Jochem De Schutter
Systems Control and Optimization Laboratory, University of Freiburg
Tuesday, May 18, 2021, 11:00
Airborne wind energy (AWE) systems are typically characterized by highly dynamic flight behavior, nonlinear and unstable dynamics with multiple in- and outputs, as well as structural and flight envelope constraints. These properties make the offline computation of dynamically feasible, power-optimal flight trajectories in combination with system design optimization an intricate endeavor. Optimal control arises as an obvious candidate to tackle this task due to its inherent ability to handle precisely these types of systems. However, despite numerous successful applications in the past decade, as of today there does not exist a generic and open-source optimal control framework tailored to AWE, thereby hindering further dissemination of optimal control as a tool in AWE-related research and industry.
We present AWEbox, an open-source Python toolbox for modeling and optimal control of AWE systems. Among others, it provides an implementation of validated models tailored for optimal control, and it takes away from the user the burden of formulating and numerically solving common AWE optimal control problems (OCP). A distinctive feature of AWEbox is that it supports optimization of multi-drone systems next to the prevalent single-drone variant as well. The second particular focus of the toolbox is robustness with respect to initialization, which we achieve by means of a formal homotopy strategy. The toolbox supports rigid-wing, lift- and drag-mode AWE systems and is capable of embedding user-provided 3D wind profiles. Building on the open-source optimization framework CasADi, it can be a particularly useful component in AWE system toolchains for performance assessment, system design, controller synthesis, or in a wide area of academic research ranging from AWE modeling to optimization algorithms.
In this presentation, we give a high-level overview of the AWEbox software package alongside a short demo. We investigate the efficacy of the specific penalty-based homotopy strategy used while optimizing a single-drone lift-mode system. We also showcase the tool's multi-drone capability by investigating the effect of adding drone-to-drone tethers on the power output fluctuation of a large-scale triple-drone system. This is relevant since power output variability increases drastically when upscaling AWE systems, due to the growing influence of gravity-induced asymmetry in lift power availability.