Near wake Fast.FARM

[oguzcrs-del]

Hello OpenFAST community,

I am using FAST.Farm to simulate five NREL 5-MW turbines arranged in a
tandem configuration. My objective is to investigate power density under different turbine spacings and yaw
misalignment angles.

The inflow wind speed is 10 m/s. In some cases, I reduced the
center-to-center spacing between consecutive turbines to values as low
as approximately 1.2D.

Here, S1–S4 are the center-to-center spacings between consecutive
turbines, normalized by rotor diameter. The yaw angles use the
FAST.Farm/OpenFAST sign convention. The fifth-turbine yaw angle was
fixed at 0 deg.
I am observing power distributions that are difficult for me to
interpret. Two example cases are shown below.:

Logically, the power output of case 7 should not exceed the power output of case 21.
And as the distance between turbines decreases, the results become increasingly bizarre.

Because all the other results told me that the power output increased significantly as the distance increased, and that yaw missalignment actually had a very low impact.

Case	S1 [D]	S2 [D]	S3 [D]	S4 [D]	Yaw T1 [deg]	Yaw T2 [deg]	Yaw T3 [deg]	Yaw T4 [deg]	Yaw T5 [deg]	P1 [MW]	P2 [MW]	P3 [MW]	P4 [MW]	P5 [MW]	Total [MW]
7	1.3551	3.0496	2.3813	1.8531	-4.7148	3.3341	26.1179	0.8877	0.0	2.4885	2.5345	1.5159	1.0620	2.1385	9.7394
21	4.2505	4.3622	4.5405	3.6657	-21.0219	-2.5484	18.8681	-7.5233	0.0	3.1034	1.8572	1.3776	1.6678	1.3710	9.3770

My questions are:

1. Is there a recommended minimum turbine spacing for which FAST.Farm
   and its near-wake correction are considered applicable?

2. Is a spacing of approximately 1.2D outside the intended or validated
   range of FAST.Farm?

3. At such small spacings, can the wake-induced rotor-averaged velocity
   decrease below the turbine cut-in wind speed, resulting in shutdown
   or intermittent operation of the downstream OpenFAST turbine?

4. Could strong variations or non-monotonic power values among the
   downstream turbines result physically from wake deflection,
   meandering and wake merging, or would they more likely indicate that
   the model is being used outside its validity range?

I would also appreciate guidance on whether a spacing sensitivity study
such as 1.2D, 1.5D, 2D, 3D and 5D would be an appropriate way to
identify the lower practical limit of the model.

Best Regards.

[jjonkman]

Dear @oguzcrs-del,

Thanks for sharing the nice results showing that FAST.Farm well-predicts the results from Bastankhah and Porté-Agel 2019 at 5D; what reference is that from?

Here are my answers to your questions:

1. The wake model of FAST.Farm aims to provide accurate solutions in the far wake, but not the near wake. The near wake is characterized by strong vortical structures, which become unstable, break down, and mix together. The far wake has a sizeable wake deficit and added wake-added turbulence, but no clear vortical structures. The transition between near wake and far wake depends on ambient inflow and turbine operational conditions, but would normally be in the 3-4D range. So, normally I would say that the wake solution in FAST.Farm is valid beyond a 3-4D spacing.
2. Yes. I would not trust the FAST.Farm solution for such small turbine spacings.
3. Again, I would not trust the FAST.Farm solution for such small turbine spacings. Regardless, even for conditions where the downstream turbines are in the far wake of upstream turbines, the situation may arise where the downstream turbine ingests wind that is below its cut-in speed, which could trigger a turbine shutdown if that mode is considered by the controller enabled in the simulation.
4. I'm not fully sure I understand your question, but power loss and power fluctuations down the row of a wind farm can certainly be caused by wake deficits, wake meandering, and wake-added turbulence.

Unfortunately, FAST.Farm does not provide any indication that the solution will be off when the turbine spacing is less than 3D. Rather, FAST.Farm simply neglects physics that is important in the near wake, i.e., the vortical structures and their breakdown, so, the solution should not trusted for spacings less than 3D.

