Model Based Optimization of a Complete Diesel Engine/SCR System
Achieving upcoming emissions legislation (Euro VI) for heavy trucks is a serious challenge for the manufacturers. Apart from the increasingly strict limits on nitrogen oxides (NOx) and particulate emissions (PM), the limited amount of fossil fuels as well as alarming climate changes simultaneously drives the demand for low CO2 emissions. A likely solution to meet upcoming NOx limits is to use a combination of exhaust gas recirculation (EGR) and selective catalytic reduction (SCR). Combining these two technologies poses new challenges and possibilities when it comes to optimization and calibration. Using a complete system approach, i.e. considering the engine and the aftertreatment system as a single unit, is important in order to achieve a good balance between regulated emissions and CO2/fuel consumption. Optimizing the complete system is a tedious task; first there are a large number of variables which affect both emissions and fuel consumption (injection timing, EGR rate, urea dosing, injection pressure, pilot/post injections for example). Secondly, the SCR catalyst has substantially slower dynamics than the diesel engine and the rest of the system, making the optimization problem time dependent. A novel approach to solve this problem is to use model based optimization which is the topic of this thesis. The first step involves developing computationally efficient models of the diesel engine, exhaust system and SCR catalyst. The engine model consists of a quasi steady gas exchange model and a zero dimensional combustion and NOx formation model. The combustion model uses a two zone concept and the NOx formation is calculated according to the Zeldovich mechanism. The exhaust system model is a dynamic mean value model and considers convective heat transfer only. The SCR catalyst model is based on a state space concept and uses customized ODE solvers to allow long time step lengths. All sub models are optimized for computational efficiency and the complete model has a simulation performance of >10 times real time on a standard PC. The agreement with measurements is excellent; specific NOx engine out is predicted with a -8.5% relative error and NOx conversion over the SCR catalyst is underpredicted with 4.0 %. The optimization problem is formulated to minimize brake specific fuel consumption including urea cost while maintaining NOx and NH3 emissions at Euro VI levels. The steady state optimization problem is solved in an efficient manner by using a simplified steady state solution of the SCR catalyst model. A non linear controller based on the MPC principle is developed to solve the optimization problem for transient sequences. An SQP-based optimization routine is used to solve the MPC optimization problem in each time step. The controller is applied to the certification test cycles (WHSC/WHTC). In the WHSC, the peak NH3 slip is reduced from 85 to 17 ppm while the fuel consumption is unchanged compared to the optimum steady state calibration. Equally good results are achieved in the WHTC. Compared to a conservative steady state calibration with NOx and average NH3 emissions on the same level, peak NH3 slip is reduced from 102 to 47 ppm, and fuel consumption is improved by 1.5%.
Source Type:Doctoral Dissertation
Keywords:TECHNOLOGY; model based optimization; MPC; real time models; model based control; SCR; optimization; EGR
Date of Publication:01/01/2009