The maintenance of power supply network is more complex, there are so many influence factors that influence the cost of it in various ways, which caused the cost difficult to control. In order to solve this problem, the data analysis is used to establish the method of forecasting model, and the confidence interval of forecast error is made, which determine the range of the maintenance cost control. To improve the accuracy of the forecast, the historical data is used for causal clustering of the factor variables. Then, principal component analysis of the variables is conducted in each type of variable, and the dimension of the variable set is reduced. The neural network algorithm is used to construct the forecasting model and the error range is determined by the error property.