The Research Of Wind Power Combined Prediction

With the growing maturity of wind power technology, declining costs of wind power, wind power capacity has been increased, but the wind power is low density and the characteristics of random. These characteristics make grid scheduling and running costs of the problems. In order to reasonable scheduling plan and ensure stable operation of power systems, and further increase the power capacity of wind power to accept, we need more accurate prediction for wind power output. The scholars of domestic and foreign has done extensive research for wind power forecast. The results show that different forecasting methods contain different information.Therefore, combination forecasting that effectively use all information can improve the prediction accuracy.At present, research of wind power prediction has some limitations:(1) the accuracy of prediction to be improved; (2) the long-term prediction accuracy is low and lack of effective forecasting method; (3) in the course of the prediction, how to effectively combine various prediction. Future trends in studying new method is effective to combine forecasts, achieve higher precision and longer prediction scale requirements.In this paper, The theory of combination forecasting were analyzed in detail. In chapter two, we analyze ARMA, Gray neural network and Wavelet neural network. Then, this paper propose entropy weight method, cooperative game method, error with absolute value and vector cosine combined forecasting model.Finally, chapter four gives examples to illustrate that the result of single prediction and combination forecasting.