| Summary: | INTRODUCTION: The uneven spatiotemporal distribution of electric vehicle (EV) charging demand and the limitations of traditional methods in adapting to dynamic correlations pose significant challenges to charging infrastructure management.
OBJECTIVES: This study aims to propose a collaborative scheduling and path planning method for charging pile groups to optimize system efficiency, reduce user waiting time, lower costs, and balance grid load.
METHODS: A spatiotemporal heterogeneous graph integrating charging station states, road networks, and grid conditions is constructed. A Graph Attention Network (GAT) with a multi-head attention mechanism is employed to dynamically capture node correlations. A joint optimization model for scheduling and path planning is established, utilizing an extended A* search algorithm within a multi-objective framework.
RESULTS: Experimental results demonstrate that, compared to the Constant Power Method (CPM) and a traditional Graph Convolutional Network (GCN) method, the proposed GAT-based method reduces average user waiting time by 30-40%, decreases total system cost by 17.9%, improves the load balancing index to 0.45, reduces grid load variance by 42.4%, and successfully serves 1038 EVs.
CONCLUSION: The proposed method effectively addresses the collaborative optimization of charging pile group resources, providing an innovative solution for building an efficient and stable EV charging network.
|
|---|
