Research On Prediction Of Surface Water Quality Based On Graph Neural Network

Water quality prediction is an important aspect of surface water pollution prevention and management,which is of great significance to reinforce the protection and utilization of water resources.The existing water quality prediction models based on deep learning mainly focus on extracting the time characteristics from the time series of water quality,but ignore the spatial characteristics of water quality.In view of this,this thesis explores the prediction of surface water quality based on graph neural network.The specific research contents are as follows:（1）For the problem that the existing water quality prediction models cannot take into account the effective extraction of both temporal and spatial features,a surface water quality prediction model GCN-Seq2 Seq is proposed that integrates temporal and spatial features.The spatial features of water quality data are extracted through GCN;the sequence dependencies between continuous time series water quality data are learned through the Seq2 Seq model,temporal features are extracted,and multi-step prediction of water quality data is performed.Experiments were conducted with data from water quality monitoring stations in the Huangshui River and its surrounding areas in Xining City,Qinghai Province.The results show that the proposed model can effectively improve the accuracy of multi-step prediction of surface water quality.（2）For the problem that the existing models are difficult to accurately extract the global spatio temporal features,a surface water quality prediction model GATDASeq2 Seq based on the attention mechanism fusion of spatio temporal features is proposed.The DASeq2 Seq model with two stage attention mechanism to extract the key features and key time points of the historical water quality time series,which further promotes the performance of the prediction model.（3）Design and implement of surface water quality prediction software,realize the main functions of data acquisition,data processing,data prediction.This thesis mainly conducts simulation experiments based on the time series data of surface water quality and the spatial data of relevant surface water quality monitoring stations.The results show that the constructed surface water quality prediction model based on graph neural network performs well and can effectively improve the prediction of surface water quality precision.This research work can provide some reference for surface water quality monitoring and pollution prevention and control.