| The problem of pollution that comes with the rapid development of the economy cannot be ignored.The concentration of PM2.5 has also become a concern.In this paper,the hourly PM2.5 concentration data of the 14 urban environmental monitoring stations in Liaoning Province from January 1,2015 to December 31,2017 and the meteorological elements of the corresponding weather stations were used to study Liaoning by statistical analysis.The temporal and spatial variation characteristics of PM2.5 concentration of pollutants were analyzed.The correlation between meteorological elements and PM2.5 concentration values was analyzed by season.On this basis,two models of stepwise regression model and BP artificial neural network model were established respectively.The city's PM2.5 concentration is predicted.The main findings of the study are as follows:(1)The concentration of PM2.5 in Liaoning province decreased from 2015 to 2017,and the concentration in the four seasons was winter>autumn>spring>summer.The average monthly concentration in 2015 is between 30~98?g/m3;the monthly average concentration of PM2.5 is lower from May to September,and the monthly average concentration of PM2.5 is higher from January to April and from November to December,PM2.5 in November.The concentration was the highest(98 ?g/m3),and the PM2.5 concentration was the smallest(30 ?g/m3)in September.(2)The research on PM2.5 concentration in special period of Liaoning Province shows that the concentration of heating period in Liaoning Province is higher than that in non-heating period in 2015,in which the concentration of PM2.5 in the heating period of Yingkou City is 1.3 times that of non-heating period,and the amount of coal burning during heating period.An increase will result in an increase in PM2.5 concentration.During the first year of 2015,the PM2.5 concentration was significantly higher than that of the weekdays.During the 2016 and 2017 Chinese Festival,the PM2.5 concentration was higher than that of the weekday,but it was also lower than the 2015 PM2.5 concentration.(3)According to the correlation analysis between PM2.5concentration and meteorological elements in Liaoning Province,the main influencing factors in spring are: average surface temperature and daily minimum surface temperature;the main summer influencing factors are: average surface temperature,daily minimum surface temperature,and daily maximum surface temperature.Etc.The main influencing factors in autumn are: the pressure of the station,the number of hours of sunshine,the concentration of PM2.5 on the previous day,etc.The main influence factors in winter are: daily minimum surface temperature,average relative humidity,sunshine hours,etc.(4)Establish a stepwise regression analysis model for the four regions of Liaoning province.The average relative error between the predicted and true values of PM2.5 concentration in the four cities was 33.4%,32.5%,32.7% and 31.8% and the average absolute error was 12.4?g/m3,10.6?g/m3,14.6?g/m3 and 20.5?g/m3.The root mean square error was 2.3?g/m3,2.0?g/m3,2.7?g/m3,and 3.9?g/m3.The model established by stepwise regression analysis predicts the concentration of PM2.5 in Liaoning Province in spring and summer,which is better than autumn and winter.The BP neural network models of different cities were established by using MATLAB software.The relative errors of the four cities were 17.2%,14.6%,18.9%,24.9%,and the average absolute error was 7.0?g/m3,4.9?g/m3,6.0?g/m3,11.9?g/m3,RMSE was 2.0?g/m3,1.4?g/m3,1.7?g/m3,and 2.9?g/m3.By comparing the accuracy indexes of these two models in Liaoning Province,the results show that the BP neural network model is better than the stepwise regression model.Therefore,the BP neural network model can improve the accuracy of PM2.5 concentration prediction. |
PDF Full Text Request