Measuring the route topography impact on real driving emissions based on neural network models

Area of Science:

• Environmental Science
• Automotive Engineering
• Data Science

Background:

• Real Driving Emissions (RDE) tests are crucial for evaluating vehicle emissions under real-world conditions.
• Factors like topography, driver behavior, and traffic influence RDE, but their individual impacts are complex and coupled.
• Hilly regions present unique challenges for accurately assessing vehicle emissions due to varied route topography.

Purpose of the Study:

• To quantify the importance of route topography as a test boundary in Real Driving Emissions (RDE) tests.
• To investigate the influence of route topography on pollutant emissions in hilly urban environments.
• To assess the adequacy of current RDE regulations concerning topographical variations.

Main Methods:

• Utilized tens of thousands of data window samples from RDE tests conducted in Chongqing.
• Applied factor analysis for data dimensionality reduction and elimination of information overlap.
• Developed neural network models to predict pollutant emissions and calculate the relative importance of input variables, including route topography.

Main Results:

• Route topography was found to be a significant factor influencing on-road emissions, comparable in importance to trip dynamics.
• Neural network predictor importance algorithms highlighted the substantial role of route topography.
• Analysis revealed that the impact of route topography on emissions is not fully recognized in existing RDE regulations.

Conclusions:

• Route topography is a critical determinant of on-road emissions, on par with trip dynamics.
• Current RDE regulations may not adequately account for the influence of varied topography, particularly in mountainous urban areas.
• Mountain cities face challenges with vehicle emissions not being effectively controlled due to the underappreciation of route topography's impact.