A Parallel Supervision System for Vehicle CO2 Emissions Based on OBD-Independent Information

Abstract

Listen

Translate article

A parallel supervision system is built in this paper in order to realize accurate vehicle <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\rm CO_{2}$</tex-math></inline-formula> emission estimation and prediction. On-board diagnostics (OBD)-independent information is used as the model input to avoid costly specialized equipment. More importantly, the OBD-independent model does not rely on an excessive number of internal variables and is capable of predicting carbon emissions from vehicle-level predictive information (future road conditions and planned speed trajectories). Based on the parallel theory, the actual traffic environment is considered the physical world, while the combined <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\rm CO_{2}$</tex-math></inline-formula> model (which consists of physical and data-driven models) is the core part of the artificial world. The physical model uses a cascaded structure with engine speeds and torques as intermediate variables, and the data-driven model relies on a modified long short-term memory (LSTM) neural network. When the historical data are sufficient in size and diversity, the data-driven model is appropriate and achieves more accurate estimations; otherwise, the physical model is preferable because of its greater robustness. In real applications, the supervision system selects the suitable model for adapting to different vehicles; this approach can leverage both the learning ability and physics-based knowledge of the system. A validation is conducted using real-world experimental results to demonstrate the effectiveness of the proposed supervision system. According to comprehensive comparisons between the obtained measurements and estimations, both the physical and data-driven models achieve sufficient accuracy. When the historical data are insufficient, the data-driven model exhibits some deficiencies, while the physical model indicates more robustness even when some primary parameters (gear ratios) are unknown. Moreover, the deterioration factor (DF) of vehicle <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\rm CO_{2}$</tex-math></inline-formula> emissions is modelled to represent the performance of aged vehicles. This parallel supervision system can monitor real-time carbon emissions from actual traffic scenarios, providing a new method for reaching the target of carbon neutrality.