The integration of machine learning techniques into battery state‐of‐health (SOH) assessment has offered improved flexibility and adaptability under varying operating conditions; however, challenges remain in capturing dynamic temporal dependencies, ensuring feature interpretability, and achieving scalable optimization. To address these limitations, this work introduces a robust and scalable framework that overcomes the constraints of existing data‐driven models—particularly their limited temporal learning capability and suboptimal parameter tuning—which often hinder accurate battery health‐state classification across diverse usage scenarios. The proposed methodology integrates data preprocessing, multiscale temporal graph embedding (MTGE) for feature extraction, temporal attention transformers (TATs) for classification, and a hybrid whale–falcon optimizer (HWFO) for parameter optimization. Comprehensive preprocessing mitigates data inconsistencies, while MTGE effectively captures both local and global temporal dependencies, enhancing feature representation. The TAT model efficiently learns complex time‐series patterns, achieving a classification accuracy of 99.5%, with precision, recall, F1‐score, and specificity of 98.9%, 99.2%, 99.1%, and 99.4%, respectively. Furthermore, HWFO exhibits superior optimization performance, reducing MAE, MSE, and RMSE to 0.015, 0.0023, and 0.048, respectively—outperforming traditional architectures. Overall, the results highlight the robustness, accuracy, and scalability of the proposed framework, underscoring its suitability for real‐time battery health monitoring and predictive maintenance applications.
G et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: