Short-term wind power prediction with signal decomposition

Wind power is widely used to replace conventional power plant and reduce carbon emission. However, the variability and intermittency of wind makes the wind power output uncertain, which will bring great challenges to the electricity dispatch and the system reliability. So it is very important to predict the wind power generation. Two different signal decomposition methods are introduced into the prediction of wind power generation in this paper. One is wavelet transform (WT), and another is empirical mode decomposition (EMD). Both of them are good at decreasing the non-stationary behavior of the signal. ANN with the capacity of nonlinear mapping is used to model the decomposed time series. The prediction models WT-ANN and EMD-ANN are compared each other and a combined model based on them is tested. The wind power data from the Saihanba wind farm of China is used for this study.