Battery Swapping Dispatch for Self-Sustained Highway Energy System Based on Spatiotemporal Deep-Learning Traffic Flow Prediction

Due to the complexity of traffic flow and the stochastic swapping behavior of electric vehicles (EVs), efficient battery dispatch is challenging. Therefore, the battery swapping dispatch framework based on traffic flow prediction is proposed to overcome this inconvenience. The framework is solved by minimizing the total transportation cost and satisfying the EV battery swapping requirement. Naturally, precise traffic flow prediction plays a vital role in efficient battery dispatch. Therefore, this article designs a deep learning prediction framework by leveraging the graph convolutional network (GCN) and the temporal convolutional network (TCN), named Spatiotemporal traffic flow network (STFNet). GCN is applied to learn the topology characteristic of the daily spatiotemporal traffic, which enables STFNet to capture the spatial feature. TCN is developed to acquire the daily traffic flow temporal dependence. Additionally, a pre-partition method based on K-means clustering is applied to improve the effectiveness of the battery dispatch framework. The experimental results indicate that the proposed battery dispatch framework is skillful. Due to the precise prediction of STFNet, the battery swapping dispatch based on STFNet prediction is the most economical, achieving a minimum of 25.82% reduction in the total transportation cost compared to benchmark models. Furthermore, the impact of the pre-partition method has been proven in the case of studies with a huge routing distance declining, dramatically reducing the total transportation cost and making the dispatch more reasonable.