Predict the effect of meteorological factors on haze using BP neural network

Abstract

Rapid urbanization and economic growth in China have resulted in severe haze. PM2.5 is a key component of haze. Using machine learning to predict PM2.5 concentrations has become a research hotspot. There are few previous studies on the correlation between PM2.5 and meteorological factors, and the accuracy of other detection results is not high. This study investigates the correlation between six meteorological factors (temperature, dew point, humidity, barometric pressure, wind speed, and visibility) and hourly PM2.5 concentration in Beijing. Pearson correlation analysis was used to analyze and found that meteorological factors had a stable non-linear effect on the change in PM2.5 concentration. Additionally, a BP neural network model was employed to examine the influence of meteorological conditions on hazy weather, considering both univariate and multifactorial aspects. Experimental results reveal a strong correlation between PM2.5 concentration and the four seasons, exhibiting minimal deviation and varying prediction errors throughout the seasons, confirming the seasonal association with haze weather. Consequently, the utilization of the BP neural network model for air quality index prediction is considered feasible. This research serves as a valuable reference for understanding the formation mechanism, implementing control methods, and establishing a haze prediction system to effectively manage urban air quality.