Based on analysis of the water consumption and water consumption per capita in Shanxi Province for years, the water demand prediction model of radial basis function neural network was set up. The nearest neighbor-clustering algorithm was adopted to decide the width of radial basis function, the cluster centers were chosen, and the weight values were calculated. Abundant water demand predicting factors were used as the input data of the model, and the RBF neural network output the water demand predicting values. The predicting results show that the model has faster calculating speed and higher predicting accuracy, which can provide basis for water resources planning and allocation.