Using full-field turbulence files.

Hi all,

Executive Summary:

I’d like your feedback on whether or not you think it is OK to change the way we use turbulence files in AeroDyn. I propose that we no longer abort the program if an analysis point lies outside the wind field and that we use the outer-most point instead. I also propose that we fix the wind field to the non-inertial hub instead of the inertial ground system.


The way TurbSim and AeroDyn work now is that TurbSim adds enough time to the requested length of the simulation to account for the fact that analysis points can be downwind of the turbine at the beginning of the simulation or upwind of the tower at the end of the simulation. The amount of time it adds is the grid width divided by the mean wind speed. For instance, if the turbine has a 90-degree yaw error at the start of the run, the tips may be one rotor radius downwind of the tower. At the end of the simulation, a 90-degree yaw error can cause a blade tip to be one rotor radius upwind of the tower.

I did it this way when I first put the full-field wind-file capability into AeroDyn; I found that the program was trying to index out of the array bounds before then. We run into a similar situation when excessive tower motion causes an analysis point to leave the grid in the lateral direction. We figured the worst-case scenario would be when a teetering turbine had a 90-degree teeter angle and and a 90-degree yaw error such that the blade tip could be the rotor radius plus the overhang to the side of the tower. In TurbSim, we currently recommend that you set the grid size to be twice that sum.

At the time I did this (which was nearly 15 years ago), we decided it was better to make the turbulence field a little bigger to avoid these problems instead of extrapolating outside the grid or simply using the outermost values.

Enter the offshore turbine.

Some turbines may be mounted on floating platforms that are moored using catenary lines that allow the platform to move around a bit. The motion may be tens of meters. To accommodate that much motion would require that we make the turbulence fields significantly larger and we would lose spatial resolution.

One possibility is to use the outermost value when an analysis point leaves the turbulence field. Doing so would allow us to use smaller grids. We won’t even have to accommodate the entire rotor diameter just enough to get to the outermost analysis points, which are usually not at the tip. However, with the possibility of so much motion for floating platforms, significant portions of the rotor may lie outside of the wind field.

My proposal today is to do things a little differently. I’d like to pin the center line of the wind field to the apex of the cone of rotation. That means that the wind field will move around in inertial space. If we ignore this issue, the only problems we would have remaining are at the start and end of the simulation where we could have analysis points outside the wind field. That’s no big deal at the start of the simulation because we usually throw away the first 10-30 seconds anyway. Only if we had a large yaw error would the amount of time of using constant wind be significant anyway. A large yaw error will cause constants winds at the end of the simulation, but in TurbSim we can automatically add a few seconds (R/u) to the wind field to accommodate the extra time we may need at the end of the simulation when we are recording the data. We are currently adding twice that much.

One thing I like about doing it this way is that the winds at T=0 are the ones that are actually affecting the rotor. Our current scheme has it so that (for an upwind turbine) the wind-file time affecting the rotor at the start of the simulation is the (overhang+0.5*gridwidth)/u. For a 100 m rotor with a 2 m overhang in a 5 m/s wind, that’s 10.4 s.

Addressing the issue of allowing the wind field to move around in inertial space, it seems to me this whole business of using stochastic winds is just a way to excite the rotor in a semi-realistic way. I don’t think anyone believes our simulated winds look anything like real winds.

So, I have two questions:

1) Do you thing it’s OK to use tighter grids and instead of stopping the program if you have a point outside the wind field, just use the outermost value?
2) Is it OK to pin the wind field to the rotor instead of holding it fixed in inertial space?


1 Like

I vote no. My reasoning is that if you fix the flowfield to the turbine motion you will lose any induced effects arising from local coherence in the flow. In effect the turbine will be operating in a “frozen turbulent field” instead of being able to spatially and temporally sample the flow as it moves through. Should the flow contain transient coherent elements (as it does in the coherent option is turned on in TurbSim) some of the short period impact will be lost. I really think you do not want to do that.

Regarding a point outside of the simulated wind field. I would vote no. I think that if the program stops due to this situation, one should first evaluate the scale of the turbine motions that caused it to see if it was the result some sort of computional instability or a realistic prediction based on the nature of the inflow dynamics. If it is the latter the grid should be expanded in order to fully document what is happening. By overiding this error, cases that need closer examination may be missed.

I voted against both of these changes. It was a wishy-washy vote against using points at the edge of the field when actually outside the field. It was a strong vote against pinning the wind field to the hub. I agree with Neil’s comments about his negative votes.

I do think that TurbSim generates a wind field quite similar to real wind. I believe the spatial variations are as important as the temporal. It is preferable to have a field large enough to cover all rotor motions, even if this means using a coarser grid. Of course, the ideal solution is a large and fine grid, but computational limits get in the way of that idea.

I’m not sure how literally Marshall intends to tie the wind to the hub. I certainly hope and expect that he did not mean that the relative wind due to hub motion would be lost. That would eliminate some important aerodynamic damping. However, moving the field with the hub while retaining the relative wind terms would still have problems. A real rotor will move into different winds as it wanders around the array. The simulated rotor should be allowed to do the same thing. It seems possible that there will be circumstances (random combinations of weird events) where a rotor might be deflected toward a higher or lower loading condition by the turbulent wind field. Instantaneously moving the wind field with the hub will lose most of that effect.

The basic idea of allowing points to go outside the field would be to allow us to use the radius of the outermost analysis points as the grid halfwidth. This would allow us to have a higher density of grid points. Normally small deflections would cause analysis points to slightly leave the volume and I think the approximation would be small. I think the larger perturbations Neil mentioned will show up in other ways such as supersonic velocities. We could also set a limit for the excursions.


Of course not. What I meant was that the point that was currently aligned with the undefelected tower at hub height at a given time in the current scheme would be located at the current hub location in the new scheme.

The point I’m trying to make with this is that I do not believe that TurbSim generates wind fields that look anything like real winds. This is just a method people have developed that will excite the turbine in ways that are statistically similar to the way real turbines are exited. There are no vortices in TurbSim’s background turbulence. Hell, Neill will tell you how unrealistic the neutrally stable spectra used in the IEC standard are.

Considering all the other gross approximations and limitations of our end-to-end simulation effort, I don’t think we will lose much with this approximation. Considering we will gain grid density in exchange, I think it is worth considering.


I guess it really depends on how far outside the field the turbine is going to move.

If the distance is large, I’d propose having the possibility of variable spacing. It seems that by adding tower points to the wind field (is this still being proposed?), we are kind of doing this anyway. This is what would be done in CFD, where the most likely positions of the turbine would be finely gridded up and the surrounding field coarsely gridded. Of course, if the turbine wanders into the coarse area, you probably will have to create another wind field anyway to maintain accuracy. Which means that having the simulation stop as it is currently written would be similar - i.e. you have to rerun the turb sim. But, having the coarse grid would enable you to see roughly where the turbine moves and could help you in your refinement of the grid. Not sure if the equations for the turbulent spectra could handle this variable spacing, but I think it would improve the tool.

If this distance is small, how about extrapolating using some kind of spline or POD method?

My $.02…


I am a graduate student of Shiraz University and I really need stochastic wind speed file (*.wnd) for testing my LMI based controller at hub height 80 m.
I want to test my controller in above rated wind speed conditions at that height.
Considering the fact that I recently join this forum, I don’t know in which relevant topic i should ask about that.

Dear Ehsan,

You should use TurbSim to generate the winds. You can get it here:

Many Thanks for your fast responce.