Accurately forecasting electric vehicle charging demand can be a foundation for research on large scale electric vehicle access to the grid. Existing methods lack consideration of the impact of the congestion factor. An electric vehicle charging demand forecasting method based on cluster analysis is proposed. On the basis of analyzing the influence factors and establishing a probability distribution model, a cluster analysis of two-dimensional travel feature data composed of mileage and travel time of each trip is made. It mines a congestion factor that is not available by statistics, considers the influence of congestion on the state of charge under different conditions and superimposes the variable into the forecasting model. Taking Beijing as an example, daily charging load curves of weekdays, weekends, summers and winters are predicted and comparative analysis is made. Results show that the method can improve the accuracy of charging demand forecasting to some extent. This work is supported by Science and Technology Project of State Grid Corporation of China (No. 52110417001G).