Accurately predicting and investigating fuel consumption and emissions is essential for effective emission control and improving fuel efficiency. Previous predictive research on fuel consumption and emissions for hybrid vehicles has primarily focussed on battery state of charge, often overlooking other critical factors, such as engine characteristics, driving behavioural characteristics, and ambient temperature, that influence fuel consumption and emissions. The inclusion of these additional factors remains a significant challenge in balancing model accuracy. Additionally, prior modelling approaches lack adequate representation of influential factors and sufficient interpretability in fuel consumption modelling. To address the aforementioned gaps, this study applied the XGBoost model to chassis dynamometer testing, optimizing it using mutual information. It also employed Shapley additive explanations to explicitly investigate fuel consumption and emissions across drive cycles. The modelling results demonstrated that for emissions, XGBoost improves the average R 2 by approximately 0.9%–13.1%, and for fuel consumption, by about 1%–2.3% over the baseline models. Feature importance reveals that battery state of charge has a firm impact on fuel consumption and CO 2 emissions, whereas a few other pollutants, such as NOx, THC and CH 4 , are primarily governed by engine operating conditions. Further, these effects vary across driving cycles, indicating that optimization strategies should be customized to specific driving conditions to achieve more accurate and efficient reductions in emissions and fuel consumption. Thus, this study is crucial as it advances understanding of hybrid vehicles’ fuel consumption and emissions.
Shaikh et al. (Wed,) studied this question.