Clustering of Electric Vehicle Charging Behaviors Using Gaussian Mixture Models for Optimized Load Profiling

Electric vehicle (EV) integration into energy grid brings challenges for demand-side management (DSM) due to extremely dynamic and stochastic nature of the loads. This work introduces a novel approach for modelling EV load profiles using the unsupervised Machine learning model, Gaussian Mixture Model (GMM) to cluster charging behaviors at public, residential, and workplace locations. Key parameters for load profile modelling such as average charging duration, start time, and power demand are analyzed. Results show that GMM-based clustering captures diverse charging behaviors, enabling precise load predictions. Peak demand at public locations reached 3,797.16 kWh at 5:00 PM, with per EV consumption peaking at 7.17 kWh. Residential and workplace charging displayed distinct demand trends, peaking at 840.09 kWh at 8:00 PM and 1,822.31 kWh at 2:00 PM, respectively. These findings provide valuable insights for optimizing grid stability, enhancing demand response mechanisms, and improving infrastructure deployment. Future work involves incorporating demand-side management strategies, renewable integration, and predictive load modeling to foster efficient and sustainable electric mobility solutions.