Session: 16-01-01: Government Agency Student Poster Competition

Paper Number: 150730

150730 - Data-Driven Modeling to Connect Adaptive Twist Morphing Wind Turbine Blades to System Performance 

As the demands of wind energy shift away from improving the cost of energy through progressively larger blades, and towards more innovative solutions, the need arises for improvements in the aerodynamic modeling of turbines to accommodate these innovations, particularly in the context of real-time controls. In recent years, Data-driven modeling has proven to be an effective way to bridge between these innovative technologies, analytical modeling, and data sources ranging from supervisory control and data acquisition (SCADA) data to simulated response data. Presented here is the groundwork necessary for the rapid and accurate development of standard data-driven models connecting system level performance criteria to Active Morphing Blades, a twist morphing adaptive aerostructure technology which uses actuation to modify the twist angle distribution of a turbine blade, changing its aerodynamic properties. Active Morphing Blades, as one such innovative solution to the demands of wind energy, need several developments to bridge from its current design space to commercialization, including materials considerations to enable the desired morphing capability, and high-fidelity, fast models to enable real-time control of the system on an active turbine. This effort contributes to the latter, both through the creation of a process necessary to develop data-driven models capable of real-time control tasks, as well as the development and evaluation of these models. In this effort, turbine performance data is generated in bulk using OpenFAST simulations which incorporate the Active Morphing Blade twist morphing architecture, which is then used to train a selection of data-driven models, including Decision Tree, Support Vector Machine (SVM), Neural Network (NN), and Gaussian Process Regression (GPR) models. These models are each unique to the performance parameter being examined and the particular turbine in use. The exact performance parameter examined is specific to a desired application, however, this effort examines aerodynamic thrust loading as the system-level performance parameter, and does so for the DTU 10 MW and IEA 15 MW reference turbines. These models are comparatively evaluated for performance, in terms of accuracy. In all cases examined, GPR is shown to be optimal for modeling the performance parameter examined for this work, aerodynamic thrust load, with accuracy nearly an order of magnitude improved over the next best model, which is NN. The developed model is also used to demonstrate applicability in an optimization problem for the examined performance parameter, aerodynamic thrust loading, the reduction of which is useful for fatigue and turbine performance applications. Using particle swarm optimization alongside the GPR model, the active morphing blade twist distribution which optimizes aerodynamic thrust loading is determined for particular wind speeds, showing a maximum improvement of 3.15% in the 10 MW reference turbine, and a maximum improvement of 2.66% in the 15 MW reference turbine.

 

Presenting Author: James Roetzer University of North Carolina at Charlotte

Presenting Author Biography: Currently enrolled in the Mechanical Engineering PhD Program at UNC Charlotte, James Roetzer is a graduate student with a Master's in Aerospace Engineering from SUNY University at Buffalo, as well as a two Bachelor's from SUNY University at Buffalo, one in Mechanical Engineering and one in Aerospace Engineering. Having previously worked as a Blade Design Intern with Atrevida Science under the NYSERDA program, and with hands-on experience in wind tunnel experimental testing for the project, James now works toward Integrating Data-Driven modeling techniques into Active Morphing Blades technology.

Authors:

James Roetzer University of North Carolina at Charlotte
Xingjie Li University of North Carolina at Charlotte
John Hall University of North Carolina at Charlotte