Abstract: A new short-term wind speed forecasting model based on rational-dilation wavelet transforms and
support vector machine (SVM) is presented. First, an idea of improving accuracy by extraction of the
oscillatory features based on wavelet transforms is discussed. Then, this paper analyzes the superiority of
rational-dilation wavelets to traditional wavelets in the power of time-frequency localization and oscillatory
feature extraction. Last, we present a construction procedure of this forecasting model. Our experimental
results show that the model has a better forecasting accuracy than all those of neural network, SVM, and the
models based on traditional wavelet transforms and SVM.