Real-Time Control Co-Design for Reconfigurable Energy-Harvesting Systems

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

This grant will fund research that enables renewable energy generation systems, such as wind turbines and kites, to operate optimally over a wide range of environmental conditions, thereby promoting the progress of science and advancing the national prosperity. Environmental variability poses operational challenges for energy-harvesting systems that can be addressed using real-time reconfigurability: simultaneous, on-the-fly changes to both the physical mechanism (plant) and the system control to ensure sustained high efficiency. Offline design methodologies can select optimal plant and control parameters for fixed conditions. No such methodology exists, however, for real-time operation in response to environmental variability and uncertainty, due to critical differences in time scale and effort required to modify plant parameters and control parameters, respectively. This project will fill this knowledge gap by developing an innovative algorithmic framework that accounts for such differences in real-time operation and will further quantify the computational requirements to make real-time reconfigurability worth additional costs and complexity. The PI will leverage engagement with the International Energy Agency Task on Airborne Wind Energy to organize annual workshops on the use of reconfigurable energy-harvesting-system models and open-source software tools created in this project. Student engagement with renewable energy technology will be promoted by implementing a “Physics of Kites” workshop in existing K-12 programming. This research aims to develop the foundations of a receding horizon co-design framework for real-time plant reconfigurability while also addressing fundamental distinctions between plant and control parameters. It accomplishes this outcome by fusing notions from nested co-design and multi-rate hierarchical model predictive control, addressing critical knowledge gaps that arise due to the simultaneous need to (i) consider an economic (rather than tracking) formulation at both levels of the hierarchy, (ii) incorporate surrogate models for tractability, and (iii) consider environmental stochasticity. Specifically, a multi-rate architecture will be investigated whereby a low-order surrogate model is used by the upper-level plant optimization to approximately capture the anticipated behavior of the lower-level control system optimization. An interconnected error system model and small gain framework will be used to address questions of convergence and efficiency under different rates of environmental parameter variation. Finally, recursive Gaussian Process modeling will be used to characterize environmental uncertainty, while reformulating deterministic objective functions into statistical ones, introducing chance constraints, and assessing theoretical properties in a probabilistic sense. The framework will be evaluated through an extensive simulation and scaled experimental validation campaign on an energy-harvesting underwater kite with real-time morphing capability. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.