A battery health state index-driven digital twin framework for battery condition assessment and predictive maintenance in electric vehicle charging environments

The increasing deployment of electric vehicles (EVs) has intensified the need for intelligent battery monitoring and predictive maintenance solutions capable of ensuring battery reliability, safety, and operational efficiency. Although battery State of Health (SOH) estimation has received considerable research attention, existing approaches primarily focus on degradation assessment while providing limited integration of charging behavior characteristics and maintenance-oriented decision support. To address this limitation, this study proposes a Battery Health State Index (BHSI)-driven digital twin framework for battery condition assessment and risk-aware maintenance prioritization. Laboratory battery cycling data obtained from an Arbin battery testing system were utilized to estimate battery health conditions and develop a normalized BHSI representation. Subsequently, a Digital Twin Score (DTS) was formulated by integrating battery health, energy consumption, and charging power characteristics into a unified assessment metric. Random Forest and Extreme Gradient Boosting (XGBoost) models were employed to estimate the normalized Battery Health State Index (BHSI) and evaluate the machine learning component of the proposed framework. Experimental results demonstrated that the Random Forest model achieved superior performance, with a coefficient of determination (R²) of 0.998, a mean absolute error (MAE) of 1.079, and a root mean square error (RMSE) of 1.310, outperforming XGBoost. The predicted BHSI values were subsequently integrated with charging behavior characteristics to compute the proposed Digital Twin Score (DTS). Correlation analysis revealed a strong inverse relationship between BHSI and DTS ( r = − 0.788), demonstrating the effectiveness of the proposed score in representing battery health and operational characteristics. The analysis of the feature importance of analysis further indicated that battery capacity and energy-related variables were the dominant contributors to BHSI prediction. The developed framework successfully classified charging sessions into low-, medium-, and high-risk categories, enabling maintenance prioritization and operational decision support. Overall, the proposed BHSI–DTS digital twin framework provides an interpretable and effective solution for battery condition assessment and intelligent maintenance support within EV charging environments.