The invention provides a dispersion distance-integrated parallelized wind power plant output power prediction method. The method of the invention belongs to the wind power technical field. According to the dispersion distance-integrated parallelized wind power plant output power prediction method, by means of the clustering of existing data, the historical data of wind power are clustered into a plurality of groups according to features, and the clustered data groups are stored on a plurality of computers; in a wind power prediction process, the computers perform prediction simultaneously andintegrate prediction results on the basis of the similarities of the data features of a time point to be predicted and the features of the historical data so as to realize more accurate and fast windpower output prediction.