Ultra-short-term wind power prediction using ANN ensemble based on PCA
Dong Lin He, Ruiye Liu · 2012
Using the ANN to forecast wind power need input so many variables that may lower the computation efficiency. Also the ANN's generalization ability is poor. It may cause “over fitting” phenomenon. To solve the first problem this paper propose to apply the principal components analysis (PCA) to reduce the variables. The advantage of PCA is that it can keep most of the information and is objective and versatility. Considering “over fitting”, the essay use boosting net ensemble. The different net ensemble is sensitive to different Weather situations. So that net ensemble has stronger generalization ability to a single ANN. By forecasting the wind power of Kuochang wind plant in Zhejiang province not only checks the correctness of ANN ensemble based on PCA to forecast wind power but also proves it has a better accuracy and will not cause “over fitting” phenomenon.