Short-term Power Prediction Based on Improved NWP Data Considering Wind Speed Distribution Inner Wind Farm

The wind farm has only one NWP data point, which cannot accurately reflect the wind resources inside the wind farm. To solve the problem of a poor match between short-term wind power prediction numerical weather forecast and wind farm measured weather. Put forward a kind of based on clustering and temporal characteristics of the modified liters of scale wind power prediction method, first of all, using EOF orthogonal decomposition method to extract the features of each fan, using hierarchical analysis method of the characteristics of the data matrix clustering, and then, each cluster sample selected representative of a fan, using the error transfer characteristic of the NWP data sequence is modified. Finally, The CNN prediction model is used for power prediction, and the weight is determined according to the cluster sample size. A wind farm in northern Hebei, China, is used to verify the effectiveness of the proposed method.