Combined prediction method and system for carbon emission of multi-region thermal power plant

The invention provides a combined prediction method and system for carbon emission of a multi-region thermal power plant. The method comprises the following steps: acquiring carbon emission data of a plurality of regions, and performing data decomposition processing on the carbon emission data by using adaptive singular spectrum analysis to obtain carbon emission trend component data of the plurality of regions; carrying out regional space correlation processing on the multi-region carbon emission trend component data, and constructing adaptive selection based on an NAS strategy and a virtual dynamic graph neural network to realize optimal spatial characteristics so as to obtain dynamic carbon emission spatial characteristics of a plurality of regions; performing time correlation processing on the carbon emission spatial features by using a dual-path enhanced Transform based on an attention module to obtain carbon emission time features corresponding to a plurality of regions; and simultaneously acquiring car