A Method of Predicting NO₂ Hourly Concentrations Near City Arteries Based on BP Neural Network

In this paper, a method based on BP neural network is proposed to predict the hourly average concentrations of NO₂ on roadside near city arteries. The main factors that affect the hourly concentrations of NO₂ include O₃ concentrations, NO concentrations, traffic volumes and meteorological factors which contain atmospheric temperature, atmospheric pressure, wind speed, wind direction, relative humidity, rainfall and solar radiation. The hourly average data of these affecting factors were selected as the input neurons of the network. And the network was trained and validated with the data form an automatic monitoring experiment on roadside. As a comparison, a multiple linear regression model was established. When predicting NO₂ concentrations with the neural network approach, the mean absolute value of relative error of training and validation result was 7.19%, and the correlation coefficient between the predicted and monitored concentrations was 0.99, which was much better than the prediction with a multiple linear regression model.