Some questions related to aerodynamic simulation for a 16MW wind turbine in Openfast

Dear Sir/Madam,

I am currently engaged in a project where I'm utilizing OpenFAST to simulate a 16MW wind turbine. My primary focus lies in examining the aerodynamic and structural characteristics of the long and flexble blade, such as: power curve, thrust curve, and blade-tip flapwise deflection,,, The core objective of my study is to conduct a comparative analysis of the simulation outcomes derived from various AeroDyn and ElastoDyn models. Specifically, I'm interested in understanding the differences in performance between the Blade Element Momentum (BEM) and the OLAF approaches for aerodynamics, as well as comparing the linear structural model to BeamDyn for the structural calculations of the blade.

Below, I have detailed some of my findings and have several questions regarding them. I would greatly appreciate your guidance and insights on these matters. Thank you very much for your time and assistance in advance.

I begin with a simple case, in which I use rigid blade, and use BEM for aerodynamic. I have successfully derived the power curve and the yaw bearing Fx, spanning from cut-in to cut-out wind speeds during power production. Upon comparison with the reference data, the alignment appears to be quite satisfactory, as depicted in the figure below:

Then, I incorporated the blade's flexibility into the simulation by employing ElastoDyn's linear modal-based model. Upon conducting the simulations, I encountered some unstable turbine response, even though I'm using steady inflow wind. I suspect that this may arise from the linear structural model's limitations in accurately representing the complex dynamics of such long and flexible blades. Have you encountered similar outcomes in your research or practice? (Question 1)

Next, I include the flexibility of the blade using BeamDyn, and I can get stable turbine response result now. I compared it with the reference, especially for power curve and blade tip flapwise deflection, it merges will as shown in the figure below. However, to achieve convergence with BeamDyn, I found it necessary to employ a very small timestep—specifically, 0.0001. This adjustment has significantly increased the computational time, resulting in simulations extending beyond 10 hours for a 300-second scenario. Is this a typical time step and computation speed you would expect for BeamDyn simulating a blade more than 120m length? (Question 2)

Then, I tried to shift to use OLAF instead of BEM for the aerodynamic calculation. If I still include BeamDyn for blade flexibility, the computation efficiency is too slow, requiring nearly three hours to process just a few seconds of simulation time series.. Have you used BeamDyn and OLAF together before? What's your experiecne of it's computation efficiency? (Question 3)

Finally, I opted to implement OLAF for the aerodynamic calculations while reverting to a rigid blade configuration. This decision was made to facilitate a direct comparison between the OLAF method and the Blade Element Momentum (BEM) approach, both applied under the condition of a rigid blade. And I have simplified the model by closing Servodyn, and prescribed a constant rotor speed, with a steady inflow wind speed. Then I compared the aerodynamic power, or Cp from OLAF and BEM. I found that the Cp calculated by OLAF is higher than BEM. I have followed the guidence on parameter settings for OLAF here: https://openfast.readthedocs.io/en/main/source/user/aerodyn-olaf/RunningOLAF.html, but there is still large difference.

The inputs and outputs for the simulation are listed below:

Simulation Inputs: Wind speed: 10.8m/s; Rotor speed: 8rpm nNWPanel: 1340 nNWPanelsFree: 335 None far wake panels. simulation length: 100s (ensure steady condition has reached)

Simulation outputs: Aerodynamic power by BEM: 16.3MW Aerodynamic power by OLAF: 17.7MW

Do you have any suggesetion here? How shuold I tune the OLAF model? (Question 4)

Thanks for your time and help! Best Regards,

Dear @WANGSUYU0809,

I'll let others comment on your OLAF-related questions, but here are my responses to questions 1 and 2:

Regarding (1), I would first say that blade mode; in ElastoDyn is not linear, but includes nonlinear effects such as Coriolis, radial shortening, gyroscopic etc. The model should be OK for "moderate deflection" of blades where torsion is not important. See the following post on our forum for more information: https://forums.nrel.gov/t/coupled-blade-modes-in-fast/314. I'm not sure I understand what instabilities you are referring to; can you clarify?

Regarding (2), you asked a similar question that was answered in the following OpenFAST issue: https://github.com/OpenFAST/openfast/issues/2106. That said, a DT of 0.0001 s is quite minuscule. Can you clarify whether your reference axis in BeamDyn is smooth (as is generally recommend), as well as the source of your 6x6 cross-sectional mass and stiffness matrices?

