The Analysis And Forecast On The Changes Of PM2.5 Concentration Value In Zhengzhou

The rapid deterioration of air quality is currently being the fact that we have to pay attention to,thus directly leading to the low visibility of the air and the low efficiency of human production and daily life.China has made great efforts to control air pollution but with limited success.Since the problem of air quality is not a temporary one,to make a better air pollution prevention and prediction,analysis playd a significant role in the factors of air quality's impacts and prediction method of study atmosphere pollution.In this paper,we focus on the evolution of PM2.5 one key pollutant of the smog which is highly issued during people's daily life.Mainly,the data of PM2.5 concentration in Zhengzhou area over a period of time is taken into account in the analysis of its evolution,which is used to quantitatively analyze the future air pollution condition in this area.Based on the spring hourly data of 2015 and the imputed missing data,together with the time series methods,the ARIMA model is given to predict.Only the influence of the time factor on the future PM2.5 concentration is considered.It is found that meteorological factors such as wind direction,wind speed and air temperature do have some certain influence on the diffusion and aggregation of pollutant PM2.5.In addition,the other remaining pollutants in the smog such as PM 10,S02,CO and so on will also influence the distribution.of PM2.5.Therefore,the meteorological factors of Zhengzhou and its surrounding areas are combined together with the above one.The control group and experimental group is established by the method of experimental,and setting four different kinds of models.Then the BP neural network model,after learning,will be used to do integral prediction about the future air condition.By the comparison of the prediction results of the two models above,it is found that the ARIMA model is limitedly suitable for the short-time prediction,and also the prediction effect is not ideal enough.In contrast,because the neural network model contains more variables,the prediction results achieve high-precision,which enables the model effective in a longer time than ARIMA model.Finally,the summary of the results of the study together with the empirical analysis is done,and meantime,the shortcomings of this paper and the method how to improve are pointed out,such as the method of filtering correction which can highly reduce the error of the predicted values.