Health (SOH). By leveraging a pre-trained CNN-BiLSTM-Attention model, we extract generic features through shallow networks and abstract features via deep networks, enabling effective knowledge transfer from a source domain (battery type A) to a target domain (battery type B). Experimental results demonstrate that compared to direct prediction methods, the proposed transfer learning-based approach achieves remarkable improvements in both prediction accuracy and computational efficiency, thereby validating the effectiveness and feasibility of the method.