Results comparison FAST and Bladed

Wind & Water Computer-Aided Engineering Software Tools
1

Hi,

I used a simple EOGR+0.0 wind file to test the model by using FAST. The wind model is 12 m/s for the first 20 seconds, then begins a gust for 10.5 seconds duration, after gust the wind keeps 12 m/s till end (50 seconds).

In the first 20 seconds, pitch angle and power keep nearly constant values( a horizontal straight line) for Bladed from 0 s on. In FAST results, the power increases from 0 till the rated power and keeps small fluctuation till 20 s (not so smooth as that in Bladed). The power is always greater than the rated value like 1%-8%, at gust position even 16%. The pitch angle is activated twice in the 20 seconds period, means it has two big peaks before gust occurs, that is different from Bladed( the pitch does not change for the 20 seconds).

The curves are 90% same around the gust position for power and pitch. Only difference is pitch angle 1 or 2 degrees. Maximum power is 16% bigger than rated power.

I can not understand the power most time is above rated value. Could you give some suggestions? Some parameters for primary input file are not correct?

Thanks

Best regards

Yinping Yang

2

Dear Yinping Yang,

The variation in the FAST output for the first 20 s of your simulation is likely related to the specification of initial conditions that are not close to the steady-state solution. Are you setting the same initial conditions in the FAST and GH Bladed codes?

The aero-elastic theories are a bit different between FAST and GH Bladed, which gives rise to differences between the codes’ outputs. We have demonstrated this in many code-to-code comparison projects. For example, one difference we know of is that FAST calculates/applies aerodynamic forces orthogonal to the deflected blade, whereas GH Bladed calculates/applies aerodynamic forces orthogonal to the undeflected blade regardless of deflection. So, I’m not surprised you are seeing difference between the codes. Of course, a portion of the differences you are seeing could also be the result of different input specifications between the codes.

Best regards,

3

Hi,

I asked the colleague and the initial condition should be same (all parameters in Initial Conditions are set to 0 except RotSpeed). I changed the GBoxEff from 100% to 95%, GenEff from 100% to 93.6% . The power seems better below rated power. Bladed uses table for both GBoxEff and GenEff. If in FAST, these two values are 100%, are they the reason that the electrical power is always bigger than rated power. It seems the controller does not work. Or maybe the non steady-state condition at the beginning could also cause the relative fluctuation above rated electrical power.

If the time is long enough before the gust occurs, could the turbine get close to steady state condition?
Like the wind lasts 60 seconds and then gust comes. I tested in this case the model. The pitch angle’s fluctuation decreases till closely constant as time proceeds. Then the machine is suddenly shut down during the gust. Could small angle violation cause this case?

How to get close to the steady state condition for the initial specification directly like in Bladed? Set OopDefl, INPDefl, TTDspFA, TTDspSS and Azimuth to the values to reach the initial steady condition? Like OopDefl=-5m and INPDefl=1m and so on. Or just long time proceeds till steady.

Could we find the reports of code-to-to comparison projects from our NREL Publication website?

Thanks

Best regards

Yinping Yang

4

Dear Yinping Yang,

Are you saying that all of the intial conditions for the structural states are the same between FAST and GH Bladed? What about the controller states, such as blade pitch angle? Regardless, the start-up transients may differ between the codes, depending on the differences in theory between them. It is always wise to allow the simulation start-up transients to die out before outputting data and comparing codes.

In FAST, the start-up transients may take 30 seconds or more, depending on the choice of initial conditions, the natural frequencies in the system, and the controller settings. If the wind-inflow is steady, the system will eventually reach a steady-state condition if there is positive damping in all system modes. Setting proper initial conditions for OoPDefl, IPDefl, TTDspFA, TTDspSS, etc. can be used to reduce the length of the start-up transient, but it is typically more convient to just run the simulation longer and only output data after the start-up transients die out.

If the efficiencies FAST is using are less than the efficiencies GH Bladed is using, than certainly the power outputs predicted by the codes will differ. GenEff in FAST is used to calculate how much less the electrical power is compared to the mechanical power in the high-speed shaft (when generating power; this is reversed for motoring during generator start-up of the turbine). GBoxEff in FAST is used to calculate a friction torque in the drivetrain, which also reduces the amount of electircal power (when generating power; this is reveresed for motoring). The intial transients in the model–and anything that causes unsteadiness–can certainly lead to fluctuations in the power.

