Jochem De Schutter
University of Freiburg
Thursday, January 25, 2024, 14:00 - 16:00
Room 02-016/18, Georges-Koehler-Allee 101, Freiburg 79110, Germany
This thesis investigates optimal control formulations and solution methods aimed at the dynamic optimization of periodically operated and nonlinear energy conversion systems. We consider two different applications: airborne wind energy (AWE) and low-temperature combustion (LTC).
Flight trajectory optimization is a central task in the design and operation of AWE systems. As a first contribution, we provide a benchmark for initial guess refinement methods for AWE optimization problems. These refinement methods greatly improve solution times and reliability, in particular in the absence of a priori expert knowledge.
Techno-economic performance at utility-scale is crucial for the commercialization of AWE. Therefore, we propose and investigate, using optimal control, two novel upscaling strategies based on multi-wing systems. The first strategy relies on a stacking concept and allows to increase system capacity independent of wing size. The second strategy is based on multi-wing systems operating in vertical wind farms and maximizes the power density per ground area. We also propose and investigate an alternative multi-wing concept that could potentially be applied in either of both upscaling strategies: a pumping rotary AWE system.
For LTC, we propose a tailored problem formulation and solution algorithm for model predictive control (MPC) and moving horizon estimation (MHE) of a gasoline engine. The approach allows the efficient treatment of the high number of simulation variables arising from modeling the engine cycle-to-cycle dynamics with differential equations. The combined MPC and MHE scheme is real-time feasible on the embedded hardware of an existing engine test set-up and it is validated in real-world experiments.
Finally, we introduce the AWE optimization toolbox AWEbox, which implements the single- and multi-wing models and solution methods investigated in this thesis, with a user-friendly interface. We also propose TuneMPC, a framework which implements an economic tuning technique for tracking MPC schemes, and we demonstrate its capability on an AWE example.