Research On Regional Air Pollution Forecasting Based On Mobile Pollution Sources’ Telemetry Data

In recent years,with the increase in the number of urban motor vehicles,the ecological and environmental problems caused by vehicle exhaust emissions have become more and more prominent,which have become a hot issue of social concern.A large number of harmful gases,such as carbon monoxide,carbon dioxide,hydrocarbons and nitrogen oxides emitted by mobile pollution sources,such as motor vehicles and engineering equipment driving on the urban road networks,have become the major sources of air pollution.Obtaining real-time vehicle emission information and air pollution in different areas is significant for controlling urban traffic pollution and protecting the ecological environment.From the aspects of atmospheric pollution caused by mobile pollution sources’ emission,two methods based on statistical learning and deep learning models are proposed to predict the air pollution in single station and multi-station area.The main contributions of this paper are as follows:(1)To solve the problem of class imbalance and sparseness of labeled samples in the air pollution data,this paper proposes a semi-supervised extreme learning machine(SS-ELM)forecasting model based on class imbalance correction technique.The pollution prediction method employs majority weighted minority oversampling technique(MWMOTE)to construct the balanced dataset.Then,SS-ELM which utilizes labeled and unlabeled data is trained to predict the air pollution of the roadside.The proposed model is tested on the dataset obtained from Xiasha monitoring stations in Hangzhou.Compared with ELM,SMOTE-ELM and MWMOTE-ELM,the proposed algorithm obtains the highest F-measure and G-mean,and improve the performance of predicting air pollution under the class imbalance scene.(2)In most cities,the monitoring stations are sparse.Also,air pollution is affected by various internal and external factors.It is a challenge to predict air pollution in urban areas.To construct historical emission data with insufficient station records,inverse distance weighted(IDW)space interpolation algorithm is introduced in this paper.Then regional air pollution prediction turns into a time-space sequence prediction problem.Based on the properties of spatial-temporal data,a deep spatial-temporal dense network model is proposed to predict air pollution in each region.Finally,several experiments are conducted on the real dataset of Hangzhou.The results show that combing IDW spatial interpolation and deep spatial-temporal dense network model can effectively predict the air pollution in each area of the city and achieve superior performance compared with ARIMA,CNN,ST-Res Net,CNN-LSTM.(3)In order to monitor the emissions of mobile pollution sources,an online monitoring software platform is built based on the developed telemetry equipment.Furthermore,according to the demand to identify mobile pollution source emission,the module of emission classification is designed and realized based on the proposed algorithm.Finally,we test on the online monitoring software platform and obtain a fine-grained classification of mobile pollution sources.This paper takes mobile pollution source emissions monitoring as the research object,an online monitoring software platform based on the developed telemetry equipment is built.Furthermore,meeting the needs of identifying mobile pollution sources that exceed the standard,the emission level classification module is designed and implemented.Finally,the proposed method in this article is embedded on our monitoring platform and tested,achieving a fine-grained division of the emission levels of mobile pollution sources.