Solar irradiance prediction method based on TCN-TRANS-ATTENION-GRU

The invention discloses a solar irradiance prediction method based on TCN-TRANS-ATTENION-GRU, and belongs to the field of deep learning prediction. The method comprises the steps of obtaining historical solar irradiance data and other meteorological data, and performing data preprocessing on time series data; according to the method, a two-channel feature extraction network is constructed, pre-processed feature data are respectively sent to different feature extraction networks, the feature data are extracted through the different feature extraction networks, then superposition is carried out, and the feature data can be extracted into deeper-level features; a channel attention network is constructed, different weights are distributed, and important features are highlighted; constructing a GRU network, and extracting time correlation of multi-sequence time sequence data; and constructing a feature output network, inputting a feature vector obtained from the GRU network into a full connection layer, and obtain