Power futures price forecasting based on improved wavelet neural network

The fluctuation of electricity futures price is affected by various factors and is mostly difficult to forecast exactly. In this paper, a model of improved wavelet neural network (IWNN) is applied to forecast the prices of power futures. Two ways are presented to accelerate local convergence and improve training ability. One is improvement of output layer function, and another is auto-adapted correction of network parameters. Then, principal component analysis (PCA) is utilized to deal with the input data of IWNN. Via PCA, primary components of input variables are extracted and insignificant components are discarded, so as to simplify the network structure as well as enhance the network´s generalization performance. The effectiveness of the proposed methods is verified by simulation and analysis.