A Cloud-Edge Digital Twin Architecture for Adaptive Battery Health Management in Sustainable Transport Systems

This paper presents a cloud-edge digital twin framework designed to enhance battery lifecycle management within electric vehicles, contributing to sustainable transportation and advanced battery system engineering. The architecture integrates a static state-of-health (SOH) model trained offline with a dynamically retrained state-of-charge (SOC) model updated periodically via cloud-based machine learning. Using a public NASA battery dataset, the system employs random forest, light gradient boosting, and deep neural networks to achieve SOH estimation errors below 1.8% RMSE and SOC errors under 0.81% RMSE while maintaining inference times under one second—compatible with onboard BMS deployment. The retrainable SOC model adapts to aging effects, ensuring continued accuracy as battery capacity degrades. This adaptive digital twin supports predictive maintenance, real-time health monitoring, and optimized battery utilization, aligning with smart manufacturing and sustainable energy system goals by extending operational life and improving reliability in EV applications.