The location and capacity planning of electric vehicle charging stations are closely related to the travel characteristics of electric vehicle loads. Therefore, only when the charging load demand is reasonably predicted can an effective charging station planning result be obtained. To this end, this paper firstly defines the state space of electric vehicle charging load in multiple dimensions, and the probability matrix of state transfer of charging load can be established consequently. Furthermore, a Markov Chain Monte Carlo (MCMC) load forecasting model based on the multi-dimensional state space of electric vehicles travelling is proposed, the spatial-temporal prediction distribution of charging load is obtained by combining the real-time sample data. Then, a two-level programming model considering the economic benefits and user satisfaction of enterprise station construction is established. With the variable weight particle swarm optimization, the optimal site and scale of charging station can be determined. Finally, the simulation results can demonstrate the rationality and effectiveness of the model and method.