FYI: We recently kicked off work to combine the free-vortex wake solution of OLAF, which accurately resolves the near wake, with the far-wake solution of FAST.Farm. This should combine the best of both worlds. More information will be forthcoming.

Best regards,

[oguzcrs-del]

Thansk your reply dear @jjonkman,

I redesigned the spacing between 3D-8D, result assume more meaningful now.
But,

I am running a FAST.Farm simulation with five NREL 5-MW turbines arranged in tandem. The freestream wind speed is 10.5 m/s, and the downstream turbines operate within the wakes of the upstream turbines.

I am currently investigating the aerodynamic operating conditions of each turbine using the AeroDyn outputs RtAeroCp and RtTSR. The attached figure shows the Cp–TSR relationship obtained from the simulation after excluding the initial flow-development period.

For the first turbine, the operating points remain close to the expected TSR region. However, for the downstream turbines, I observe significantly larger variations in both TSR and Cp. In particular, some downstream turbines reach instantaneous RtAeroCp values of approximately 0.65–0.74 and TSR values above 10.

I also checked RtTSR independently using

[
\lambda = \frac{\Omega R}{U_{\mathrm{rotor}}},
]

where I used RotSpeed and RtVAvgxh. The independently calculated TSR agrees almost exactly with the OpenFAST RtTSR output. Therefore, the large TSR values appear to be associated with the reduced rotor-averaged velocity in the wake rather than an error in the output.

My questions are:

2. In strongly waked and unsteady conditions, is it expected that instantaneous RtAeroCp can become significantly larger than the conventional steady-state maximum Cp of the NREL 5-MW turbine?

3. Are RtAeroCp and RtTSR appropriate quantities for constructing and comparing Cp–TSR curves of upstream and downstream turbines in FAST.Farm?

4. If the objective is to compare the aerodynamic performance of all turbines using a common reference, would it be more appropriate to calculate

[
\lambda_{\infty}=\frac{\Omega R}{U_{\infty}}
]

and

[
C_{p,\infty}=
\frac{P_{\mathrm{aero}}}
{\frac{1}{2}\rho A U_{\infty}^{3}},
]

using the ambient/freestream velocity rather than the local rotor-averaged velocity?

5. Alternatively, is there an existing FAST.Farm/AeroDyn output that provides Cp and TSR referenced to the ambient or undisturbed wind velocity?

I would particularly appreciate guidance on the recommended way to report Cp–TSR behavior for downstream turbines operating in strongly waked conditions.

Thank you.

[jjonkman]

Dear @oguzcrs-del,

Here are my responses to your questions:

1. The baseline controller of the NREL 5-MW baseline turbine maximizes power in Region 2 (from 7.8 to 11.4 m/s local hub-height wind speed), but uses higher-than desired rotor speed (hence higher TSR) in Region 1.5 (from 3 to 7.8 m/s local hub-height wind speed) to limit rotor-speed variations and avoid resonating the tower. So, I'm not surprised you are seeing higher TSR for the waked turbines downstream compared to the upstream turbine.
2. As discussed in other topics, e.g. see: issue Study on cp-lambda curve #2932 or https://forums.nlr.gov/t/drive-train-model-validation-compared-to-fast/2017, by nature of their calculation within the aero-elastic solution, I would only use AeroDyn output RtAeroCp in a time-averaged sense to avoid apparent violations in the Betz limit that may occur instantaneously as a result of the calculation approach.
3. As long as you are using these outputs in a time-averaged way.
4. When examining turbine performance of different turbines in the wind farm, you could certainly do as you propose and use the same ambient (undisturbed) wind speed when computing Cp and TSR of each turbine. But I would say for farm-level performance, it is more common in the wind community to examine the absolute power of each turbine directly, rather than the Cp of each turbine.
5. No, FAST.Farm does not currently have outputs for Cp or TSR for each turbine based on the ambient (undisturbed) wind speed.

Best regards,