I don’t know enough about what you are doing to answer your questions regarding “electrical power is always bigger than rated power / controller does not work”.

FAST’s small angle assumption/violation will not trigger a turbine shut down. (For information regarding the small angle assumption/violation, please refer to this post: http://forums.nrel.gov/t/nrel-5mw-controls-dll-interface/183/1.) If the machine is shutting down during a gust, this must be initiated by the control system. What controller settings are you using?

The following papers show some of the latest results in which FAST, GH Bladed, and other codes are compared side-by-side. These papers were written under the IEA Annex 23 Offshore Code Comparison Collobaritive (OC3) project:

nrel.gov/docs/fy07osti/41930.pdf
nrel.gov/docs/fy08osti/42471.pdf

Best regards,

5

Hi,

The structural initial states are the same between FAST and GH Bladed. Also they use the same DLL controller file for the model. I think the controller states should also be the same. ( I compiled the FAST source codes with BladedDLLInterface.f90, in which the parameters of module BladedDLLParameters are set to meet the model. ) You mean max. and min. Pitch angle ? Then the FAST.exe will call the DLL controller and its setting automatically.

I ran a simulation for 50 seconds with EOGR wind condition at 30 second. The transients during the first 20 seconds are strong. The electrical power is always bigger than the rated value. Take 5 MW onshore machine for example, the power is most time between 5.1 MW and 5.5 MW. I understand that power should always strictly be under 5 MW level with the help of control system. So I got the conclusion that the controller does not seem to work efficiently. Or the controller did not keep the electrical power under the 5 MW. If it is bigger than rated power, maybe the generator will be damaged. After the transients die out, the power is strictly under 5 MW.

Thanks for your link of the two reports.

Best regards

Yinping Yang

6

Hi Yinping Yang,

I didn’t mean the min/max blade-pitch angles. I meant the initial blade-pitch angles (at time zero). Regardless of whether or not FAST is configured to use GH Bladed-style DLL controllers, the intial rotor speed and initial blade-pitch angles in FAST are specified within FAST’s primary input file. The DLL controller may or may not use these to initialize its own internal states. I can’t comment on how the DLL controller initializes its own internal states because I don’t know what DLL you are using. Neither can I comment on how GH Bladed initializes its rotor speed and blade-pitch angles.

The DLL controller that I wrote for the NREL 5-MW wind turbine does not allow the electrical power to exceed 5 MW, regardless of the conditions. Is this the controller you are using? If so, perhaps you are not checking the correct output parameter? What output parameter in FAST are using to check the electrical power? Output “GenPwr” is the only FAST output that gives electrical power. The other power outputs–such as “RotPwr” or “HSShftPwr”–are mechanical powers (mechanical torque times shaft rotational speed), which can exceed the electrical power.

Best regards,

7

Hi,

The initial pitch angles are set to 0 deg. The controller was written by a company and I can not know the algorithms inside.

I tested a EOG wind profile for 5MW onshore turbine. And the power is OK. Only when the gust happens, the peak of power is 5.24 MW. I understand that this happens in short time and it should not damage the generator of turbine, or the turbine has protection measures. Is this correct?

Best regards,

onshore.jpg
onshore.jpg1280×692 35.1 KB

8

Dear Yinping Yang,

I cannot comment much on your questions because I don’t know anything about your turbine, controller, or protection system. Many systems are OK with short and slight overloadings of the generator. But you should ask the developer of your controller why the controller permits overloading of the generator.

Best regards,

9

Hi,

I ran a simulation under constant wind speed 12m/s for 150s just for loads comparison. I compared the load results with Bladed. The results " RootFxb1, RootFyb1, RootFzb1" “RootMxb1, RootMyb1” are ok compared with those of Bladed. The " RootMzb1" is different from that of BLaded. “YawBrFxp, YawBrFyp, YawBrFzp, YawBrMyp” of tower are also different. I do not know the reason of the difference, especially for the tower. Maybe the tower is not modeled correctly. Or FAST’s own modelling of tower is different from GH Bladed and this causes the defference. The frequency of YawBrFxp is bigger. Maybe you can give me some suggestions.

