Study On Prediction Method Of Nitrogen Dioxide Concentration In Chengdu-Chongqing Economic Zone Based On Time Series

NO₂ is a kind of important polluted gas in the atmosphere and has been designated as a“standard pollutant”by the Environmental Protection Agency（EPA）.It is also an important component of smog.At present,the data for studying the spatial and temporal distribution of NO₂ concentration are mainly derived from ground-based and satellite.Since the ground-based needs to draw support from ground stations,and the monitoring scope is limited.Therefore,the full coverage by satellite,all-weather monitoring and analysis of the spatial and temporal distribution becomes a mainstream method.In recent years,people has been paying attention to air quality while demanding to improve their living standards.In order to control and prevent large-area pollution,China began to forecast air quality,and NO₂ concentration is an important indicator of monitoring.There are two main ways to study future concentrations of pollutant,which are simulating by establishing physicochemical correlation models and predicting historical data using statistical models or neural network models.At present,physicochemical models are used more to predict the future concentration of NO₂,while statistical models and neural network models are used less.Therefore,it is very important to try these two methods.In order to analyze the NO₂ air pollution situation in the Chengdu-Chongqing Economic Zone and predict the future atmospheric NO₂ concentration,this paper mainly has done two aspects of work:（1）This paper studies the temporal and spatial variation trends of NO₂ column concentration in the11-year（2008-2018）of the Chengdu-Chongqing Economic Zone.The Chengdu-Chongqing Economic Zone showed obvious seasonal characteristics,with the most serious pollution in winter,the least pollution in summer and slightly higher pollution in autumn than spring.From 2008 to 2012,the concentration of NO₂increased year by year,and from 2013 to 2018,the concentration of NO₂ decreased year by year.In 2013,the annual mean concentration of NO₂ was the highest,which was 485.01×10¹³molec/cm².Chengdu was the most polluted city in the Chengdu-Chongqing Economic Zone.The annual average value of NO₂column concentration in 11 years was 858.07×10¹³molec/cm².Ya’an was the least polluted city in the Chengdu-Chongqing Economic Zone.The annual average value of NO₂ column concentration in 11 years was 243.41×10¹³molec/cm².All these indicate that the concentration of NO₂ in the regions with developed industries is higher than that in the regions with non-dominant industries.From the multi-year annual average distribution of NO₂ and the distribution of population density,the pollution in areas with high population density is higher than that in sparsely populated areas.（2）In order to predict the monthly average NO₂ concentration in the atmosphere in the next year,the historical NO₂ monthly average data is used to predictive learning using the multiplicative seasonality model（SARIMA）in statistical model and the radial basis neural network model.For the problem of small amount of data and the inability to predict the monthly average concentration of NO₂ in multiple steps,the data are firstly grouped by simplified model in this paper,and then predicted by the radial basis neural network model,and finally,based on the grouping,the data with obvious accidental factors were summed and averaged,and then hen the data are predicted by the radial basis neural network model.In Chengdu-Chongqing Economic Zone,by using SARIMA model and grouping radial basis neural network model and summation average radial basis neural network model,this paper calculate that the root mean square error of predicted and actual values were 41.69,43,32.38 respectively,and the correlation coefficients were 0.88,0.90 and 0.95 respectively.The experimental results show that the summation average radial basis neural network has the best prediction effect.