Short Time Forecast of Wind Speed Based on EMD and SVM

As an important renewable energy, wind power is paid great attention by all countries. It is of great significance to predict wind speed accurately for the power system which includes large amounts of wind power. The wind speed time series appears typical non-stationary, as a result, the outcome obtained from applying single prediction method directly will be unsatisfied. In order to improve the accuracy of wind speed prediction, a model based on empirical mode decomposition (EMD) and support vector machines (SVM) is proposed in the paper. The wind speed time series is made by EMD at first, then appropriate support vector machine model is established with different frequency bands, finally the output value of each model is summed equal right to get the final prediction result. The radial kernel is selected by the SVM, the parameters that necessary are obtained through cross-validation. The actual cases are employed to demonstrate the validity of the proposed approach. The results are compared with those obtained by the single SVM model, which shows that the given model can effectively improve the accuracy of wind speed forecasting.