PM_2.5 Prediction In Taishan District Based On BP Neural Network

In the context of the new era,the concept of harmonious coexistence between human being and nature has been engraved more deeply in people’hearts.As a necessity for human’s survival,air is critical to human whose health is closely related to the air’s safety.However,some changes have been made in response to the environmentally friendly policy upholding low-carbon travel that a growing number of people are choosing new energy vehicles in recent years.Whether it is in a big city or a small town,the discharge of pollutants directly affects the safety and health of local people.The impact of air pollution on life is obvious to us all.A series of tragic car accidents due to greatly shortened visible distance caused by heavy haze have brought unbearable distresses to numerous families.This thesis mainly conducts the following research regarding the prediction over PM_2.5:1.Through the analysis of the data collected from five air quality monitoring points in Taishan District,the samples of daily pollutant data from 2014 to 2020 are obtained,including PM_2.5,PM₁₀,SO₂,NO₂,CO,O₃-8h,and meteorological data of daily average temperature,weather conditions,wind force and direction.The correlation between these data and data of PM_2.5acquired in the next day is analyzed.2.Based on the BP（back propagation）neural network,it analyzes and introduces the structure and realization process of the original BP neural network and the neural network optimized by artificial bee colony algorithm and genetic algorithm.The basic data of Taishan District is used as the data sample.The three types of networks are trained iteratively,and after comparison,it is concluded that the two algorithms have obvious optimization effects on the BP neural network.Among them,the improved genetic algorithm has more advantages than that,and the variance is further reduced to 15.8298.3.A GUI interface is built.Through visual operation,the user can type in the pollutant data and weather conditions of the previous day to obtain the prediction over the PM_2.5of the next day,and the predicted result can be presented through visual interface.