question concerning TurbSim (V1.5)

Hi,

I have a question concerning TurbSim (V1.5):

I specified the wind speed at reference height (Uref) to 8.5 m/s in the .inp file and generated a wind field.
After a simulation with FAST the output no. 21 (‘uWind’) which should be the wind speed at hub height has a mean of 9.8 m/s.
I tried different Uref, but uWind always seems to be larger.

I’ve just tried to set the turbulence intensity to zero, then Uref and uWind are equal.
So why doesn’t a turbulent wind field has a mean(uWind)=Uref?

Kind regards,

Stefan Kapp

Hi, Stefan.

There are typically two things that will cause the average hub-height wind speed to differ significantly from TurbSim’s reference wind speed:
(1) the reference height input in TurbSim is not the turbine hub height (i.e., RefHt /= HubHt), or
(2) the record length (AnalysisTime) is not the same as the length of the time series you are averaging over.

Based on your description of setting the turbulence intensity to zero, I’d say the problem is record length. My guess is that the mean of your uWind time series will vary based on the random seed used in TurbSim. Some seeds will be result in a mean higher than Uref; others lower.

If the AnalysisTime is roughly the same as the length of your FAST simulation, Uref will get closer to the calculated average of uWind. However, keep in mind that FAST (actually AeroDyn) is interpolating the wind fields to get the velocity at hub height on the tower centerline at the specified output times. There will be a filtering effect, which will be particularly noticable if you use an even number of points to create your TurbSim grid. The way AeroDyn interpolates the 3-d grids also means that the FAST output time will always be less than the total length of the wind file being used.

Hope that helps.

Bonnie

Hi Bonnie,

thank you for your reply and your ideas.

(1) I chose RefHt = HubHt = 90m
(2) for testing I chose the same full time ranges
I have an uneven number of grid points.

In the .sum-file TurbSim confirms my chosen Uref=8.5 m/s in the “Mean Wind Speed Profile:” table in line Height=90m in column “Wind Speed” as well as in column “U-comp (X)”.

Averaging the full output uWind results in 8.43 m/s, which is now more close to what it should be.

I can live with that now, but I have another problem, which is probably misplaced here (concerning speed control of the FAST model)

I generated a wind field with zero turbulence and no shear (here, the mean is perfect) in “region 2”. Running now the FAST-code in Simulink, the rotational speed should result in omega = lambda_opt*u/R = 7.55 * 8.5m/s / 63m = 9.727 rpm
Unfortunately the baseline speed controller (which I checked and should be ok) produces omega = 9.921 rpm

Can you explain this remaining offset? Consequently, lambda is not 7.55, but 7.7. I think it should be nearer to the optimal speed.

Thank you very much for your support, I’ve already been trying a lot to solve this…

Stefan

Hi Stefan,

I’m not sure why your model does not reproduce the optimal TSR exactly, but at least the response is close. Here are some possible reasons for differences:

*I calculated my TSR using the EQUIL wake model of AeroDyn. The DYNIN wake model may give different results.
*I calculated my TSR will all proper DOFs in FAST enabled. Disabling DOFs (e.g., making the rotor rigid) may give different results.
*Also, with all proper DOFs enabled, varying the wind speed varies the TSR because different wind speeds lead to different aerodynamic loads, and thus, different turbine deflections.
*The baseline controller for the NREL 5-MW turbine is supplied as a DLL. I’m not sure how you’ve implemented this in Simulink, but perhaps there are differences in our implementations.

By the way: while the difference will be small, the rotor radius used in the TSR calculation should use the effective radius normal to the shaft axis, which in this case is (63m)*COS(2.5deg).

Best regards,

Hi Jason,

thank you very much!
Your first idea was the reason why my results differed from yours, I have used the DYNIN model. What is the main difference between the two models?

I implement the controller as a simple math-function block in Simulink or as a look up table.
And I added the precone of the rotor (2.5°), just a small effect on the controller.

Amazing fast support in this forum!

Best wishes from Europe,

Stefan

Hi Stefan,

The EQUIL model is a quasi-steady model that assumes that the wake is always in equilibrium with the inflow and structural response. The DYNIN model is a dynamic model such that wake responds dynamically to changes in the inflow or structural response.

Under steady conditions, both models would ideally predict the same response. But this is not true in practice as the Prandtl hub- and tip-losses that are part of the EQUIL solution are not used in the DYNIN soluion. Instead, DYNIN calculates the variation of induction along the rotor disk (radially and azimuthally) using internal flow states and there are not enough states implemented in the current version of AeroDyn to properly capture the hub- and tip-losses (more states would mean slower execution time). We plan to add more DYNIN states as an option in a future version of AeroDyn.

