Short-term wind speed prediction using support vector regression

This paper presents a new approach to short-term wind speed prediction. The chaotic time series analysis method is used to capture the characteristic of complex wind behavior in which a correlation dimension method is employed to calculate embedding dimension of the time series, then a mutual information method is used to determine the time delay. Based on the embedding dimension and time delay, support vector regression (SVR) is trained to perform the prediction. The proposed method is evaluated using the real-world data collected from a wind farm. The results have demonstrated the accuracy of the proposed wind speed prediction method in comparison with that offered by an artificial neural network (ANN).