Wake model parameters calibration in FLORIS

[fbelli997]

Dear all,

I am currently working on calibrating the wake model parameters in FLORIS. To this end, I am considering using WindSe to generate steady-state flow conditions. Do you think this is a reasonable approach to pursue? Do you believe it is meaningful to perform this calibration using RANS, or would LES be necessary for reliable results?

I also have some cross-plane measurements collected during wind tunnel campaigns, but the number of available cases is limited and does not fully cover the entire operational range.

Thank you in advance for your feedback.

Kind regards,
Federico

[Bartdoekemeijer]

Hey Federico,

If you are looking to make FLORIS accurate for a single site, then your best shot is to calibrate the parameters against measurements of that specific site.

If you are looking for a "generally applicable" (i.e., not just accurate for a single site) set of model parameters, I would generally recommend:

1. Calibrating against field data from full-scale commercial wind farms
2. Calibrating against a large-eddy simulation model that has been compared/validated against field data in the past.
3. Using the default model parameters (which were tuned against SCADA/LES in the past).

Following my gut feeling, I would stay away from calibrating against lower-fidelity models and wind tunnel experiments. You will undoubtably end up with a different set of model parameters, but I seriously wonder whether they will actually be better.

[misi9170]

Hi Federico,

To chip in here: I agree with Bart that calibrating against (quality-controlled) field data is best, and that comparing to LES is probably next best and preferable to wind tunnel experiments (even though several of the wake models were originally derived and validated based on wind tunnel experiments).

What I'm less sure about is whether LES data is necessarily better than RANS data as a calibration benchmark. I'm not a fluids dynamicist by any stretch of the imagination, but if the RANS simulations can accurately capture time-averaged flows, they may be a reasonable substitute for LES (which we usually perform some time averaging over before calibrating analytical wake models to anyway). There is always a question of "how long" the tie averaging represents, which is a nuance you may need to be careful with.

Please don't take this as a recommendation that calibration to RANS will work well---it's more of a statement that we have not tried to do this in the past, so I don't have a strong opinion about it. I haven't personally worked with WindSE very much at all, but I'll tag @bayc here, because he has worked with both FLORIS and WindSE.

Misha

[fbelli997]

Hi Misha,
Thanks a lot for your message and for sharing your thoughts, much appreciated!
It would indeed be very helpful to hear @bayc opinion on this.

Federico

[bayc]

Hello @fbelli997 ,

I also second what Bart said.

Regarding using WindSE, to get the best results, you would want to tune your WindSE turbine power/thrust curves to match a turbine's performance. Though to claim WindSE as a ground truth would still be somewhat suspect without validating it against either field or LES data, is my understanding. One of the developer's passed along to me that they would trust the relative power production of turbines, but not the absolute power production without some calibration.

Chris