Another limitation of the DYNIN model in the current version of AeroDyn is that it is numerically unstable for heavily-loaded rotors (i.e., low wind speeds). So, only the EQUIL model can be applied at low wind speeds. In fact, when using full-field turbulent winds, the current version of AeroDyn will automatically switch the option from DYNIN to EQUIL when the mean wind speed is less than 8 m/s (the mean wind speed of the entire full-field file is used in this calculation).

Given the limitations of the DYNIN model in the preceeding two paragraphs, I decided to apply the EQUIL model (across all wind speeds) when calculating the steady-state behavior of the NREL 5-MW baseline turbine for Chapter 9 of its specifications report.

Best regards,

Hello there,
I am going to run Turbsim in order to generate a wind field to be read by InfloWind in FAST v8.Basically as I am working on OC4 jacket (Test #21) I use the sample input files extracted out from FAST archive at the first and modify some parameters in case of needed.Now there is a doubt for me about URef which has been set to 12 in Turbsim input file relative to test #21 and the question is why it has been set to 12 ?! I have searched many times so far but have not found any reference or recommendation about it.In “Upwind Design Basis” related to K13 deep water site for OC4 project there is nothing mentioned about U10 at reference height(hub height) as well.Would you please help me about this and let me know if this value(URef=12) is a constant one for OC4 project wind field generation ?
Kind Regards,
Arsalan

Dear Arsalan,

URef = 12 is just an example for this sample test case. Normally in wind turbine analysis for normal operation one would run TurbSim and FAST multiple times with different mean wind speeds (URef) between cut-in and cut-out wind speed and multiple simulations with different seeds (in TurbSim) within each wind speed bin.

Best regards,

Dear Jason,
I really appreciate for your detailed answer.Considering cut-in and cut-out wind speeds which are 3 m/s and 25 m/s respectively for 5MW wind turbines analysis within operation time one that is between two values above ,there would be 22 individual wind speed neglecting the 25 m/s.In addition, according to IEC 61400-3 the wind bins contain intervals with length equal two(i.e. Vw 3-5 ,5-7, 7-9 m/s and so on).If you could help me in the following questions I would be pleased.

  1. Is it a good idea considering each interval’s mid-point as a 10-m mean wind speed at hub-height (URef) for each of simulation firstly and then for each of them change the seed values in Turbsim input file for every single simulation based on a particular URef and run the code as many as needed?
  2. If yes is the answer to question above, how many times is needed to change the seed values and run the code again for an OWT in operation?
    Best regards,

Dear Arsalan,

  1. Yes.
  2. IEC 61400-3 recommends 6 seeds per wind speed bin for load case 1.1; however, experience has shown that to get statistical convergence one would need at least 10 seeds.

Best regards,

Dear Jason,
I am so grateful with your answer.But what do you exactly mean by statistically convergence?and how should we determine this convergence ?
And finally last issue about last post ;
should we change the Hs and Tp values and set new ones those have highest percentage of time in scatter digrams relative to each wind bin presented in “Upwind Design Basis”(K13 site) in HydroDyn input file as we change the URef(wind speed bins’ mid-point) for each simulation ?
Sincerely,
Arsalan

Dear Arsalan,

By “statistical convergence”, I mean that the statistical quantities of interest you are after e.g. mean, standard deviation, skewness, 95% quantile, etc. to a fixed value.

Yes, the sea state should be changed to relate to the mean wind speed. IEC load case 1.1 requires that take the expected (mean) value of Hs associated with the mean wind speed and the associated range of wave periods. IEC load case 1.2 requires that you take the full joint probability distribution into account. IEC 61400-3 explains the details.

Best regards,

Dear Arsalan,

By “statistical convergence”, I mean that the statistical quantities of interest you are after e.g. mean, standard deviation, skewness, 95% quantile, etc. to a fixed value.

Yes, the sea state should be changed to relate to the mean wind speed. IEC load case 1.1 requires that take the expected (mean) value of Hs associated with the mean wind speed and the associated range of wave periods. IEC load case 1.2 requires that you take the full joint probability distribution into account. IEC 61400-3 explains the details.

Best regards,

Dear Jason,
You did clarify my doubts completely and I appreciate you again.
Best regards,
Arsalan