LS-SVM Short-Term Wind Speed Forecasting Based on Wavelet Decomposition Within the Bayesian Evidence Inference Framework

In view of the quasi-periodic,non-stationary and non-linear features of wind speed,the original wind speed sequence is decomposed into a series of sub-sequences based on the multiresolution analysis feature of wavelet.For each of these sub-sequences,a different forecasting model is established.The optimal parameters of every model can be found through the three-layer Bayesian evidence inference and they are used to establish the least squares support vector machine(LS-SVM) short-term wind speed forecasting model based on the wavelet decomposition and the Bayesian evidence inference framework.When the proposed method was applied in the one-hour-ahead wind speed prediction in a wind farm in the northeast region,the mean average percentage error of the predicted wind speed was only 7.63%,a large improvement of the prediction precision.The results verify the effectiveness of the proposed method.