Dear @Wonjun.Choi,
From your plots, my guess is that in the OpenFAST simulation, nonzero wave forcing at time zero “kicks” the system and given the low damping, it takes a long time for the initial transient resulting from this “kick” to dampen out. It also looks like the OrcaFlex model gets around this by ramping up the wave elevation over the first 200 s, which is not a feature supported within OpenFAST’s SeaState module (without a source code change).
A few comments:
- Your ElastoDyn model has the platform and tower degrees of freedom (DOFs) enabled. If you look at the mean response of the platform displacement in all 6 DOFs and tower fore-aft and side-side deflection after all start-up transients die out, and set these as initial conditions for a subsequent simulation, does the length of the start-up transient reduce?
- Your OpenFAST model has quite small damping, limited to radiation damping in HydroDyn, tendon drag and structural damping in MoorDyn, and perhaps tower structural damping in ElastoDyn. Not having hydrodynamic viscous drag is likely unrealistic, and by adding it, the damping level would like be more realistic, which should reduce the start-up transient.
- Adding aerodynamic loads by enabling AeroDyn will also add damping that will reduce the start-up transient.
Best regards,
1 Like
Hello, dear Dr.Jonkman,
First of all, I appreciate your kind and thoughtful response.
Based on your comments, I investigated the following three points.
[Tower bending effects]
- I examined TTDspFA (Tower top fore-art) and TTDspAA (Tower top side-side).
- The mean tower-top deflections are very small.
[Morison viscous drag effects]
- When Morison viscous drag was included, the dynamic convergence time was reduced to around 1,000 seconds.
- I looked into the basis of the Cd values used in the glue code, and they appear to be consistent with typical values from API RP 2SK. However, since they are not based on CFD or experimental data, I am unsure about their suitability.
[Aerodynamic effects]
- As you mentioned, including aerodynamic effects reduced the convergence time.
- However, the RNA is modeled as a rigid RNA, so aerodynamic effects were excluded.
Thank you again for your helpful guidance.
Best regards,
WonJun CHOI
1 Like
Hello, dear Dr.Jonkman,
First of all, thank you very much for your continuous help.
I have completed the comparison analysis with the RNA modeled as a rigid body.
As the next step, I am now performing simulations under operating conditions, with the turbine rotating.
I have a question regarding the coordinate systems of the OpenFAST NREL 5MW turbine.
Based on the FAST User Guide and the input file, I created the figure below.
Could you please let me know whether my interpretation of the global, nacelle/yaw, shaft, and hub coordinate systems is correct? (NacYaw = 180° and Wind PropagationDir = 180°)
Best regards,
WonJun CHOI
Dear @Wonjun.Choi,
For the situation where NacYaw = 180degrees and PropagationDir = 180degrees, I agree that the wind vector would now point along negative X and an upwind rotor will remain upwind. However, the local coordinate system of the nacelle, shaft, and hub, which are normally nominally aligned with the global coordinate system, should now be 180degrees misaligned with the global coordinate system. That is, the nacelle, shaft, and hub coordinate system should have their x and y axes flipped 180degrees (about z) relative to what you’ve shown.
Best regards,
Dear Dr. Jonkman,
Thank you for your detailed explanation. 
I understand your point regarding the 180degree misalignment of the nacelle, shaft, and hub coordinate systems relative to the global coordinate system.
[Modified]
Best regards,
WonJun CHOI
Dear Dr. Jonkman,
I have a question regarding the controller DLL used for the NREL TLP model.
- In Matha’s paper, Modal Development and Loads Analysis of an Offshore Wind Turbine on a Tension Leg Platform, with a Comparison to Other Floating Turbine Concepts, I found the following statement:
- “Additionally the original baseline controller used with the land-based and TLP system is modified for the OC3-Hywind system.”
