Forecasting EV Charging Demand: A Graph Convolutional Neural Network-Based Approach

Electric vehicles (EVs) are expected to revolutionize the global transportation sector by promoting sustainability and eco-friendliness. The continuous proliferation of EVs requires an expansion of the existing charging infrastructure to meet the corresponding increase in electricity demand. Such an expansion requires an accurate forecasting of charging demand in both space and time domains for a well-planned allocation of charging stations (CSs). This paper proposes a Graph Convolutional Neural Networks (GCNN) based approach combined with a Long Short-Term Memory (LSTM) to predict the future charging demands of EVs. The proposed architecture fuses the benefits of both GCNN and LSTM to extract the underlying spatio-temporal features from the collected dataset. The training dataset reflects the coupling between the power and transportation systems, and thereby it helps the proposed deep learning architecture to capture the spatio-temporal patterns of the inter-connected environment. A comparative analysis is conducted with other state-of-the-art EV charging load prediction models to assess the prediction performance of the proposed load forecasting strategy.