A machine learning-based optimal charging strategy for PV-assisted electric vehicle systems incorporating second-life...

The rapid expansion of electric vehicles (EVs) and residential photovoltaic (PV) systems has created new challenges in battery charging management, particularly due to the variability of renewable energy, grid limitations, and battery aging effects. In this work, we present a machine learning-based Energy Management System (EMS) designed for PV-assisted smart charging of EVs and second-life batteries. The proposed system estimates the Optimal Charging Duration Class (OCDC) using an XGBoost model trained on real-time operating variables such as state of charge (SOC), battery temperature, available PV surplus, and degradation-related indicators. Rather than relying on conventional continuous power control, the proposed approach adopts discrete charging modes that dynamically adjust to operating conditions, aiming to improve both energy utilization and battery health. Degradation considerations are incorporated in a practical, control-oriented manner by avoiding operating regions associated with accelerated aging, instead of explicitly modeling electrochemical processes. The system is evaluated within a simulation framework that includes realistic PV generation profiles, load demand, and thermal behavior. Although hardware implementation is not yet included, it is identified as an important direction for future validation. The results show that the proposed EMS increases PV self-consumption by around 22% while reducing exposure to high-temperature operation, high state-of-charge conditions, and unnecessary cycling. These outcomes indicate a potential reduction in degradation risk, especially for second-life battery applications. Overall, the study demonstrates that integrating machine learning with real-time energy management can provide an efficient and health-aware solution for smart charging in residential and hybrid renewable energy systems.