Home/WWT. Wind, Wakes and Turbulence/WWT. Wind, Wakes and Turbulence 1

The Science of Making Torque from Wind (TORQUE 2018)

20 June  |  21 June  |  22 June  |  Events View
AA | CM | DSE | MET | MST | WWT

Loading Events
This event has passed.

Room: BL.28 Carassa e Dadda
Chaired by: Stefan Ivanell | Uppsala
Topic: WWT. Wind, Wakes and Turbulence
Form of presentation: Oral
Duration: 90 minutes

Authors:
Paul van der Laan, Søren Juhl Andersen

Abstract:
Wind turbine wakes can cause energy losses and increase wind turbine loads in wind farms. Computational Fluid Dynamics (CFD) can be employed to calculate these effects. A relatively affordable CFD method is Reynolds-averaged Navier-Stokes (RANS), where the averaged flow is directly calculated and all turbulence is represented by a turbulence model. The widely used k-epsilon turbulence model is known to under-predict the velocity deficit in the near wake due to an over-prediction in turbulent mixing. Therefore, RANS modelers have developed new and applied existing extended k-epsilon models that often limit the turbulent mixing in the wake by reducing the turbulent length and/or time scales. However, not much is known about how well the limited turbulent length and time scales compare with high fidelity CFD models as Large Eddy Simulation (LES), which resolve most of the important turbulence scales. Moreover, it is not well known what the turbulent length and time scales of wake actually is. In this work, we use LES data of a number of single wind turbine wake cases operating in atmospheric turbulence to answer these questions.

Authors:
Tuhfe Göçmen, Gregor Giebel

Abstract:
One of the ancillary services the wind farms are required to provide to the system operators is reserve power, which is achieved by down-regulating the wind farm from its possible power. The most recent grid codes dictate the quality of the possible power at the wind farm level to be assessed within 1-min intervals for offshore wind power plants. Therefore, the necessity of a fast and reliable wake model is more prominent than ever. Here we investigate the performance of two engineering wake models under 1-sec resolution SCADA data on three different offshore wind farms, given the quantified input uncertainty. The preliminary results show that, even wind farm specific training of the model parameters might fail to comply with the strict criteria stated in the grid codes, especially for the layouts with significant wake losses observed. In order to tackle the inadequacy of the engineering wake models to capture some of the dynamics in the wind farm flow due to the embedded assumptions, purely data-driven techniques are evaluated. The flexibility of such an on-line model enables ‘site-turbine-time-specific’ modelling, in which the parameters are defined per turbine and updated with each time-step in a specific wind farm.

Authors:
Michael F. Howland, Aditya Ghate, Sanjiva K. Lele

Abstract:
Wind turbines with large rotor diameters create wakes which are affected by the rotation of the earth. Aside from creating horizontal mean velocity veer, the Coriolis force, caused by earth’s rotation, also results in wake deflection. In atmospheric turbulence, the horizontal Coriolis component is often neglected since its magnitude is small compared to buoyant forces. However, at lower latitudes, the influence of this horizontal forcing will cause vertical wake deflection of the same order as the horizontal wake deflection imposed by the vertical Coriolis component. Using both uniform and stable atmospheric boundary layer inflow conditions on wind turbine arrays, the effects of the horizontal Coriolis component are investigated in large eddy simulation. Results indicate that at these lower latitudes, the horizontal Coriolis force cannot be neglected in the study of wind farms. A low-order Coriolis force-induced wake deflection model is proposed and tested against numerical results.

Authors:
Ryan King

Abstract:
We introduce a data-driven machine learning framework for improving the accuracy of wind plant flow models by learning a model correction based on data from higher-fidelity simulation or experimental field data.  First, a high dimensional optimization problem is solved using adjoint gradients to determine model correction fields that improve performance of the low-fidelity model.  A supervised learning problem is then solved to train a Gaussian process regression model that can predict the correction field based on local flow field velocities and positions parameterized by the rotor diameter.  We train the Gaussian process model on nacelle-mounted lidar data and show its predictive capabilities on large eddy simulations of turbines with different rotor diameters.

20 June  |  21 June  |  22 June  |  Events View
AA | CM | DSE | MET | MST | WWT

Go to Top