Scholarlink

<p>Air pollution has become a significant global environmental concern, particularly in the field of transportation where the high proportion of nitrogen oxide (NOx) emissions from heavy-duty diesel vehicles (HDDV) is a main contributor to the deterioration of air quality. This study combines vehicle remote monitoring data with deep learning technology and proposes a long short-term memory (LSTM) model based on on-board diagnostics (OBD) data for real-time prediction of NOx emissions from HDDV. The research considers various window widths, model structures, and multiple influencing factors, including a comprehensive analysis of vehicle engine parameters and operational characteristics, demonstrating excellent performance. The mean absolute error (MAE) is 3.963 ppm, root mean squared error (RMSE) of the model is 7.927 ppm, mean squared error (MSE) is 62.841 ppm<sup>2</sup>, and the coefficient of determination (<i>R</i><sup>2</sup>) is 0.997. Comparisons with random forest method reveal the superior performance and scalability of this model. Predicting NOx emissions from HDDV contributes to make up for the missing key item of SCR sensor downstream NOx concentration in OBD data. This study lays the groundwork for future identification and prediction of high NOx emissions to address and improve urban traffic conditions and air quality.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Real-Time Prediction and Control of NOx Emissions in Heavy-Duty Diesel Vehicles Using OBD Data and Deep Learning Models

  • 期刊论文

  • 发表日期：

    2025年7月4日

查看全文
  • Tao Li,
  • Xin Lou,
  • Baoli Gong,
  • Guoquan Xie,
  • Kui Wang,
  • Honghao Zhang,
  • Yong Peng

摘要

Air pollution has become a significant global environmental concern, particularly in the field of transportation where the high proportion of nitrogen oxide (NOx) emissions from heavy-duty diesel vehicles (HDDV) is a main contributor to the deterioration of air quality. This study combines vehicle remote monitoring data with deep learning technology and proposes a long short-term memory (LSTM) model based on on-board diagnostics (OBD) data for real-time prediction of NOx emissions from HDDV. The research considers various window widths, model structures, and multiple influencing factors, including a comprehensive analysis of vehicle engine parameters and operational characteristics, demonstrating excellent performance. The mean absolute error (MAE) is 3.963 ppm, root mean squared error (RMSE) of the model is 7.927 ppm, mean squared error (MSE) is 62.841 ppm2, and the coefficient of determination (R2) is 0.997. Comparisons with random forest method reveal the superior performance and scalability of this model. Predicting NOx emissions from HDDV contributes to make up for the missing key item of SCR sensor downstream NOx concentration in OBD data. This study lays the groundwork for future identification and prediction of high NOx emissions to address and improve urban traffic conditions and air quality.