Research on NOx Emission Prediction Model for Delivery Trucks Based on Machine Learning and Its Application in Smart Transportation

With the rapid development of smart transport and green emission concepts, accurate monitoring and management of vehicle emissions have become the key to achieving low-carbon transport. This study focuses on NOx emissions from transport trucks, which have a significant impact on the environment, and establishes a predictive model for NOx emissions based on the random forest model using actual operational data collected by the remote monitoring platform.The results show that the NOx prediction using the random forest model has excellent performance, with an average R² of 0.928 and an average MAE of 43.3, demonstrating high accuracy. According to China's National Pollutant Emission Standard, NOx emissions greater than 500 ppm are defined as high emissions. Based on this standard, this paper introduces logistic regression, k-nearest neighbor, support vector machine and random forest model to predict the accuracy of high-emission classification, and the random forest model has the best performance on high-emission classification with an accuracy of 93.7%, effectively identifying vehicles with excessive emissions. In order to gain more insight into the key factors affecting NOx emissions, the study used partial dependency diagrams to analyse the important variables. The results of the study show that SCR outlet temperature, DPF exhaust temperature and urea injection rate have a significant effect on NOx emissions. This study not only provides a theoretical basis for the optimisation of the emission control system, but also provides scientific support for the realisation of intelligent and low-carbon traffic management policy making, which helps the green emission management in the intelligent traffic system.