<p>Electric vehicle charging demand prediction is an important prerequisite for researching the interaction between electric vehicle (EVs), the power grid, and the transportation network. However, most existing work does not use real-world traffic data to analyze EV charging demand. Therefore, this paper proposes a data-driven model for EV charging demand prediction. Firstly, original EV travel trajectory data is used for data mining and integration modeling, including area selection, spatial grid modeling, trajectory data mapping, POI retrieval data identification, urban functional area clustering, and traffic network modeling. Through modeling and data processing, regenerated feature data such as functional area division, travel pattern distribution, and actual driving paths are obtained. Secondly, considering the mobility load characteristics of EVs, a single EV model is established, which contains driving characteristic parameters and charging characteristic parameters. And the evaluation model of virtual energy storage schedulable operation region of EV cluster is established. Finally, with a certain area as an example, path planning experiments and charging demand experiments in different scenarios were designed. And a day-ahead peak shaving strategy with EVs is proposed. The results show that the proposed model can effectively predict the spatial and temporal distribution characteristics of charging demand and load for different date types and functional areas. It also lays the theoretical foundation for subsequent research on EV charging control and guidance.</p>