Short-term Prediction Of Wind Power

Wind power prediction is important to the operation of power system with comparatively large mount of wind power. With more and more research taken in the field, several remarkable wind power prediction systems are utilized abroad. However, satisfactory improvements have not yet been proposed by domestic researchers.This paper gives a study on popular prediction methods and an analysis on wind speed data, and then an SVM-based wind power short-term prediction method is proposed, which is able to arrive at a prediction precision equal to or better than the best results of domestic ARMA- and ANN-based methods. More precise prediction models can be trained out with the information of inertia and daily periodicity of wind speed serial added to feature vector. In addition, the precision of 24 to 72 hours ahead prediction can be improved with weather forecast messages added to feature vector.This paper also gives an clustering analysis on daily wind speed serials of different days. The result shows that the clustering of the serials has something to do with weather forecast. The results of clustering analysis can be implemented in wind speed prediction after they are merged into several classic algorithms.Finally, a web system for wind power prediction is developed based on the result of this paper.