Wind power prediction method based on spatial-temporal feature extraction and secondary decomposition aggregation

The invention provides a wind power prediction method based on spatial-temporal feature extraction and secondary decomposition aggregation. The method comprises the following steps: establishing a graph network model according to the correlation between feature data and latitude and longitude information of a wind power plant cluster; selecting an optimal feature subset of the space-time correlation features of the recombined graph network model by using a dual feature selection method; sequentially utilizing a variational mode decomposition algorithm and a group decomposition algorithm to decompose the historical wind power data twice to obtain component data; measuring the component data by using a multi-scale permutation entropy algorithm to obtain a fusion subset; and reconstructing a future predicted value obtained by predicting the fusion subset by using a parallel Stacking model so as to realize power prediction of the wind power plant cluster. The invention provides a decomposition and aggregation met