Research On Air Quality Prediction Based On Improved BP Neural Network

With the rapid development of global economy and the increasing number of urban population,motor vehicles and industrial development zones,the problem of urban air pollution becomes more and more serious,and the pressure on the atmospheric environment is also increasing.The air quality of a city not only affects the production and life of residents,physical and mental health,but also affects the overall competitiveness of the city.In view of the air pollution problem,if we can dig the law of air quality change from the historical environmental data,we can predict the future air quality situation in advance,and prevent the possible air pollution,provide reliable help for the environmental management department to formulate air pollution prevention and control measures.In addition,air quality prediction can let the public know the future air quality situation in advance,provide a certain reference for their travel,and improve the public's attention to the change of air quality,strengthen the public's awareness of environmental protection.Traditional air quality prediction methods are mainly statistical prediction and numerical prediction.Although the prediction accuracy of these two methods is higher than that of early potential prediction,but they also have some shortcomings.Statistical prediction often needs a large number of long-term monitoring data,while numerical prediction takes a lot of time and is difficult to achieve.Artificial neural network(ANN)is a kind of information processing system which imitates the functional characteristics of human brain.It has strong nonlinear processing ability and good fault tolerance ability.This paper combines the artificial neural network with the prediction of air quality index(AQI),and uses the strong nonlinear mapping ability of artificial neural network to find out the relationship between historical pollutant concentration,meteorological data and future air quality index,establish prediction model,realize the prediction of urban air quality,and improve the shortcomings of the model,further improve the model's training efficiency and prediction accuracy.The main work of this paper is as follows:1.Research the development of artificial neural network in the field of air quality prediction at home and abroad,research the basic principles of artificial neural network and BP neural network,study the related concepts and calculation methods of air quality index(AQI).2.Collect and sort out the air pollutant concentration data and meteorological data from January 2016 to December 2019 in Nanchang,design a prediction model using historical pollutant concentration data and meteorological data to predict the future air quality index,meanwhile,construct training samples and test samples,and process the missing values and special values of samples.3.Select the structure of BP neural network,establish the air quality index prediction model based on BP neural network.After training the model with training samples,predict the air quality index of a few months in Nanchang in 2019,and analyze the prediction effect and shortcomings of BP neural network.4.Study the principle of principal component analysis(PCA)and LM(Levenberg-Marquardt)algorithm.Establish the BP neural network prediction model improved by LM algorithm,establish the BP neural network prediction model improved by PCA and LM algorithm.First of all,in order to explore the difference of training efficiency between the common BP model and the two improved models,compare the training time and the size of the occupied space of each model,and analyze the experimental results.Secondly,in order to explore the difference of prediction accuracy between the common BP model and the two improved models,compare the AQI prediction accuracy of each model in three aspects of month,quarter and year,and analyze the experimental results.Finally,explain the advantages of the two improved models,and give some suggestions for how to further improve the prediction model.The final experimental results show that the two improved BP models have more advantages than the common BP model.Among them,BP neural network model based on PCA-LM algorithm has the fastest training speed,the smallest occupation space,and the best performance in the case of strong interference;while BP neural network model based on LM algorithm has the best prediction effect in most cases,but it is not as good as PCA-LM-BP model in the case of strong interference.At the same time,the experiment also proves that it is feasible and effective to use the air quality index prediction model based on the improved BP neural network to predict the air quality of Nanchang,which provides some ideas and methods for the follow-up air quality prediction research.