A Study On Application Of Neural Network Method In City Air Quality Monitoring

Air is closely related to humans,air quality directly affect people's health and quality of life,China's air pollution emissions total quantity had increased once.In order to strengthen prevention and control of air pollution and reduce air pollution on human health and environmental risks,former state environmental protection administration focused on the establishment of urban air quality daily and air quality forecast system.The air quality evaluation is a very important link,and a variety of evaluation methods came into being.API methods widely used especially but there are some defects:(1)To highlight the role of a single pollutant and not a variety of pollutants into the air as a synergy of quality and grade of the main factors, Therefore,physical meaning is the lack of clear;(2) One-sidedness of the area;(3) Evaluation criteria did not keep pace with the country to change the subject,and so on.There are a variety of air pollutants in the environment,in which each of the pollutants in the atmosphere of a certain quality of the environment damage.When the accumulation of pollutants to a certain extent.That is,when its concentration reaches a certain level,the atmosphere will inevitably cause serious harm to the quality of the environment.L.D's research has shown that James D:With a special concentration of pollution,its air quality also the beginning of the damage caused by the slow increase. The concentration of pollutants reached the critical level,its air quality caused by the loss of much more strongly,when its concentration increases to a certain value,and damage by a new flat.The article adopted the concept of pollution damage as the rate of evaluation of the theory,very easy to understand,and the expansion of the application.The article applied to improve the air pollution damage evaluation method of Changchun city air quality first,enhanced the evaluation of the physical meaning:1.Air pollution damage rate R_i=1/(1+63.93e^-0.3401Xi);2.The new range of air pollution damage rate; 3.From using the evaluation process,we took the monitor data into it and got the results of the annual evaluation,And the actual results were compared.The results showed that there were lots of consistent ones between the method of atmospheric pollution damage rate and API evaluation.Research and analysis shows that the method of atmospheric pollution damage is not only achieving air quality assessment but also being better than the API evaluation method in some ways.Then,from air pollution damage rate advantages,to strengths and weaknesses of the API method we proposed the model which based on the rate of air pollution damage in air quality daily API method and air pollution damage rate method in air quality daily and introduced the concept of secondary pollutants.This made "daily" state of air quality statements was more accurate and objective and People on the secondary pollutants can be informed,it Is helpful to people's environmental consciousness increase.This kind of attempt is both retained the API simple and intuitive method of advantages and comprehensive expressesed on a variety of pollutants results of the damage to the urban air quality combined action really.At the same time,this method is not limited on the type of pollutants and the number of restrictions,As long as there are pollutants and their monitoring data,we can develop and establish the evaluation model,and thus is of great versatility and universality.China attaches great importance to environmental protection.Economic development Is turned "both FAST and GOOD"model to "both GOOD and FAST" model.Therefore,the future of the department of environmental protection monitoring capacity and the scope to monitor will certainly step up.This will have new pollutants into the air quality monitoring.At that time,as long as the rate of pollution damage to the R and the scope of the evaluation procedured to add amendments would be easier to deal with the new rapid evaluation missions.OIFELman(OUTPUT-INPUT FEEDBACK ELman) that the output-input ELman feedback artificial neural network is a network of scholars ELman on the basis proposed by an improved neural network model,in addition to the model of the input layer,hidden layer and output layer units,there is also a special liaison unit,the unit was linked to memory before the moment of hidden unit output value.ELman regression neural network is a typical dynamic neural network,with the characteristics of the dynamic mapping function,in particular,are highly non-linear mapping capability.It can be their own learning and memory of all input and output volume of the relationship between the characteristics of the factors that can be avoided with the goal judge to describe the complex relationship,particularly the expression of the formula,which automatically close to the best portrait of the law of the sample data function,Regardless of how they function with what form.And if we consider the performance of the system's function in the form of more complex neural network that features the more obvious that the air quality forecast is the conventional way to the characteristics of the factors(meteorological factors:wind direction,temperature, pressure,etc.) to select the individual and The custom application to function within the forecast,and OIFELMAN-based network of air quality forecasting is a new neural network in the field to try.This article in using of OIFElman neural network forecasted a wide range of pollutants concentration.This is a good exploration for the actual work.It applied OIFElman neural network model to realize the value of the concentration of pollutants in the next day's forecast,To the atmosphere of the three major pollutants:PM10 (particulate matter),SO₂ and NO₂ subject to monitoring,we selected the city of Changchun on January 1,2007 to December 31,2007 for 365 days to monitor the value of the sample data and OIFELman network was used in air quality monitoring value of the forecast and carried on the contrast with the actual result.As the OIFELman network model to increase the output level of feedback node,thus increasing the neurons at all levels of feedback and improve the network's signal processing capabilities,to enhance the reliability of the forecasts.The results show that the fitting effect is very good.OIFELman network is having excellent performance forecast,with the increase in the number of iterations,the superiority is becoming the more obvious.At the same time that it is precisely because OIF ELman network model to increase the output level of feedback node,the increase in the neurons at all levels of feedback and improve the network's signal processing capabilities,thereby enhancing the reliability of the forecasts.Atmospheric concentrations of pollutants in the value of the forecast on OIFElman advantage of the network model,unlike other air qualityforecasting methods can not be separated as weather forecasting,that is,it does not need the input of various weather factors.Even so,a few days ago due to the input node is the value of the pollution monitoring,and the size of its value on the day of meteorological factors must exist in a variety of contacts and laws,and OIFElman network based on the recent weakness of the value of monitoring to predict that the use of these links, Which is not separated from the concentration of pollutants in value by meteorological factors,in essence,just cleverly bypassing the meteorological factors,weather forecasting is based on a variety of recent has been monitoring the weather factor to the trend of the situation and to use their experience and calculation that,Air quality forecasts can be based on the recent value of the monitoring samples,to adopt the appropriate tools to predict.Therefore OIFElman network model so that the calculation becomes simple, weather factors completely hidden in the mathematical expressions within.As the OIFElman network model with artificial neural network has a self-learning, and self-organizing adaptive characteristics of capacity,Therefore uses in the air pollution forecast domain is having the feasibility,the validity and the accuracy.Still there is a very good prospect of application and the application of a certain value.