Wind Speed Prediction Based On Wavelet Packet Analysis And SVM

The ever-disregarded clean, renewable energy has in recent years been out of laboratory and integrated into our life every day. With the green energy race heating up, many new wind energy companies are taking big steps forward in making wind produced electricity more efficient, affordable, and generally more accessible for individuals, energy companies, and governments. According to demands for making decision supports for a wind farm and helping engineer to site selection of wind farms, more and more researchers have been attracted by technology of wind velocity prediction. This dissertation combined with existed experience of prediction to build a prediction model based on processing the original wind signal, wavelet packet analysis and support vector machine, meanwhile, using this model to predict the wind period which we may need in order to enhance the precision of prediction.At first, pre-processing the original signal to let it can be identified by SVM. Then a second process by wavelet packet analysis was needed in order to dividing the signal into two parts of training samples. Therefore, some groups of testing samples can be selected from training samples.Secondly, an air data system was developed based on embedded technology which consisted of touch-screen, communication circuit and respectively independent data gathering circuit, engineers can utilize the system to obtain a high-precision data of wind velocity on the top of wind turbine.At last, a prediction model can be built with samples trained by SVM, to get the precision of prediction by using testing samples.