Unexpected? transient behavior using RANS (kL)

[ews-ffarella]

Hi!
I have been trying to use AMR-Wind for wind resource assessment in complex terrain.
I have created a library to help with the setup (can share the code on demand), but the workflow is pretty much the same as AMRTerrain. I have run a test case in rather complex terrain, and fail to reach a steady state, even after 1200s of physical time. Ok, Ellipsys 3D from DTU also seemed to have a hard time for some wind directions, and the flow field looks crazy when the wind comes from 330° to 50°.

AMR-Wind gives somewhat more plausible results, but at the cost of huge transient behavior.
I would kindly as if you could check my setup and results to see if I am missing anything. 🙏

I have attached a PDF with all the relevant information.

project_hwr_amr_wind_postpro_1.pdf
project_hwr_amr_wind_postpro_61.pdf

I use static refinement instead of the AMRTerrain procedure. See a description if this issue #1629

Domain

inflow angle 0° (huge transient behavior)

inflow 270° (no transient behavior)

[hgopalan]

AMR - Wind RANS modeling is not a steady-state solver. It is a URANS solver. So when you have complex structures, the large scale flow separation from these structures are captured and they propagate downstream and this causes unsteadiness pattern in your wind speed profile and direction. We cannot use this instantaneous value to look for convergence. We need to compute a time-averaged solution. In some cases, when the upstream terrain is not steep, you do not see significant kinks in the wind speed profile.

[ews-ffarella]

Yeah, I figured. Sadly, I dont have any nice monitoring of what the DTU code Ellipsys3D did (it is a commercial offer, through WindPRO, called WAsP-CFD). I just saw that for southern directions for example, the flow field looked rather ok.
But this is true, i have no idea if the steady state solver actually did reach a steady state 😆

This is why I use probes to monitor the evolution of the flow field. I was surprised to see that most of my directions exhibit a transient behavior.

[ews-ffarella]

This is what the flow field using WAsP CFD looks like

I actually sorted the flow cases by the direction at a "reference point", which will be actually used to extrapolate time-series of WS/WD to other locations. You can see that the solver predicts huge recirculation for some directions (340° to 30°)

For wind resource assessment, this is critical. An idea case would be to have a bijection between the reference wind direction and the cfd case direction. See below a plot. X axis is the cfd inflow direction, y axis is the observed wind direction

basically, one needs to build a lookup table, to find for each observed wind direction, which cfd case should be used.
You basically have a transfer function for Speed up and direction deflection, based on the CFD results. So idealy, on the graph above, you always hope that a horizontal line would only cross 1 line (meaning that if i observed let s say 31° at my mast, i need to fecth results from CFD case 20° and 30°, and combine them using simple linear interpolation). If you see where case 20° is inserted, you can imaging how crazy are the results of wind resource assessment.

This is just some feedback about my use-case, not a bug report or anything. This is why I was really looking forward to using the averging feature of AMRWind, but somehow I only get NaNs (also in the solution)