The Design And Implementation Of Air Quality Prediction And Warning System Based On BP Neural Network

Air pollution has become a problem of great concern to people all over the world.Air pollution in China is more serious than that in other countries.Air pollution has seriously damaged people's lives and health.Therefore,how to establish a scientific,effective and accurate air quality prediction system has become a hot issue.In recent years,our team has been engaged in the research of air quality forecast,and has put forward an air quality forecast model based on artificial neural network(in this paper,it is called old method).Although this method can effectively solve the complex and nonlinear problem of the air quality forecast,and improve the effectiveness and practicality,it also has some deficiencies.In order to deal with them in the old method,this paper presents an improved method for air quality forecast based on BP Neural Network.The system is implemented in Java.The Echarts visualization technology is used to display a large amount of data in a graphical way to observe the trend of pollutant concentration.Then it uses Matlab's neural network toolbox to implement BP neural network and conduct training,prediction and error analysis.Finally,in the early warning,the predicted values are counted according to the state-prescribed warning level,and the WebSocket technology is used to push the warning information to the foreground.The data,from January 2016 to December 2017,collected by Shijiazhuang City Air Quality Monitoring Point was taken as the research object,besides,the air quality data and meteorological data required for the experiments were obtained from the API interfaces published by the third-party website in the paper.In the aspect of prediction,the old method is improved from three aspects: the determination of impact factors,the establishment of prediction model and the determination of experimental data items.Firstly,about impact factors determination,the improved method uses the chi-square check method to replace the data mining method in the old method.Secondly,in the prediction model establishment,the improved method uses the first two months of air quality and meteorological data to predict the air for the next ten days.In the determination of experimental data items,the improved method increased the experimental data items of the old method from 26 items to 37 iterms.Finally,the paper uses the improved method and the old method to predict the air quality levels of Shijiazhuang and analyzes errors of them,from Jan.1st,2017 to Des.31 st,2017.The experimental results show that the air quality prediction model established in this paper has a high prediction accuracy,improves the practicability and effectiveness of the prediction,and can provide more reliable scientific basis for environmental protection departments than the old method.