Supervisory control for emission compliance of heavy-duty vehicles

Licentiate thesis, 2020

Heavy freight trucks globally contribute to a significant proportion of transport-related air pollution. The dominant air pollutants from heavy freight trucks with diesel engine and exhaust aftertreatment system (EATS) are CO2, hydrocarbons (HC), CO, particulate matter (PM), NOX (NO and NO2), and NH3. The greenhouse gas emission legislation limits the amount of CO2 emission, and Euro VI emission legislation limits the other dominant air pollutants. Emission legislation is gradually becoming more and more stringent to reach the long term goal of near-zero-emission. Several parties are working together to reduce the emissions, keeping both the short and long term goal in mind. Any step which results in ICE downsizing contributes to the reduction of all dominant emissions. But, with size and type of the ICE decided, there is a trade-off between NOX emission and other emissions: reduced NOX emission means reduced fuel efficiency (i.e. increased CO2, PM, and HC emissions).

It is a challenge to fulfil the current emission legislation—especially real-driving NOX emission legislation—with existing control functionalities in the engine management system (EMS). However, better control of NOX emission is possible by exploiting predictive driving information and considering the coupling between the engine system and EATS. This work pursues this idea and concludes that fulfilling real-driving NOX emission legislation is possible, considering the coupling between the engine system and EATS while using predictive information. The work provides a mathematical formulation of the concept and then develops, evaluates, and implements an engine-EATS supervisor which optimizes total fuel consumption and fulfils both the world harmonized transient cycle (WHTC) based and real-driving NOX emission legislation. The developed supervisor is a distributed economic nonlinear model predictive controller (E-NMPC). This work develops and analyzes two different versions of the distributed E-NMPC based supervisory control algorithm. The more efficient one of the two is again compared for three variants. Considering the computation time of the three algorithms and processing speed of the existing EMS, one algorithm is selected for implementation.

The supervisor performs much better compared to a baseline controller (optimized offline). Simulation results show that the supervisory controller has 1.7% less total fuel consumption and 88.4% less NH3 slip, compared to the baseline controller, to achieve the same real-driving NOX emission.