A spatiotemporal graph neural network dissolved oxygen prediction method, system, device and medium cooperating with multiple meteorological elements

Abstract

The application discloses a kind of synergic multi-weather element's spatio-temporal graph neural network dissolved oxygen prediction method, comprising: obtaining the multi-source data for the dissolved oxygen prediction of target monitoring section in the preset time range, and pre-processing;Multi-source data is constructed as graph representation for representing spatial correlation;Based on graph representation, multi-scale spatial dependence modeling is carried out on multi-source data, to obtain spatial context;Based on multi-source data and the spatial context, multi-scale time dependence modeling is carried out through time convolution network and time Transformer encoder, to obtain spatio-temporal feature representation;Based on spatio-temporal feature representation, prediction output processing is carried out, to obtain dissolved oxygen concentration prediction result.The application solves the problem that existing data-driven model is difficult to model the multi-scale spatio-temporal dependence of dissolved oxygen concentration affected by regional meteorological environment, and improves the accuracy of short-term and long-term prediction of dissolved oxygen concentration.The method of the application provides technical support for water environment monitoring and early warning of water quality.

Images