فهرست نشریات علمی دسترسی آزاد ایران

Lithium-ion batteries are essential for electric vehicles and renewable energy systems; yet, accurate battery state estimation remains critical for effective battery management systems. Although deep learning models have advanced state of health estimation, their comparative performance in accuracy, computational efficiency, and sustainability is underexplored and not discussed in detail. In this research, the proposed model achieves an optimal balance for real-time, resource-constrained BMS applications, resulting in high accuracy and superior efficiency compared to traditional models. Across four datasets, CS2_35–CS2_38, the proposed deep learning models were evaluated for predictive accuracy using evaluation metrics, including training time for efficiency and RMSE variability for generalizability. The proposed model outperformed the others, reducing RMSE by up to 26% compared to the traditional model, which exhibited consistent performance across all datasets. Furthermore, it trained 45–55% faster, reduced computational overhead by nearly half, and showed the lowest RMSE variability, indicating robust generalizability. These results highlight the proposed model as an ideal choice for resource-limited applications. By leveraging efficient models like CNN, this research advances state of health estimation while encouraging sustainable, eco-friendly BMS practices that minimize computational energy demands, aligning with Green AI principles for environmentally conscious battery management.