Best regards,

Hi, Regarding OLAF, the best is to follow the guidelines (though they are a bit conservative). When running with BeamDyn, you need the DT of OLAF to be a multiple of the DT of the glue code (the guideline script should take care of that).

Simulations with BeamDyn+OLAF is be slow at the moment not so much because of OLAF but because of the AeroDyn implementation which computes inflow at every wake points for each glue code time step.

I hope that helps.

Dear @WANGSUYU0809,

I'll let others comment on your OLAF-related questions, but here are my responses to questions 1 and 2:

Regarding (1), I would first say that blade mode; in ElastoDyn is not linear, but includes nonlinear effects such as Coriolis, radial shortening, gyroscopic etc. The model should be OK for "moderate deflection" of blades where torsion is not important. See the following post on our forum for more information: https://forums.nrel.gov/t/coupled-blade-modes-in-fast/314. I'm not sure I understand what instabilities you are referring to; can you clarify?

Regarding (2), you asked a similar question that was answered in the following OpenFAST issue: #2106. That said, a DT of 0.0001 s is quite minuscule. Can you clarify whether your reference axis in BeamDyn is smooth (as is generally recommend), as well as the source of your 6x6 cross-sectional mass and stiffness matrices?

Best regards,

Dear @WANGSUYU0809,

I'll let others comment on your OLAF-related questions, but here are my responses to questions 1 and 2:

Regarding (1), I would first say that blade mode; in ElastoDyn is not linear, but includes nonlinear effects such as Coriolis, radial shortening, gyroscopic etc. The model should be OK for "moderate deflection" of blades where torsion is not important. See the following post on our forum for more information: https://forums.nrel.gov/t/coupled-blade-modes-in-fast/314. I'm not sure I understand what instabilities you are referring to; can you clarify?

Regarding (2), you asked a similar question that was answered in the following OpenFAST issue: #2106. That said, a DT of 0.0001 s is quite minuscule. Can you clarify whether your reference axis in BeamDyn is smooth (as is generally recommend), as well as the source of your 6x6 cross-sectional mass and stiffness matrices?

Best regards,

Thanks for your reply! To clarify my Question1, I have attached a figure below, to show the instable response I got from elastodyn blade, there is large oscilation on blade root moment load.

To clarify my Question2, I have attached a figure below, to show the reference axis I used to build beamdyn blade, I think it could be called as 'smooth'. I have set the structural twist angle to 0 along the blade, but the mass and stiffness matrix is calculated considering the principal direction of mass and stiffness. I have sectional mass and stiffness data refers to sectional mass center, elastic center and the principal direction from a Bladed model. I transfer it to a 6x6 matrix based on this paper 'BeamDyn inputs from sectional beam properties'.

Hi, Regarding OLAF, the best is to follow the guidelines (though they are a bit conservative). When running with BeamDyn, you need the DT of OLAF to be a multiple of the DT of the glue code (the guideline script should take care of that).

Simulations with BeamDyn+OLAF is be slow at the moment not so much because of OLAF but because of the AeroDyn implementation which computes inflow at every wake points for each glue code time step.

I hope that helps.

Hi, thanks for your reply! Actually I have already followe the guidelines you mentioned while setting OLAF parameters file, including the number of panels (I have listed the numper of panels I used above) and regularization parameters, but still I think the aero power calculated by OLAF is a bit high compared to BEM result. Do you have any comments or suggestions here? Should I try another way to tune OLAF till its result are close to BEM result, or should we accept this is indeed the difference between 2 different aerodynamic method, and it's hard to tell which one is more accurate without comparison with real field measured data?

Dear @WANGSUYU0809,

Regarding (1), I'm not sure what is driving the large oscillations in your blade-root moments, but I would suggest simplifying the model (e.g,, disabling the generator DOF with a fixed rotor speed rather than relying on the controller, disabling other structural DOFs, or using simpler aerodynamic models) to better isolate the issue.

Regarding (2), I a bit concerned about the "kink" in kp_yr near the root. To ensure it is smooth, I would recommend fitting the reference curve with a low-order polynomial.

Best regards,

OpenFAST / openfast

Some questions related to aerodynamic simulation for a 16MW wind turbine in Openfast #2132