Numerical Instabilities in Turbulence Intensity Calculations with FAST.Farm

[fbelli997]

Hi everyone,

I'm currently working on extracting the turbulence intensity from a set of FAST.Farm simulations, with wind speeds ranging from 3 m/s to 25 m/s in steps of 1 m/s.
For these calculations, I post-processed the VTK files generated by the simulations and calculated the turbulence intensity on the wind field using the formula: TI = std/U where std is the standard deviation of the wind speed, and U is the mean wind speed.
For wind speeds up to 12 m/s, using a dt_low of 2 seconds, the turbulence intensity is stable and provides accurate results, as shown in the first figure:

However, as I increase the wind speed beyond 12 m/s, I begin to observe issues and numerical instabilities in the data, as demonstrated in the second figure:

These instabilities worsen as the wind speed increases.
I tried reducing dt_low, and this seems to improve the results somewhat for wind speeds between 17 and 20 m/s. However, when simulating wind speeds above 20 m/s, even with dt_low set to 0.2 seconds, the results remain problematic.
Has anyone encountered similar issues when working with FAST.Farm? Do you know why these instabilities occur and how to mitigate them?

Thank you in advance for your insights and suggestions!

Best regards,
Federico

[jjonkman]

Dear @fbelli997,

Which wake model are you using? How are you generating the inflow?

Are you following the FAST.Farm modeling guidance (https://openfast.readthedocs.io/en/main/source/user/fast.farm/ModelGuidance.html) when setting up the spatial-termporal discretizations of the FAST.Farm model, e.g., by using the Python scripts from the OpenFAST Toolbox (https://github.com/OpenFAST/openfast_toolbox/tree/main/openfast_toolbox/fastfarm)? In general, the FAST.Farm modeling guidance recommends using a DT_Low that scales inversely with the wind speed (the faster the wind speed, the smaller DT_Low should be).

What is a bit surprising to me in your results is that the oddities you are seeing in TI are not in the region of the wake, but rather in the free stream beyond the boundaries of the wake.

Best regards,

[fbelli997]

Dear Jason,
I'm actually using the "Polar" wake equation model, with the WAT unofficial version of FAST.Farm, now I'm trying to implement on the 4th version(The most recent one). I generated the Inflow with Turbsim, with KAIMAL turbulent model, no shear, TI= 3% settings (the simulation of the first picture that I attached was 4%, second was 3%).
I'm not using the Python scripts, however I am aware of the equations which stabilize the numerical approach, so for my simulation the "maximum allowed" DT-low = 2.96 s, based on wind of 17 m/s, D-wake = 240 m, C-meander = 2.1.
However for calculating TI each node of the Low resolution domain, I had no choice but to decrease the DT-low=0.2 s for having more frequent temporal .vtk files, and also the corresponding num-planes = 1271 for 15D becomes.
Therefore, I appreciate if you come up with some idea about this inconsistency in my simulation results.

[jjonkman]

Dear @fbelli997,

Thanks for clarifying. Your DT_Low sounds sufficiently small.

Again, what is a bit surprising to me in your results is that the oddities you are seeing in TI are not in the region of the wake, but rather in the free stream beyond the boundaries of the wake. So, perhaps there is no issue with the wake calculation at all, but rather in your calculation of TI?

Best regards,

[fbelli997]

Dear Jason, thank you for your feedback. It's possible that I made an error during the post-processing. I'll review it to confirm.