Data-driven Traffic Prediction and Pricing Scheme for Shared Electric Vehicles

To improve the economy and operational efficiency of a shared electric vehicle (SEV) system, it is necessary to fully explore the user's response mechanism to the pricing and the user's waiting behavior during charging. Simultaneously, the accurate prediction of user traffic demand is a fundamental issue in shared pricing decision-making. Based on this, a data-driven user traffic prediction model was established. By combining a graph convolution neural network and long-term and short-term memory models, the temporal and spatial characteristics and dynamic correlations of SEV users' demand can be captured, and an accurate multi-time step traffic flow prediction can be realized by combining the time and weather characteristics. Second, considering the user's adaptive waiting behavior in a shared system, a service quality model of the shared system was established based on the user's waiting cost. To maximize the profit and service quality of the sharing system, a multi-objective pricing decision-making model was established for determining the price of shared electric vehicles at different times and paths. Finally, the accuracy of the proposed users' flow forecasting model and the effectiveness and economy of the multi-objective pricing model were demonstrated based on the case of the EVCARD traffic system in the Hongkou district of Shanghai City. The experimental results show that the proposed SEV multifactor fusion demand prediction model could obtain more accurate multi-time-step traffic prediction than common existing prediction methods. In addition, by setting the space and time change price signals for shared driving reasonably, the operational profitability of shared electric vehicle systems can be improved, and the operational demand for a better balance between system profit and user service quality can be realized. © 2023 Xi'an Highway University. All rights reserved.