Air Quality Forecast Based on Neural Network
Download as PDF
DOI: 10.25236/mse.2018.044
Author(s)
Shengping Zhao And Jingrui Li
Corresponding Author
Shengping Zhao
Abstract
There is a wide public concern on air quality. The precise forecast of air quality is of significant importance for production, life and atmospheric control. This study takes the air quality in Yuxi City as the research object. And the Pearson correlation coefficients between 6 conventional meteorological factors, the concentration of 6 pollutants and AQI are analyzed. AQI value is forecasted based on the values of previous n days. Air quality forecast can generally be done through BP network, but there are problems like overfitting. Therefore, some improvement is made in BP network. Base on the analysis of the three years AQI tendency chart, it is proposed in the study to add historical contemporaneous meteorological data to the Input. And it is proved that this addition improves the forecast precision. Besides, the study proposes the ideas about improving excitation function and using cascade network for the further improvement of model forecast performance.
Keywords
Bp Neural Network, Air Quality Forecast, Historical Data, Improved Algorithm