Luo, R., Zhang, Y., Zhou, Y., Chen, H., Yang, L., Yang, J., & Su, R. (2021). Deep Learning Approach for Long- Term Prediction of Electric Vehicle (EV) Charging Station Availability. 2021 IEEE International Intelligent Transportation Systems Conference (ITSC). doi:10.1109/itsc48978.2021.9564633
Traffic prediction with high accuracy has significance towards traffic facilities scheduling, adaptive traffic control logic, even the urban economic development. EV charging station availability prediction is one of the traffic facilities usage forecasting challenges in the urban planning. With the accurate long-term prediction of each EV charging station availability, drivers could schedule the charging activities wisely and avoid range anxiety. Data driven models are commonly applied to deal with the similar questions. However, due to the complex charging station distribution with topological road network structure and time-variant charging station availability, many widely used algorithms, such as recurrent neural network (RNN), could only extract the pure time-series information during the model training. One direct effect is that the prediction accuracy decreases rapidly over time. The Spatial-Temporal Graph Convolutional Network (STGCN) that consists the Graph Convolutional Network (GCN) considering edge connectivity and the Gated Recurrent Unit (GRU) is proposed to process both spatial and temporal dependence of relevant transportation data in this paper. Then, the spatial-temporal model is deployed on a real testing dataset in Dundee City and experiments show a better performance for EV charging station availability forecasting over a long-term period compared with other baselines and the possibility of real-time application.
This research / project is supported by the A*STAR - RIE2020 AME IAF-PP
Grant Reference no. : A19D6a0053