Diffusion Model Of Neural Network-based Street Canyon No_x

As the number of city motor vehicle increase and that of building intensive degree, the pollution on the street canyon becomes more and more serious.So, the study on the model of pollution distribution is an important work. In previous study, most of the air quality modeling work has been so far oriented towards deterministic simulations of ambient pollutant concentrations, which is based on the use of one selected model. Finally, the model parameters had been modified and met the requirements in the local area. But the importance of various internal model parameters must be taken into account in different city or country and their values are always different and they can't give satisfactory short-term predictions. The algorithms of neural networks are accurate and the structure is simple. It provides a new method to the street canyon pollution prediction.In this paper, the attribute database was collected based on the pollutant concentration data measured in the Zhongsan Road of Chongqing. The database concludes some important parameters: temperature, pressure, the number of the vehicle, the street-level wind speed, the emission of the pollutant from the traffic in the street, the angle of the roof-level wind with respect to the street and so on. Correlation analysis on pollutants concentration of the canyon street with the utilization of the sample database had been completed and build the model.Firstly, OSPM(Operation Street Pollution Model)as a mathematical simulation of Vehicle Pollutants Dispersion in Urban Street Canyon was studied. The model can calculate diffuse concentration from vehicle exhaust and predict air quality. Monitoring data of NOx concentration of street canyon along the main road in Chongqing was used in this paper to evaluate OSPM. The linear correlation coefficient between the simulation values and the expected values was 0.775 95. After the modification of the conversion factor of wind speed,the model presented a better precision (R=0.862 58). The amended model was fit for simulating the diffuse concentration on vehicle pollutants dispersion in urban street canyon in Chongqing,and met the requirements of air quality assessment to a certain extent. By the analysis of influencing factors, some practicable methods were proposed which would good for controlling and improving air quality of Urban Street Canyon.In this paper, after studding the network algorithm and architecture, the Newton method of artificial neural networks was used to simulate the NO_X. The three-layer back propagation network architecture was composed of three parts- input layer,hidden layer and output layer. The input layer is composed of he street-level wind speed,the emission of the pollutant from the traffic in the street,the angle of the roof-level wind with respect to the street。The output layer is composed of two neurons which represent the NO_X concentration on the leeward side and the windward side respectively. The simulated result and correlation coefficient test show that the ratio of the test set error to the training set error is 1.11. The R-values of training sample is around 0.93433, while that of the test sample is 0.87015, which is higher than the R-values of 0.01 and 0.05 significant level and the model has high generalization capacity.Meanwhile, the network architecture of traffic pollution controlling was building with the soft MATLAB 7.5 neural networks tools box and the simulation system was developed with the soft Visure Foxpro9.0.The geographic data were collected and the special database and GIS system were established. Artificial neural networks and the GIS technology were used to build the Visualization technology on traffic pollution controlling. The result showed that Visualization technology on traffic pollution controlling based on artificial neural networks could well simulate the pollutants and showed the concentration distribution.