Advanced Cluster-Based Load Forecasting and Peak Demand Management for Electric Vehicle Charging Networks

The rapid adoption of Electric Vehicles (EVs) necessitates advanced solutions for flexible charging (FC) systems to meet diverse user needs and optimize grid efficiency. This study introduces a data-driven, cluster-based load forecasting framework using machine learning techniques. Charging sessions are categorized into residential, workspace, and instantaneous station clusters via HDBSCAN and Elbow methods. Regression models, including Random Forest, SVM, and Logistic Regression, are applied to forecast load demand, leveraging refined datasets and feature engineering techniques. The proposed methodology, implemented in Python and MATLAB, integrates a Battery Energy Storage System (BESS) for peak demand curtailment, demonstrating scalability for 1000 EVs. Results highlight significant improvements in load prediction accuracy, operational efficiency, and demand response outcomes. This integrated framework provides a scalable, adaptable, and efficient solution for EV charging systems, aligning with future energy management needs.