Short-Term Wind Power Prediction Based on DBSCAN Clustering and Support Vector Machine Regression

Wind power forecasting (WPF) is of great significance for guiding power grid dispatching and wind farm production planning. The intermittency and fluctuation of wind tend to cause the diversity of training samples, which has a great influence on the prediction accuracy. In this paper, a novel short-term wind power prediction data mining method, called DBSCAN-SVR, is proposed to solve the dynamic problem of training samples and improve prediction accuracy. Different to the traditional algorithms, the new algorithm innovatively combines the advantages of DBSCAN clustering analysis and support vector machine regression (SVR). First, according to the similarity of historical days, the training samples are classified by DBSCAN clustering method. In the process, Principal Component Analysis (PCA) algorithm is used to reduce the dimension of training samples, which can improve the performance of the traditional DBSCAN clustering method. Second, the SVR algorithm is used to solve the problems of overfitting and local optimization of traditional networks. The performance of DBSCAN-SVR was evaluated through actual wind power data records. The experimental results demonstrate that the proposed method has a better performance than the traditional methods.