Related Experiment Video
Updated: Jun 28, 2026
09:34
Implementation of Portable Emissions Measurement Systems PEMS for the Real-driving Emissions RDE Regulation in Europe
Published on: December 4, 2016
28.1K
Urban road BC emissions of LDGVs: Machine learning models using OBD/PEMS data
Xin Wang1, Zhaowen Qiu1, Zhen Liu1
1School of Automobile, Chang'an University, Shangyuan Road, Xi'an, 710016, Shaanxi, PR China.
Chemosphere
|September 15, 2024
Summary
Quantifying urban Black Carbon (BC) from light-duty gasoline vehicles (LDGVs) is difficult. This study used PEMS and ML models, finding Random Forest best predicts BC, enabling real-time monitoring for emission reduction.
Area of Science:
- Environmental Science
- Automotive Engineering
- Data Science
Background:
- Urban Black Carbon (BC) emissions from light-duty gasoline vehicles (LDGVs) are a significant air quality concern.
- Accurate quantification of real-world BC emissions from LDGVs remains a challenge.
- Existing methods may not capture the dynamic nature of urban driving conditions.
Purpose of the Study:
- To assess real-world BC emissions from LDGVs using a Portable Emission Measurement System (PEMS).
- To develop and evaluate machine learning (ML) models based on On-Board Diagnostics (OBD) data for predicting BC emissions.
- To identify key engine parameters influencing BC emissions and compare emission factors between different fuel injection technologies.
Main Methods:
- Utilized PEMS to measure BC emissions from five LDGVs during urban driving.
- Developed five ML models, including Random Forest (RF), using OBD data to predict BC emissions.
- Analyzed correlations between BC emissions and engine parameters (speed, load).
- Compared BC emission factors (EFs) for gasoline direct injection (GDI) and port fuel injection (PFI) engines.
Main Results:
- The Random Forest (RF) ML model demonstrated the highest predictive accuracy for BC emissions (R² > 0.6) across all tested LDGVs.
- Strong correlations (R² between 0.5 and 0.9) were found between BC emissions and engine parameters like speed and load.
- China VI compliant LDGVs showed minimal variation in BC emissions across different urban road types.
- Gasoline direct injection (GDI) engines exhibited higher BC EFs (0.141 ± 0.038 mg/km) compared to port fuel injection (PFI) engines (0.114 ± 0.049 mg/km), a 23.7% increase.
Conclusions:
- OBD-based ML models, particularly RF, offer a viable approach for real-time BC emission monitoring in LDGVs.
- Integrating these models can support the development of effective emission reduction strategies.
- Engine parameters significantly influence BC emissions, and GDI technology leads to higher emissions than PFI.
