Research On Method Of Air Quality Source Receptor Analysis Based On Neural Network

In the process of economic,social and environmental sustainable development,people's living standards and quality of life have been greatly improved,and it is required to have good air quality corresponding to it,so as to protect human health and ecological health.It is the key of air quality management to study the relationship between source and receptor in the atmosphere and identify the key pollution sources that affect air quality.Because the influencing factors of air quality are dynamic,complex and changeable,and the local air quality(receptor)is affected by the comprehensive effects and coupling effects of geographical factors,meteorological factors and economic factors(source points),its evolution law presents the characteristics of variability,volatility and non-linearity.How to effectively characterize the spatio-temporal dynamic correlation of air quality characteristics,and quantitatively analyze the contribution rate of source points to the air quality of receptors,so as to effectively predict the air quality of receptors is of great significance.In this paper,a modeling and analysis method of air quality source and receptor based on neural network is proposed,which fully considers the correlation between source and receptor in space and the influence of historical concentration in time to analyze and predict the evolution trend of air quality more accurately and stably.Firstly,the factors influencing the air quality of source and receptor were studied.The spatial correlation factor was defined as the meteorological factors and pollutant concentration at each source point,and the temporal correlation factor was defined as the historical pollutant concentration of the receptor.On the basis of considering the temporal and spatial correlation of air quality,a new method to construct the air quality source-receptor model is proposed.Secondly,the spatial and temporal characteristics of source and receptor air quality are analyzed.The influence factors between each source point and the receptor were quantitatively analyzed,and the correlation analysis was carried out by using the method of numerical fitting,and the contribution rate between the source and the receptor was effectively characterized by combining the gray correlation analysis.Thirdly,a prediction model of source-receptor analysis based on neural network is proposed.The air quality of monitoring stations was abstracted as network nodes,BP neural network prediction method was proposed on the basis of spatial correlation,and LSTM prediction method was proposed on the basis of time correlation.The method of source-receptor air quality feature mapping based on neural network is implemented.Finally,the model was trained and adjusted.After fully considering the influence factors of air quality,the model was used to predict the air quality index of the receptors in the next day,and the experimental results were analyzed and compared.