Using wavelet transform to improve generalization ability of neural network in next day load curve forecasting

The net day load curve forecasting plays an important role for electric power system operation. Because of affecting by many factors, daily curve is composed by many regular wave trends and stochastic ones. This makes the poor efficiency and generalization capacity of neural network adopted in forecasting. By using discrete wavelet transform, the complicated load curve could be extracted to many simplex ones. After abnegating stochastic series, other extracting results are simulated by radial basis function (RBF) neural networks. Adding the forecasting results of neural network together, it will get the forecasting load. The tests show that the models brought forward in this paper is feasible.