Short-Term Load Forecasting Based on Rough Set and Wavelet Neural Network

This paper aims for developing a method, based on rough set (RS) reduction and wavelet neural network (WNN), to improve the efficiency of short-term load forecasting (STLF). The RS reduction could erase redundant characters and this makes it possible to take many influential factors of power load into account, although the learning ability of neural network is limited. Furthermore, WNN is brought forward to improve the efficiency of neural network in load forecasting. The mid-layer of WNN adopts Morlet wavelet and this makes it has a good convergence rate than traditional radial basis function (RBF) neural network. At the same time, because of spending less time in simulating the historical training data, the generalization capability of WNN is of good performance, too. The experimental results show that the method developed in this paper suits STLF and the forecasting result is satisfactory.