Research on Spatio-temporal Charging Load Prediction and Charging Station Planning of Urban Electrical Vehicles

In allusion to the fact of low utilization rate of partial urban electric vehicle (abbr. EV) charging stations and low user satisfaction due to unreasonable construction and planning, to meet the needs of large-scale EVs charging stations planning under the dual carbon target, based on the fitting of Monte Carlo simulation with echo state network (abbr. ESN) a method to forecast urban EV space-time charging load was proposed to further conduct the research on EV charging station planning. Firstly, considering the structure of urban traffic network and the main functions of different regions, the area to be planned was divided into meshes as the alternative locations of charging stations to be built. The Monte Carlo method was utilized to simulate various modes of travel chains for all kinds of EVs to obtain EV charging load data set in each mesh. To fit the diversified distribution characteristics of EV charging load in each mesh, an ESN learning algorithm-based EV temporal and spatial charging load prediction model was established to realize the prediction of EV temporal and spatial charging loads within the region to be planned under a certain inventory of EV. Further taking the maximum predicted charging load and so on within the mesh to be planned as constraint conditions, and taking the minimized cost of construction, operation and maintenance, the charging and trip cost of EV users and the composite cost of the loss of distribution network as objective functions, an EV charging station planning model was constructed. The swarm clustering optimization was used to solve the constructed model to obtain the sites in the region to be planned for the construction of charging stations, the amount and the capacities of them. Finally, taking EV charging load prediction for a certain urban district and the planning of the charging station planning as example, the effectiveness of the proposed method and the constructed model is verified.