- I am currently using DISCON_OC3Hywind.F90 for analysis of operating condition. Is this controller appropriate for the NREL TLP case, or should I use the original NREL 5-MW baseline controller instead? (land-base, DISCON.F90)
Best regards,
WonJun CHOI
Dear @Wonjun.Choi,
When we’ve run OpenFAST simulations with the NREL 5-MW baseline wind turbine atop the MIT/NREL TLP, we’ve used the original DISCON.dll controller for the land-based and fixed-bottom offshore turbines. The DISCON_OC3Hywind.dll controller was developed for the OC3-Hywind spar, and also used for the OC4-DeepCwind semisubmersible due to the low natural frequencies in surge and pitch for these models.
Best regards,
1 Like
Dear Dr. Jonkman,
First of all, as you mentioned, I used a controller designed for land-based use.
I would like to ask about the response after applying the original baseline controller to the NREL 5MW turbine on the NREL TLP.
[Process]
-
To check whether the response is related to the controller operating, I performed steady-wind simulations for three wind speeds:
8m/s: below-rated
10m/s: near below-rated
12m/s: above-rated
-
wave condition follows:
Hs 4m
Tp 14s
gamma 1
wave direction 180deg
[Results & questions]
- The rotor speed and generator speed remain close to the rated values for all three wind speeds. The generator power also remains close to the rated power, although the fluctuation level changes with wind speed.
- However, the surge motions in the 10m/s and 12m/s cases increase with time.
- [Q1] Could this surge growth be caused by the baseline pitch controller near the rated operating region, or is it more likely related to my TLP model settings or initial conditions?
- [Q2] Is this the right way to check if the controller is operating properly?
Are there any other results I should consider?
I am still studying the aerodynamic and controller-related details, so I would appreciate your guidance if I am missing something basic.
Best regards,
WonJun CHOI
.fst
ElastoDyn
SeaState
ServoDyn
AeroDyn
InflowWind
Dear Dr. Jonkman,
I conducted a simulation based on your advice.
- I checked the controller behavior for each wind speed condition around the rated wind speed. As shown in the attached plots, the pitch controller starts to operate at a wind speed of 12 m/s, and the generator power appears to remain within a reasonable range for each wind speed condition.
- My question is about the surge motion at 12 m/s. The surge response does not seem to diverge with time, but its magnitude is quite large. Since I have already checked the model behavior up to the condition just before the pitch controller is activated, I suspect that the large surge motion may be related to the controller.
- As you mentioned previously, would it be reasonable to adjust the pitch controller gains in this case? Or are there any other settings that I should check first?
Thank you again for your help. I am still learning about aerodynamics and do not have a strong background, so your guidance is very helpful.
Best regards,
WonJun CHOI
[AeroDyn]
Dear @Wonjun.Choi,
The response you are showing at 12 m/s indeed looks like a controller-induced instability of the platform-surge mode, reaching a limit-state oscillation due to nonlinearities within the model.
I agree that reducing the controller gains to a target natural frequency below the surge frequency should eliminate this instability. That said, I don’t recall this being a big problem for the MIT/NREL TLP, so, perhaps the issue is limited to a very specific wind-speed band, which you are focusing on for this test. Do you see similar signs of an instability if you switch from a steady wind speed at 12 m/s to a steady wind speed at 14 m/s? Also, do you see a similar instability if you switch from a steady wind speed at 12 m/s to a full-field turbulent wind inflow with a mean wind speed of 12 m/s?
Best regards,
Dear Dr. Jonkman,
Thank you for your help. I conducted a simulation based on your advice.
-
First, I tested the surge instability under a steady wind speed of 14 m/s and a full-field turbulent wind inflow with a mean wind speed of 12 m/s. In both cases, the surge response still showed a growing oscillation.
-
Following your suggestion, I adjusted the controller gain to target 0.08 rad/s, which is lower than the surge natural frequency (0.1 rad/s). The modified gain values are shown in the attached figure.
-
After reducing the gains, the surge response was improved compared with the original controller. However, a relatively large surge response still appears in the 2000–2500 s range. In addition, the generator power occasionally drops close to zero, which seems abnormal to me.
