Optimizing Neighborhood Electric Load Forecasting with Advanced Machine Learning Algorithms: A Pathway to Sustainable EV Charging Infrastructure

Chenglong Tang
School of Automation Jiangsu University of Science and Technology, zhenjiang,212100, jiangsu, China
Jin Zhao
School of Automation Jiangsu University of Science and Technology, zhenjiang,212100, jiangsu, China

Abstract:

This study introduces a sophisticated methodology to optimize load management at electric vehicle (EV) charging stations, leveraging the power of advanced machine learning tech- niques. Central to our approach is a detailed data collection and preprocessing phase, utilizing the comprehensive” EV Charging Station Data” from the U.S. Department of Energy’s Alternative Fuel Data Center.  We categorize and preprocess this data meticulously, focusing on the charging station’s capacity and geographical distribution. Following data cleaning, normalization, and feature engineering, we employ the K-means++ algorithm for clustering analysis, segmenting the charging stations into meaningful groups based on their load profiles. For each cluster, Support Vector Regression (SVR) is then utilized to forecast the charging demand accurately. These predictive insights feed into an optimization model aimed at minimizing peak load across the network while ensuring all EVs receive adequate charging.