Nonlinear model predictive control of an automotive waste heat recovery system

This thesis considers a waste heat recovery system for heavy-duty trucks that is based on the Organic Rankine Cycle and uses the exhaust gas heat to evaporate an organic working fluid. The evaporated working fluid expands over an expansion machine, which drives an electric generator, and finally condenses. The considered system differs from similar concepts because it is operated with a hydraulically closed low-pressure part. This allows decreasing the outlet pressure of the expansion machine below the atmospheric pressure, and thus increases its efficiency.
The aim of this thesis is to develop a real-time capable nonlinear model predictive control strategy for the evaporation and the condensation process that considers a significant model-plant mismatch and uses power maximizing reference values for the system states.
A detailed first-principles mathematical model is presented, which allows to easily adapt the system model to different system configurations. To determine power maximizing reference values, the optimal steady-state system states are determined. Subsequently, these operating points are used to derive an optimal control reference for the dynamic system operation.
The system dynamics is approximated by a suitable gain scheduling approach. The calculation of the optimal control inputs finally results in a quadratic program, which can be solved in a numerically very efficient way. The significant model-plant mismatch is modeled by unknown but constant disturbances, which are estimated by a suitable observer and then considered in the controller. The simulation results on the validated simulation model show that the proposed control concept achieves a high tracking performance of the power maximizing control reference, even for a large model-plant mismatch.