The output of wind turbine generator and photovoltaic has the characteristics of uncertainties because of the presence of random wind speed and solar irradiance. And the distribution network load level also has non-constant characteristics. The uncertainty has a significant impact on the distribution network during distribution network planning. This paper takes these uncertainties into account to the optimal allocation of distributed generatiions, divides a year running states of distribution network into 8760 hour-scenes using scene division method, and clusters scenes using the improved K-means clustering method. The objective function is the minimal comprehensive cost taking the environmental factor into account, and the improved PSO algorithm is employed to solve the model. Finally, the 33-bus distribution system is simulated to verify the effectiveness of the proposed model and method.