Hello,
I m currently facing issues while simulating load cases with deterministic winds like DLC 2.3 (EOG with grid loss) at higher wind (i.e. 25m/s). The load timeseries accross the turbine are showing severe spikes driving the loading. I think these are numerical issues coming from AERODYN. I observe also spikes in the Cl values of blade node 13 to 15. Also when I switch the model to DBEMT, I can smooth out the spike to a reasonable level. Unfortunately the turbine behavior is changing a lot with DBEMT, making me a bit uncertain regarding the further use of DBEMT. Also I noted that a reduction of the time steps DT does not affect this issue.
Best Regards,
Florian
morewind engineering solutions GmbH
Dear @Florian.Stache,
I would suggest rerunning this case with the newest version of OpenFAST (currently v3.2.1). There have been several upgrades to AeroDyn since OpenFAST v2.2 that should improve its robustness for this shutdown situation, such as Updates to AeroDyn by bjonkman · Pull Request #538 · OpenFAST/openfast · GitHub.
Best regards,
Hi Jason,
Thank you for your hint.
We tested with v3.2.1. The spikes are still there. Anyhow not exactly same load cases and timing.
We are now adding additional outputs to Aerodyn to understand what is going on. Obviously it does not happen in normal operation DLC 1.1 but more after severe transients like in DLC1.4 or DLc2.1 with one blade stucked.
Best Regards,
Florian
Hi Jason,
we made a couple of trials and obviously the issue is in the dynamic stall model. When we set AFAeroMod=1 the spikes are disappearing. The DBEMT was only smoothing out the spikes but they still where there. In the next picture you see the comparison of full Aero Model on vs AFAeroMod=1
We tried further on blade and tower DOF freeze, tower upflow and shadow off and WakMod=0. All still had spikes and even sometimes aborted the simulation.
We did output Alpha and CL at different stations and we observe that Alpha is stable but CL is getting spikes at some sections. Anyhow the parameter of the airfoils in those section do not appear to be very different than the other sections. Here CL vs alpha for the AFAeroMod=2 in blue and =1 in red for blade node 15 at 46m BlSpn.
max value is not visible in the above picture but CL goes up to 355 and down to -117 within one time step.
Here the parameter of the airfoil file (steady curve looks fine):
10.000000 Re ! Reynolds number in
0 Ctrl ! Control setting (m
True InclUAdata ! Is unsteady aerody
!..
-2.744419 alpha0 ! 0-lift angle of at
14.299403 alpha1 ! Angle of attack at
-12.302492 alpha2 ! Angle of attack at
1.000000 eta_e ! Recovery factor in
7.659511 C_nalpha ! Slope of the 2D no
Default T_f0 ! Initial value of t
Default T_V0 ! Initial value of t
Default T_p ! Boundary-layer,lea
Default T_VL ! Initial value of t
Default b1 ! Constant in the ex
Default b2 ! Constant in the ex
Default b5 ! Constant in the ex
Default A1 ! Constant in the ex
Default A2 ! Constant in the ex
Default A5 ! Constant in the ex
0.000000 S1 ! Constant in the f
0.000000 S2 ! Constant in the f
0.000000 S3 ! Constant in the f
0.000000 S4 ! Constant in the f
1.929310 Cn1 ! Critical value of
-1.091417 Cn2 ! As Cn1 for negativ
Default St_sh ! Strouhal’s sheddin
0.011517 Cd0 ! 2D drag coefficien
-0.085619 Cm0 ! 2D pitching moment
0.000000 k0 ! Constant in the \h
0.000000 k1 ! Constant in the \h
0.000000 k2 ! Constant in the \h
0.000000 k3 ! Constant in the \h
0.000000 k1_hat ! Constant in the ex
Default x_cp_bar ! Constant in the ex
Default UACutout ! Angle of attack ab
Default filtCutOff ! Cut-off frequency
Best Regards,
Florian
Dear Florian,
Which UAMod have you enabled when AFAeroMod = 2?
Best regards,
We are using UAMod=3. This is the default setting on Github for the IEA10MW. We tried the other models as well but could not really eliminate the spikes. Anyhow I will give it another try as the first trials where with openFASTv2.2
Dear @Florian.Stache,
I’m not fully following. The results in your post dated Sep 7, 9:25 AM are with OpenFAST v2.2 or v3.2.1? If they are from the former, I would definitely see if upgrading to v3.2.1 resolves the issue.
Best regards,
No. we switched to v3.2.1 since your first comment from 26.08 and we did not resolve the issue as the results from today 07.09 show.
I was just saying in my last post, that we did also a study on the three different UAmodels before and I would just like to re-do it as this one was from before 26.08.
sorry for the confusion…
Dear Jason,
we now did run in v3.2.1 UAMod=2 and it appears stable. Here a comparison to the run with de-activated UA (AFAeroMod=1) (blue):
We will check if we can generalise to all the DLC where we encountered spikes.
The good old Oye Model (UAMod=6) is also free of spikes but shows less deviations from steady.
Would be interesting to know why IEA model was set to UAMod=3 by default on github.
Best Regards,
Florian
I don’t have a good explanation as to what is happening . I’ve sometimes observed spikes here and there in simulations but I never dived into them thoroughly.
In the examples you sent, the lift overshoots quite a lot (up to Cl=7), which clearly indicates that something is wrong with the UA model (the linear region is “too big” maybe).
I have two guesses:
-
the polar data is not that well represented by a linear region, which challenges the model because a lot of the model assumes a linear region, and uses the difference between the linear region and the polar data. I usually use pyDatView to open the polar files, and compare the different “Cn” curves that it computes, making sure the linear/potential curve is close to the regular Cn curve, and making sure that the 0,1,2 regions are close to the zero lift, and stall angles (see picture below)
-
The UA parameters in the airfoil file were not computed properly. I don’t know how the parameters of the IEA 10 MW were computed. You can try to use the following python script to recompute the unsteady parameters and see if they change (you can compare the new and old polars by opening both in pyDatView).
Experimenting with different UAMod=3,4,6 and comparing them is a good way to see what can be wrong. Another thing to try is to use the UA driver with the problematic polar file (at the radial station you plot), and try to reproduce oscillations of angle of attack similar to the ones you have in your time series. If you still see weird spikes, we can use this as a test case and dive a bit more into what the issue can be.
Keep us in touch,
Emmanuel