Long short-term memory network-based emission models for conventional and new energy buses
Zhuoqun Sun, Chao Wang, Zhirui Ye, Hui Bi
Abstract
Public transportation is regarded as a mitigation measure for addressing climate change and air quality deterioration. However, it is still necessary to estimate the emissions of transit buses, in particular when heavy loads and long periods of operation result in increased emission levels. Consequently, the primary objective of this study is to establish a method to estimate CO, CO2, HC, NOx emissions of buses with four different fuel types including gas-electric hybrid electric buses (GEHE buses), compressed natural gas buses (CNG buses), EURO 4 heavy-duty diesel engine buses (EURO 4 buses) and EURO 5 heavy-duty diesel engine buses (EURO 5 buses) based on the long short-term memory network. The proposed models can fully consider the time dependence of emissions response to vehicle operation situation. In addition, to evaluate the performance of the proposed models, an effective emission model which also addressed the time dependence of emissions by taking the elapsed time of acceleration and deceleration into account, was developed for emissions on each route for comparison. According to the estimation results, the emission models developed in this study performed better than the compared method in terms of emission rates and average emission factors, whose root mean squared errors (RMSE) were explicitly lower than the compared method, mean absolute percentile errors (MAPE) were lower than 50% of the compared method, and the predicted average emission factors were relatively more accurate than those of the compared models.