Wind power prediction gale data enhancement method considering extreme gale weather

Abstract

The invention discloses a gale data enhancement method considering extreme gale weather. The method comprises the following steps: acquiring historical actually measured wind speed and power data, and generating original wind speed-power data samples; performing local abnormal point identification on the original wind speed-power data samples according to the wind speed data, and removing outlier samples to obtain gale data initial samples; clustering the gale data initial samples by using an F-DBSCAN clustering algorithm to obtain core point samples of each cluster; calculating a cluster density according to core points of clusters, and synthesizing new gale data samples by using an oversampling algorithm based on the core points and the cluster density of the clusters where the core points are located; expanding the original wind speed-power data samples by using the synthesized new gale data samples to serve as training samples of a wind power prediction model. The invention can be used for effectively supplementing unbalanced samples of model training, so that the reliability of a wind power prediction model can be improved, and the wind power prediction precision can be improved.

Description

technical field [0001] The invention relates to the technical field of wind power forecasting, in particular to a strong wind data enhancement method for wind power forecasting considering extreme windy weather. Background technique [0002] Wind power prediction usually adopts physical methods or statistical methods. Physical methods consider NWP data, wind farm terrain and climate factors to predict wind power; statistical methods usually use historical wind speed and historical power to predict wind power. With the development of machine learning and deep learning, wind power prediction methods based on machine learning and deep learning, such as using extreme learning machines, support vector machines, and least squares vector machines; Some progress and breakthroughs have been made in wind power forecasting. [0003] However, most of the traditional wind power forecasting does not consider special treatment for wind power forecasting under extreme weather conditions. ...

AI Technical Summary

Problems solved by technology

[0003] However, most of the traditional wind power forecasts do not consider special treatment for wind power forecasts under extreme weather conditions.

A few papers mentioned that, for example, it is possible to find a training set with a high similarity to the prediction set based on Euclidean distance, entropy, correlation coefficient, etc.; to find out data similar to wind speed and weather patterns according to the clustering method as a training set Make predictions; select similar day data as the training set for prediction according to the model orientation, etc., but these methods not only involve a large number of similarity calculations, resulting in high system overhead, but also do not fundamentally solve the error caused by the scarcity of samples due to extreme windy weather Therefore, in the field of wind power forecasting, wind power forecasting under extreme windy weather is an urgent problem to be solved.

Images