| This paper used air quality monitoring data of Shenyang from 2015 to 2017 and the ground meteorological data for the same period to analyze the temporal variation characteristics of AQI,carry out correlation analysis of AQI and influencing factors in different seasons to further understand the formation reasons of temporal variation characteristics,and on this basis,three models are established to predict AQI in Shenyang of four seasons.This study has the following conclusions:(1)From 2015 to 2017,most AQI daily values in Shenyang fluctuated within the range of 30-300;The monthly variation curve is approximately v-shaped.From January to August,the change trend is downward,and from August to December,the change trend is upward.Pollution is more serious in winter and spring,and less in summer and autumn.The annual average AQI in 2015,2016 and 2017 was level 2,level 3 and level 3.The number of days in good category is the largest in each season.The number of days with level 2 of AQI in each year is the most.(2)Correlation analysis was conducted on the AQI and factors of four seasons,correlation coefficients were calculated,and Monte Carlo test was conducted.The results showed that the influence factors of each season are different.The main meteorological factors in spring are evaporation and temperature.The main meteorological factors in summer are evaporation,surface temperature,humidity and sunshine duration.The main meteorological factors in autumn are evaporation,surface temperature,air pressure,humidity,sunshine duration,air temperature and wind speed.The main meteorological factors in winter are evaporation,surface temperature,humidity,sunshine duration,wind speed and direction.Evaporation and temperature had an effect on AQI in four seasons,and the characteristic factor of AQI in the previous day also had an effect on AQI in the four seasons.On this basis,models are established.(3)Three prediction models of AQI were established and the models were used for prediction.The mean absolute errors of stepwise regression model of four seasons were 22.46,20.21,20.76 and 33.77;the mean relative errors were 26.49%,33.20%,29.00% and30.19%;the root-mean-square errors were 29.51,27.06,27.60 and 46.79;the correct rate oflevel forecast was 58.70%,57.61%,67.03% and 37.78%.The mean absolute errors of BP neural network prediction model of four seasons were 17.06,17.05,18.87 and 30.05;the mean relative errors were 17.52%,29.38%,25.85% and 27.94%;the root-mean-square errors were 23.56,23.87,26.07 and 48.36;the correct rate of level forecast was 71.74%,69.57%,75.00% and 64.44%.The mean absolute errors of genetic algorithm optimized BP neural network prediction model were 15.51,16.95,18.49 and 28.18;the mean relative errors were19.47%,23.10%,23.86% and 27.17%;the root-mean-square errors were 22.49,23.35,24.04 and 44.47;the correct rate of level forecast was 72.83%,72.82%,75.82% and 68.89%.The prediction effect of genetic algorithm optimized BP neural network model in each season is better than that of BP neural network model and stepwise regression model,indicating that the prediction effect of neural network model for Shenyang AQI is better than that of regression model,and the prediction effect of BP neural network optimized by genetic algorithm is better. |
PDF Full Text Request