Unsupervised Learning-Driven Charging Station Deployment: Optimizing Electric Vehicle Networks For Sustainable Development

Abstract

The fast growth of the electric vehicles (EVs) makes the strategic implementation of charging facilities necessary in support of a sustainable transport ecosystem. The conventional methods of charging station’s location typically take the position to be set beforehand without giving much consideration to the dynamic traffic patterns, the user behavior, and constraints in grid capacity. The paper suggests an unsupervised learning-based approach to optimal deployment of charging stations taking into consideration the clustering algorithms, dimensionality reduction techniques, and graph-based optimization to determine strategic points on which to place EV charging stations. Its methodology uses three tier methodology: data aggregation and pre-processing, unsupervised pattern discovery and multi-objective optimization. The framework employed is a K-means clustering with DBSCAN to perform spatial analysis, Principal Component Analysis (PCA) as a feature re- duction method and graphene neural networks to optimize network topology. It is experimentally validated on actual traffic data of metropolitan regions that the pro- posed UL-EVCD (Unsupervised Learning Electric Vehicle Charging Deployment) model reimbursement 23% of the coverage efficiency and 18% of the mean distance to the charging stations in comparison with traditional grid-based models. The framework also guarantees 94% grid stability and as much as possible integration of renewable energy towards a sustainable city development system.