Research on the Planning of Electric Vehicle Charging Stations Based on Multi-View Graph Convolutional Networks

In 2020, electric vehicle charging infrastructure was included in “new infrastructure,” and its incremental growth has continued rapidly. However, there is still significant room for improvement in public charging pile density. Public charging infrastructure needs to be strategically located and expanded within urban spaces based on the specific needs and charging behaviors of different types of users. This study first utilizes heterogeneous data from multiple sources to analyze the behavioral characteristics of electric vehicle users and the impact of urban spatial attributes on charging demand. Subsequently, a modularity-optimized community detection algorithm is employed to partition urban spaces. Finally, a multi-view graph convolutional neural network, integrated with the XGBoost algorithm, is established to predict charging demand in specific areas within these communities. The model presented in this paper outperforms traditional graph neural networks in terms of regression and classification performance.