DC4: Fully Autonomous AWE operation over Long Time Horizons

Objectives

For AWE to be commercially successful, they must operate for long time periods fully autonomously with minimal human interaction. Accordingly, the system must be capable of monitoring its own conditions, deciding when to take-off, when to land, reconfiguring operations based on a continuous monitoring and estimation of the wind field and accomplishing plant-wide requests regarding the desired power output. To achieve long-term autonomous operation, the system must complete these operations reliably, in the face of changing conditions (e.g., weather) and faults (e.g., grid loss). This research will establish a systematic, theoretically solid, and computationally viable approach to design such a decision-making system. The research will start by establishing the full range of AWE operational modes and the conditions where the supervisory control must act. Next, the DC will formalize a hierarchical and distributed control structure, according to the principle of increasing intelligence and decreasing accuracy when climbing up the hierarchy, from precise low-level control systems such as kite attitude and reeling force, to learning-based high-level decision-making systems to choose the most suitable operating phase. Finally, a computational method to verify the designed automation system will be researched, able to isolate possible single points of failure in the decision-making logics, to provide the designer with a feedback of which hardware/software modifications shall be made to increase fault tolerance. The effectiveness of the developed approaches will be verified using highly accurate system models combined with Monte Carlo simulations, as well as ad-hoc simulations of particularly challenging operating conditions and faults. The DC involved in this project will also interact with the various stakeholders in the definition of regulations and standards regarding the high-level control functions of fully autonomous AWE and the related specifications.

Expected Results

Hierarchical and distributed control layout for all-round AWE autonomy; Design methods for the control systems involved at the various levels; Extensive validation of the performance via detailed simulations.

Supervisory team

Lorenzo Fagiano is main supervisor, Alessandro Croce is co-supervisor.

Planned secondments

Technical University of Denmark (M15-M21) to set up realistic simulations of unsteady wind conditions and account for such situations in the automation system, supervised by Michael McWilliam and Mark Kelly. Jointly between HM/kiteKRAFT in Munich M33-M37 to develop automation strategies to account for grid compliance norms and requests from the grid operator and testing on the kiteKRAFT system, supervised by Christoph M. Hackl and Florian Bauer.