Wind power plant ultra-short-term power prediction method based on independent sparse stacked auto-encoders

The invention discloses a wind power plant ultra-short-term power prediction method based on independent sparse stacked auto-encoders. The method comprises the following steps: collecting historical power observation data sets of multiple wind power plants at multiple moments; carrying out dimensionality reduction on the spatial multi-dimensional wind power data set to obtain dimensionality-reduced historical feature data, combining thedimensionality-reduced historical feature data and meteorological information to jointly serve as influence factors of power prediction, and further carrying out dimensionality reduction on the influence factors of power prediction to obtain low-dimensional features; selecting independent features from the low-dimensional features, and performing prediction by using a neural network with dynamic mapping capability to obtain low-dimensional prediction data of the power of the wind power plant; and decoding and reconstructing the low-dimensional prediction data to obtain a power