Electric vehicle charging demand forecasting taking into account multiple time scales and dynamic road network information
Minan Tang, Changyou Wang, Hanting Li, Xi Guo, Kaiyue Zhang, Mingyu Wang, Xuemei Liang
Abstract
Spatial-temporal prediction of electric vehicle (EV) charging loads is crucial for the optimization of urban charging station scheduling and layout planning. However, with the rapid growth of EV penetration, the spatial-temporal distribution of charging loads has become increasingly random. This paper proposes a novel model that combines data-driven and model-driven methods for predicting regional EV charging loads across multiple time scales. First, ArcGIS is used to divide urban land parcels into different areas and establish a time-flow model based on road intersection characteristics. Secondly, dynamic Dijkstra algorithms, unit mileage power consumption and other method models are introduced to construct a perfect single EV mobility model. Then, the advantages of the Bass diffusion model and the GM (1,1) model are integrated through a recursive weighting mechanism to establish a hybrid prediction model for EV ownership. Finally, the Monte Carlo method is used to simulate the charging behavior of EV in a region of Lanzhou City, and compared and analyzed with the speed-flow model and the fixed power consumption model. The results indicate that the RMSE of Bass-GM(1, 1) decreased by 28.80 % compared to Bass model, while the prediction accuracy of the MC model for residential evening peak loads improved by 12.02 %. The feasibility and validity of the proposed model are verified, and it can more intuitively reflect the demand distribution of EV charging load short-term and development trend medium- and long-term.