Thanks

Best regards

Yinping

Fig10.jpg
Fig10.jpg1281×692 73.1 KB

Fig12.jpg
Fig12.jpg1281×692 91.8 KB

Fig11.jpg
Fig11.jpg1281×692 39.1 KB

10

The other figures are attached in this post.

Fig15.jpg
Fig15.jpg1281×692 45.7 KB

Fig14.jpg
Fig14.jpg1281×692 56.2 KB

Fig13.jpg
Fig13.jpg1281×692 40.1 KB

11

The last one.

Fig16.jpg
Fig16.jpg1281×692 55.8 KB

12

Dear Yinping Yang,

I don’t know anything about the turbine you are modeling or what your model settings are, so, it is difficult to know what is cuasing the differences you are seeing. As I said in a prior post, the aero-elastic models are different between FAST and Bladed, so, basic differences in their outputs are not surprising. With my quick skim through your plots, here is what I first notice:

*The fact that the axial force at the tower-top (YawBrFzp) is different between FAST and GH Bladed suggests that the models do not have the same rotor-nacelle assembly mass.
*It appears that FAST is predicting a slightly higher shear force and rotor torque than GH Bladed.
*With regards to the blade pitching moment, the mean of your FAST solution is near zero, but the mean of your GH Bladed solution is -45. Are you using the same aerodynamic offset and aerodynamic pitching moment data between the models?

I hope that helps.

Best regards,

13

Dear Jason,

Thanks for your helpful advice. I model a 2.5 MW wind turbine for only FAST time-marching simulation. I take the 5 MW onshore primary input file as my template. Turbine Control section is nearly not changed, only the Blade intial pitch is set to 0 degree for each blade. The same external controller of Bladed Style is used in FAST model. The controller uses its own text input file. Yaw DOF is deactivated and the other DOFs are the same as 5MW DOF settings. I changed only the rotor speed for the intial conditions. Turbine Configuration and Mass and Intertia and Drivetrain are set according to the model. Outputs are changed to meet the needs.

The blade mass is not equal to that of GH Bladed. I found this from the summary output file. I understand FAST integrates the Distributed Properties to obtain the mass of blade and tower. I adjust the AdjBlMs(factor to adjust blade mass density) to obtain the same mass as GH Bladed. Is there another way to correct this mass difference of blade ? Then the results of YawBrFzp agree better.

The aerodynamic pitching moment data are the same. Maybe the AeroCent parameter(aerodynamic offset ?) is not correctly calculated and given. I do not know how to get the aerodynamic center of each airfoil section. Is it always 0.25 or it changes also from section to section?

Could you please tell me which kinds of differences are basic differences? I will be sure they are not caused by some missing or false parts of my model. The tower has obvious different results. I do not know if the tower model is correctly modeled. For example, the YawBrFyp and YawBrMyp are totally different.

Thanks very much
Best regards

Yinping Yang

14

Dear Yinping Yang,

The FAST input “AdjBlMs” was added to support fine tuning of the overall mass, as you’ve done. You could of course be more sophisticated in adjusting the mass (e.g., by modifying the mass distribution to not only match the overall mass, but also the center of mass, moment of inertia, etc.), but this is more difficult.

I cannot comment on how to calculate the aerodynamic center for a given blade; I can only describe how the input in FAST is defined. To summarize, the quantity (AeroCent - 0.25) is the fractional distance to the aerodynamic center from the blade-pitch axis along the chordline, positive toward the trailing edge. So, setting AeroCent to 0.25 results in a pitching moment only from aerodynamic Cm data. When AeroCent is different from 0.25, not only does Cm influence the pitching moment, but the lift and drag forces also contribute to the moment due to the offset. See the FAST User’s Guide for a more detailed explanation.

It is hard to define “basic differences” in a measurable way. For examples of FAST and Bladed comparisons (along with results from other codes) that I’ve been involved in, you can see these papers: nrel.gov/docs/fy08osti/42471.pdf (both from the IEA Task 23 OC3 project).

When you changed the system masses/stiffnesses/geometry/etc. of the NREL 5-MW model in order to make your 2.5-MW model, did you also recalculate the mode shapes? The mode shape accuracy also contributes to the accuracy of FAST’s solution.

