A Vehicle Emission Factors Concentration Inversion Method Based on Deep Neural Networks

In this paper, a vehicle emission factors concentration inversion method based on deep neural network(DNN) is performed to estimate the vehicle emission factors carbon monoxide (CO); hydrocarbons (HC) and nitric oxide (NO) volume concentration ratio. The emission data used in this paper includes the remote sensing monitoring vehicle emission data collected by the Hefei Urban Road Network Vehicle Exhaust Emission Monitoring System in Hefei in 2014 and the vehicle environmental inspection data collected by Hefei Environmental Protection Department. We complement part of the important incomplete exhaust emission data with a semi-supervised learning co -training regression method and train the DNN emission factor inversion models to estimate the vehicle emission factor concentration ratio. Finally, the comparative experiments show that the inversion performance of DNN model is better than that based on Support Vector Regression(S VR) or Multiple Linear Regression(MLR). Therefore, the DNN model is more suitable for estimating vehicle emission factors volume concentration ratio in complex environment.