EV Charging Station Demand Analysis

Electric vehicle (EV) charging networks are usually monitored through separate hardware logs and spreadsheets, so demand surges, queue delays and capacity needs are noticed only after they have occurred. This paper presents the EV Charging Demand Analytics Platform, which brings the session records and station metadata of a charging network into one interactive web application. The data of 8,354 charging sessions across 20 urban and highway hubs are validated, merged and enriched with time, revenue, utilization and demand-level features. Six scikit-learn pipelines are trained: a Random Forest queue waiting-time regressor, a Gradient Boosting port-capacity advisor based on a 70% target-utilization rule, a Random Forest peak-hour classifier, a Gradient Boosting sessionenergy forecaster, a Random Forest high-demand hotspot classifier and a revenue regressor. A Streamlit and Plotly interface gives network key performance indicators, a station leaderboard, a day-of-week by hour-block density heatmap, a six-way non-restrictive filter and what-if prediction forms. The network totals are 8,354 sessions, Rs. 3,495,296 revenue, 226,603 kWh energy, 53.2% average load and 9.5 min average waiting time. The peak-hour and demand-tier classifiers reach 99.3% and 94.0% accuracy, while the waiting-time model does not beat the mean baseline (R2 = −0.008). The reasons for these results, including target leakage in several models, are discussed openly, and a plan for a stricter evaluation is given.