Best regards,

15

Hi,

In the summary file, the mass of blade calculated by FAST is about 10% less than that of GH-Bladed. But the results of the first and second mass moment of blade are nearly the same, the error is about less than 1%. I suspect the loss of mass is due to the discretisation into different segments of the blade, while GH-Bladed uses the continous blade shapes. When I change the “AdjBlMs” to adjust the mass to the same value like GH-Bladed calculated, the first and second mass moment increase about 11%. Does the increase of first and second mass moment influence the results, like loads of tower top ? If so, can I avoid this increase of mass moment when I match the blade mass?

I calculated AeroCent using the formular : AeroCent=0.25-[(fraction of chord from leading edge to actual pitch axis)-(fraction of chord from leading edge to actual aerodynamic center)] to convert from my notation to FAST’s notation. FAST has assumed the pitch axis passes through the 25% chord positon. GH-Bladed only uses the input of pitch axis, which is equal to 0.25(aerodynamic center)+(the fraction of chord from aerodynamic center to pitch axis). By applying the above formular, I got AeroCent=0.25-(the fraction of chord from aerodynamic center to pitch axis). These values are smaller than 0.25. The pitching moment is different from the result of Bladed. I changed the AeroCent to bigger vlues for all sections, like 0.45 bigger than 0.25. The pitching moment calculated by FAST is about -100 kN. If I adjust the AeroCent to make the result be -45 kN like Bladed, AeroCent will not obey the above formular. I am now confused with the discrepancy of pitching moment FAST and Bladed calculate differently.

The mode shapes of blade are calculated the same between Modes and GH Bladed. The mode shapes of tower are taken from GH Bladed. For Modes the frequencies agree better than the mode shapes . BModes produces smaller frequencies for the second mode, like only 1.3 Hz rather than 1.65Hz, with the fake inertial and mass offsets for generating the adams datasets. I added the point mass distribution into the input of Modes and it has no problem. I did the same for the tower input of FAST. I found the tower mass is greatly bigger than the result of Bladed. The summary file shows the tower is interpolated into different stations according to the nodes number. FAST user guide suggests 20 TwrNodes. I use 40. Is it allowed more than 20? I found the added point mass distribution is also taken into account for the interoplation. I think it is not correct to add the point mass distribution for FAST.

Could you tell me the reason of the difference of YawBrFyp between FAST and Bladed? The response of GH Bladed is like a beat phenomenon learned in the vibration course.

Thanks for your paper suggestions.
Best regards

Yinping Yang

16

Dear Yinping Yang,

In FAST, you could, of course, modify the blade mass distribution to not only match the overall mass, but also the center of mass, moment of inertia, etc. These will all influence the resulting dynamic response, but it is difficult to state this influence in general terms.

In the Modes, BModes, and FAST codes, the tower mass is defined as a distributed mass per unit length (varying along the length), which gets integrated along the tower in order to find the overall mass. None of these codes allow one to specify separate point masses. In FAST, 20 tower elements is typically suitable. You can specify more than 20 elements, but the code will become slower.

I cannot comment on how GH Bladed defines the aerodynamic offset nor can I comment on how it calculates the blade pitching moments. In FAST, the blade (bending and pitching) moments are calculated by integrating along the blade the ( applied aerodynamic pitching moments ) + ( radial vector ) x ( applied aerodynamic forces + gravitation forces - inertia forces ). All of these forces, moments, and distances are vectors and are oriented with the blade as the blade deflects. I know that GH Bladed ingores the influence of the blade deflection on at least the aerodynamic forces. Perhaps this is one reason why the pitching moments don’t agree between your models.

Neither can I comment on why GH Bladed predicts a “beat phenomenon” in YawBrFyp. I suggest you ask GH about this.

Best regards,

17

Dear Jason

I am trying to model the NREL baseline 5MW concept in Bladed.
Can I have the “prj” file of that?
Have you put that in the public domain?

Thank you
Mahdi

18

Dear Mahdi,

Neither I nor NREL support GH Bladed. I suggest you contact Garrad Hassan for help.

Best regards,

19

Dear Jason,

I have seen the abstract of your recent paper entitled " Offshore Code Comparison Collaboration within IEA Wind Task 23: Phase IV Results Regarding Floating Wind Turbine Modeling".
I could not find the full version, have you put that in the public domain for download?

Regards,
Mahdi

20

Try here:

http://www.nrel.gov/docs/fy10osti/47534.pdf