Utilization of machine learning approaches for enhancing sustainability of electric vehicles with optimization of lithium-ion battery health status

Lithium-ion batteries play a central role in electric vehicles (EVs), renewable energy storage, and modern power networks due to their high energy density and efficiency. However, accurately estimating their State of Health (SOH) remains a major challenge, as battery degradation is governed by complex electrochemical and thermal processes influenced by dynamic operating conditions. Although numerous machine learning (ML) approaches have been proposed, many existing methods rely on narrowly scoped datasets, struggle with nonlinear degradation behavior, or lack robustness under real-world variability. These limitations hinder their applicability in large-scale sustainable energy systems. To address these gaps, this study introduces an Ensemble Stacking Regressor designed to provide accurate, generalizable, and noise-resilient SOH estimation. The framework integrates Extreme Gradient Boosting Random Forest (XGBRF), Histogram-based Gradient Boosting Regressor (HGBR), and Extra Trees Regressor (ETR), combined through a Support Vector Regression (SVR) meta-model. Extensive feature extraction from voltage, current, temperature, and internal resistance profiles enables the model to capture multi-dimensional degradation patterns essential for reliable SOH assessment. Experimental results on four MIT battery datasets reveal consistently high accuracy, with R² values above 0.990 and low RMSE, MAE, and RSE metrics. Additional validation on NASA and Oxford datasets confirms strong generalization, while noise-perturbation tests demonstrate high resilience under uncertain measurement conditions. These findings indicate that the proposed framework could enhance battery reliability, support smarter energy management strategies, and strengthen the integration of EVs and storage systems into sustainable energy networks. The model can offer a robust and transferable solution for improving SOH monitoring across diverse applications.