With the increasing growth of charging facilities for electric vehicles, the power supply capacity prediction for electric vehicles required by the social users in a district become a key issue for charging station’s construction planning. Combined with the data feature theory of depth learning, a machine learning based capacity planning forecasting method based on charging station capacity impact factors is presented. The method, considering environment impact factors such as traffic around charging station, regional development, power grid security and so on as the basis, trains and builds a neural network mapping model of service environment and charging requirement. Experiments show that the model can give the ideal charging capacity of the to-be-built charging station after analyzing the impact factors of the to-be-built power station surroundings, which could solve the sizing problem of the station charging capacity.