AI-Driven Predictive Control for Load Balancing in EV Charging Networks

Abstract

EV charging networks are increasingly behaving as dynamic load systems in which simultaneous charging demand creates local peaks, feeder stress, queue imbalance, and uneven charger utilization. Existing studies have examined machine learning for charging behavior, coordinated charging, model predictive control, and reinforcement-learning-based station management, showing that AI can support both forecasting and adaptive decision making in EV charging systems. However, much of the literature still treats prediction and control as partially separate functions, leaving a gap in unified frameworks that directly couple short-horizon load forecasting with balancing action. To address this issue, this article presents an AI-driven predictive control framework for load balancing in EV charging networks based on rolling load prediction, chargerstate awareness, node classification, adaptive control action, and network-level balance optimization. The results show improved load-balancing performance, reduced localized peak formation, more stable charger utilization, and lower queue imbalance than static and reactive approaches. Overall, the study demonstrates that predictive AI-based control provides a scalable foundation for proactive balancing in EV charging networks.