A park carbon emission prediction method and system based on a graph neural network

Abstract

The application relates to a park carbon emission prediction method and system based on a graph neural network, and steps include: acquiring and constructing a park topology data set; a GraphRNN model is built, the GraphRNN model is trained using the topology data set, and a graph structure generation model is obtained; carbon emission time series data of each infrastructure in the park is acquired, and a carbon emission data set is constructed; an STGCN model is built, a park graph structure adjacency matrix is generated based on the graph structure generation model, time series data in the carbon emission data set is converted into a feature matrix, the STGCN model is trained, and a carbon emission prediction model is obtained; real-time carbon emission of park infrastructure is input into the carbon emission prediction model, and carbon emission prediction of each infrastructure in the park is obtained. Compared with the prior art, the application can construct a structured network graph consistent with a real park, effectively consider the space-time relationship between park infrastructures, and improve the prediction accuracy.

Images