Link-based Traffic Volume Forecasting for Dynamic Emission Estimation Based on Pattern Clustering and Recognition

In China, air pollution has become serious and frequent in many megacities, which puts an urgent need for real-time air quality modeling. Concentrated emissions from industry like power plants and steel mills can be directly monitored on-line, but for road transportation, as a major source of pollution, the dynamic emission cannot be obtained from each vehicle, which became the bottleneck in air quality modeling. Dynamic traffic volume is crucial for the estimation of dynamic emissions. However, limited by traffic monitoring cost and coverage, it is a considerable challenge to obtain wide-ranged and real-time traffic volume by observation or current forecasting methods. Therefore, by analyzing currently available data sources including the traffic survey, floating car data, and the average daily traffic (ADT), a method based on traffic flow pattern clustering is proposed to predict dynamic traffic volume for each link. Firstly, by mining sample traffic volume data, the daily traffic flow presents characters of repeatability and finiteness, then using the K-means clustering, a library of traffic patterns on different road and date types are obtained. Secondly, the normalized fundamental diagram is proposed for traffic volume pattern recognition, then the link-based dynamic traffic volume can be predicted by combining the traffic volume pattern with the ADT of the link. Comparing the predicted traffic volume by proposed method and the speed inversion using fundamental diagram with the field data, the average root mean square error and the mean relative error using proposed method are 43.0% and 82.6% lower than the speed inversion respectively.

• Record URL:
  https://annualmeeting.mytrb.org/OnlineProgram/Details/15667
• Supplemental Notes:
  • This paper was sponsored by TRB committee AMS10 Standing Committee on Air Quality and Greenhouse Gas Mitigation.
• Corporate Authors:

  Transportation Research Board

  ,    
• Authors:
  • Ge, Mengxue
  • Song, Guohua
  • Zang, Jinrui
  • Wu, Yizheng
  • Yu, Lei
• Conference:
  • Transportation Research Board 100th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2021-1-5 to 2021-1-29
• Date: 2021

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: 20p

Subject/Index Terms

• TRT Terms: Air quality; Average daily traffic; Cluster analysis; Forecasting; Network links; Pollutants; Traffic volume
• Geographic Terms: China
• Subject Areas: Environment; Highways; Operations and Traffic Management; Planning and Forecasting;

Filing Info

• Accession Number: 01763916
• Record Type: Publication
• Report/Paper Numbers: TRBAM-21-04182
• Files: TRIS, TRB, ATRI
• Created Date: Feb 4 2021 10:57AM