Non-steady Dynamics of Atmospheric Turbulence Interaction with Wind Turbine Loadings through Blade-boundary-layer-resolved CFD
Open Access
- Author:
- Vijayakumar, Ganesh
- Graduate Program:
- Mechanical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- April 27, 2015
- Committee Members:
- James Gordon Brasseur, Dissertation Advisor/Co-Advisor
James Gordon Brasseur, Committee Chair/Co-Chair
Eric Paterson, Committee Member
Philip John Morris, Committee Member
Marcelo Chamecki, Committee Member - Keywords:
- Atmosphere
Wind Turbines
LES
Hybrid URANS-LES
ABL
Atmospheric Boundary Layer - Abstract:
- Modern commercial megawatt-scale wind turbines occupy the lower 15-20% of the atmospheric boundary layer (ABL), the atmospheric surface layer (ASL). The current trend of increasing wind turbine diameter and hub height increases the interaction of the wind turbines with the upper ASL which contains spatio-temporal velocity variations over a wide range of length and time scales. Our interest is the interaction of the wind turbine with the energetic integral-scale eddies, since these cause the largest temporal variations in blade loadings. The rotation of a wind turbine blade through the ABL causes fluctuations in the local velocity magnitude and angle of attack at different sections along the blade. The blade boundary layer responds to these fluctuations and in turn causes temporal transients in local sectional loads and integrated blade and shaft bending moments. While the integral scales of the atmospheric boundary layer are O(10−100m) in the horizontal with advection time scales of order tens of seconds, the viscous surface layer of the blade boundary layer is O(10 − 100 μm) with time scales of order milliseconds. Thus, the response of wind turbine blade loadings to atmospheric turbulence is the result of the interaction between two turbulence dynamical systems at extremely disparate ranges of length and time scales. A deeper understanding of this interaction can impact future approaches to improve the reliability of wind turbines in wind farms, and can underlie future improvements. My thesis centers on the development of a computational framework to simulate the interaction between the atmospheric and wind turbine blade turbulence dynamical systems using a two step one-way coupled approach. Pseudo-spectral large eddy simulation (LES) is used to generate a true (equilibrium) atmospheric boundary layer over a flat land with specified surface roughness and heating consistent with the stability state of the daytime lower troposphere. Using the data from the precursor simulation as inflow conditions, a second simulation is performed on a smaller domain around the wind turbine using finite volume CFD with a body-fitted grid to compute the unsteady blade loads in response to atmospheric turbulence. Analysis of the precursor LES shows that the advective time scales multiple rotation time scales of the rotor. From blade element momentum theory coupled with LES of the ABL, we find that the energy-containing eddies were found to cause large temporal fluctuations (±50%) in the integrated moments, primarily due to changes in the local flow angle relative to the local chord sections. A low-dissipation pseudo-spectral algorithm was applied to the ABL LES. A finite volume algorithm was required to resolve the flow features around the complex blade geometry. The effect of the finite volume algorithm on the accuracy of it’s prediction of the rough-surface ABL was assessed using the method of Brasseur and Wei [1]. We found that finite volume algorithms need finer horizontal grid resolution to retain the same accuracy as the corresponding pseudo-spectral simulations. These results were used to design our computational framework to accurately propagate the turbulence eddies through the finite volume domain. The ability of our computational framework to capture blade boundary layer dynamics in response to atmospheric turbulence is intimately associated with the extreme care taken in the design of our grid and with the development of a new hybrid URANS-LES turbulence model. The new turbulence model blends a 1-equation LES subgrid model in the far field with the k−w−SST−SAS URANS model to the blade boundary layer adjacent to the blade surface. With this computational framework, we simulated a single rotating blade of the NREL- 5MW wind turbine in the moderately convective daytime atmosphere using blade-boundary-layer-resolved CFD simulations. The analysis of load fluctuations on a single rotating blade in a daytime atmosphere using blade-boundary-layer-resolved CFD has yielded two key results: (1) Whereas non-steady blade loadings are generally described as the response to non-steadiness in wind speed, our analysis show that time changes in wind vector direction are a much greater contributor to load transients, and strongly impact boundary layer dynamics; (2) largest temporal variations in loadings result from three distinct dynamical responses with disparate time scales: advection of atmospheric eddies through the rotor at the minute time scale, blade response at the rotor rotation time scale (5s) and blade response to turbulence-induced forcings as the blades traverse internal atmospheric eddy structure at sub-blade rotation time scales. In our simulations at rated wind speed, quasi-2D blade boundary layer separation is observed over most of the outer 50% of the blade with chordwise motions, correlated with time changes in relative wind vector angle, which itself is strongly correlated with changes in blade sectional and integrated loads. Thus, tools based on sectional “table lookups” like FAST [2] and Actuator Line Methods [3], improved using data from high-fidelity simulations and experiment, have the potential to capture the major fluctuations in integrated loads from daytime atmospheric turbulence.