Estimation of transport CO2 emissions using machine learning algorithm

This study investigates carbon dioxide emissions from light-duty diesel trucks using a portable emission measurement system (PEMS) and a global positioning system. Two LDDTs are selected for data collection, and a novel CO2 emission model is developed using deep learning techniques, specifically an LSTM architecture. The model is trained on PEMS data to predict CO2 emissions based on various factors such as vehicle specific power, speed, road slope, and acceleration. Results indicate significant effects of these variables on CO2 emission rates, with a strong positive correlation between vehicle speed, road slope, and CO2 emissions. CO2 emission rates substantially increase when vehicle acceleration exceeds five m/s. The proposed model demonstrates high accuracy in predicting on-road CO2 emissions, with correlated factor (R2) values ranging from 0.986 to 0.990 and RMSE values ranging from 0.165 to 0.167. These findings have implications for developing strategies to mitigate emissions in the transportation sector.