-
Could this indicate that the controller-gain reduction was not applied appropriately, or should I also check other parts of the controller?
Best regards,
WonJun CHOI
[TurbSim.inp]
Dear @Wonjun.Choi,
I believe your results are expected. By significant reducing the pitch controller gains, the surge instability is eliminated at the expense of large generator speed and power fluctuations.
Best regards,
Dear Dr. Jonkman,
I understand that the reduced pitch-controller gains can eliminate the surge instability, but this comes at the expense of larger generator-speed and generator-power fluctuations.
- In the current comparison, my main focus is on the platform motions and tendon tensions, rather than the generator-power performance itself. In this case, would it be reasonable to accept the increased generator-power fluctuation as a trade-off, as long as the platform motion and tendon tension responses become stable?
- Or should I continue tuning the controller gains, or consider additional controller modifications, to find a case that satisfies both platform/tendon response stability and acceptable generator power regulation?
Ideally, I would like to satisfy both, but I am not sure what the appropriate approach would be.
Best regards,
WonJun CHOI
Dear @Wonjun.Choi,
You could trying varying the target frequency driving the controller gains to see if you can find a better compromise between acceptable surge motion and generator speed/power fluctuations. But you’ll have to decide what compromise is reasonable for your given project.
Other modifications, such as switching to the more advanced controls options available with the ROSCO controller (GitHub - NatLabRockies/ROSCO: A Reference Open Source Controller for Wind Turbines · GitHub) would require additional effort.
Best regards,
1 Like
Dear Dr. Jonkman,
First of all, thank you for your advice on the controller.
Your help was very much appreciated.
I have a question about determining tower mode shapes using BModes.
- Since the TLP has a freeboard of 9 m, I changed the tower height from 87.6 m to 78.6 m in order to set the hub height to 90 m.
- For verification, I compared the tower mode shapes from the original Tower_ElastoDyn file with those recalculated using BModes.
- As shown in the attached plot, the 1st mode agrees reasonably well, but the 2nd mode is significantly different. I found some similar posts on the forum, but I could not find a clear resolution.
Could this be an issue related to the 2nd tower mode in BModes, or is it more likely due to my BModes setup process?
The BModes input file and tower property file are attached.
Best regards,
WonJun CHOI
.bmi
.dat
Dear @Wonjun.Choi,
In BModes, the freeboard can be set through the setting of a negative-valued draft, so, you shouldn’t have to change radius to change the tower height in BModes. That said, the OpenFAST model of the NREL 5-MW baseline wind turbine atop the MIT/NREL TLP uses does not consider the freeboard, so, you could set draft = 0m and radius = 87.6m for this model. I haven’t confirmed the other values in your BModes input files.
Regardless, I’m not too surprised by your results. The tower mode shapes for the MIT/NREL TLP were originally derived through MSC.ADAMS, not BModes and the ADAMS model considered blade flexibility. The 2nd tower mode shapes are much more sensitive to blade flexibility.
Best regards,
Dear Dr. Jonkman,
Thank you very much for your detailed explanation.
- Your comments were very helpful for understanding the limitations of using BModes for the MIT/NREL TLP tower mode shapes, especially regarding the second tower modes and the effect of blade flexibility.
- As a next step, I will consider either using the existing ElastoDyn_Tower file by ignoring the freeboard and matching the hub height to 90 m, or trying MSC.ADAMS with a student license.
Best regards,
Wonjun Choi
Dear Dr. Jonkman,
-
To validate the tower mode shapes generated using BModes, I compared the OpenFAST model with an OrcaFlex model, in which the tower is represented using a finite element formulation. I performed both free decay and regular wave simulations, and the responses from the two models showed good agreement.
-
Based on your explanation, I understand that if differences between OpenFAST and OrcaFlex appear in future comparisons, one possible reason may be that blade flexibility was not considered when generating the tower mode shapes using BModes.
Thank you again for your help.
Best regards,
Wonjun Choi
1 Like
