Simulation length for FOWT

[Simon.Wiedemann]

Hello,

I have a question on the needed simulation length for a floating offshore wind turbine with turbulent wind and irregular waves.

In DNV-OS-J103 Design of Floating Wind Turbine Structures 6.2.1 (p.41) it is stated, that simulations should be made for at least 3 hours and not only 10 minutes as it is common for onshore turbines.
But then in NREL/CP-5000-58153 (Simulation-Length Requirements in the Loads Analysis of Offshore Floating Wind Turbines) it is concluded, that 10 min duration is sufficient for transient simulations and that the method of treating unclosed cycles is more important (for fatigue).

In the later document some method of producing “Periodic Wind Files” is used.
Can anyone comment what simulation length and how many seeds should be used in order to capture also low frequent effects?
I am a little bit confused by that.

Thank you and best wishes,

Simon Wiedemann

[Jason.Jonkman]

Dear Simon,

At this time, I don’t think anyone can offer specific recommendations on the exact simulation length that is sufficient for loads analysis of floating offshore wind turbines (FOWTs). We have taken the results of our research project documented in NREL/CP-5000-58153 (and related publications) to write the section on simulation requirements for the draft IEC Technical Specification (61400-3-2) for the design of FOWTs (this TS is currently under review, hopefully to be published later this year). Here is what is relevant text written regarding this topic (from section 7.5.6):

Within each bin, it is important to perform the loads analysis with an appropriate number of adequate-length simulations to ensure the statistical reliability of the calculated structural loads. The appropriate number and length of simulations shall be determined for each DLC based on the FOWT support structure and the site-specific offshore conditions, but should not be less than those specified in IEC 61400-3-1 clause 7.5.6. The IEC 61400-3-1 recommends 10-minute simulations for most DLCs, with at least 6 random wind and wave seeds, resulting in 60 minutes of stochastic wind and wave inputs for each environmental condition. The 10-minute simulation length is based on the spectral gap of wind variation, which occurs between the turbulent and diurnal peaks in the wind spectrum. Ten minutes of turbulent wind can be approximated as stationary within this frequency band. Similar reasoning in the offshore oil and gas industry has led to common practice of applying 1 to 6 hours per simulation for floating systems to account for the spectral gap of waves at a lower frequency, the low natural frequencies of floating sub-structures, and second-order slow-drift hydrodynamic effects.
Simply running longer wind turbine simulations may not be satisfactory. Turbulent wind simulations often assume a stationary wind condition, so, the turbulent wind generated with simulation times much longer than 1 hour is unphysical. Also, generating turbulent wind data with adequate spatial extent, adequate spatial and time resolution, and simulation times much longer than 1 hour is too computationally expensive for most computers to generate and store. Moreover, increasing the simulation length introduces additional stochastic information (and larger extremes) that itself will result in larger ultimate structural loads, independent of offshore considerations.
To avoid these wind data problems, the use of repeated periodic wind data is recommended if FOWT considerations necessitate running simulations much longer than 10 minutes. Turbulent wind data generated through Fourier-transform techniques are periodic with a period equal to the length of the dataset (typically 10 minutes). This periodic wind data can be successively repeated for simulations involving combined wind and wave excitation longer than 10 minutes using wave data based on the total simulation length (up to 6 hours). But it should be ensured that the use of periodic wind data does not excite low-frequency response of the FOWT.
For ultimate loads on FOWTs with negligible slow-drift effects, it is possible that the length of individual simulations need not be longer than 10 minutes as long as the number of simulations is sufficient to ensure the statistical reliability of the calculated structural loads. That is, it is possible that the same ultimate loads can be calculated using simulations of different length, as long as the total amount of random information in the stochastic wind and wave data is kept constant by varying the number of simulations. An assessment of the simulation-length requirements may require one to compare ultimate loads between simulations of different length. To compare ultimate loads between simulations of different length, the averaging technique is important. One should either compare the ultimate load from the same total simulation length, or divide the longer simulations into the length of the shortest simulation and compare the average maxima.
For fatigue loads on FOWTs, it is possible that there is greater sensitivity to the method of counting unclosed cycles compared to the simulation length—see clause 7.6.3.
In FOWTs the range of support-structure frequencies may be considerably lower than for fixed systems. Therefore, in order to capture potentially extreme loads or enough fatigue cycles during the transients, the simulation length for start-up and shutdown events may need to be increased for FOWTs.

And from section 7.6.3:

Unclosed cycles play an important role in fatigue analysis of FOWT, particularly for shorter simulations. Unclosed cycles, also called half or partial cycles, are generated by rainflow-counting algorithms when peaks cannot be matched with equivalent but opposite amplitude valleys. Unclosed cycles are created at the beginning and end of time-domain simulations, and for large amplitude cycles. A weighting factor between zero and one is applied to these unclosed cycles when the final damage is calculated. If a weighting factor of one is used, each unclosed cycle is treated as if it was a full cycle, and if zero is used, the unclosed cycles are disregarded, having no effect on the fatigue calculation. A factor of 0.5 is commonly recommended as a compromise. It is possible that there is greater sensitivity in the fatigue loads to the method of counting unclosed cycles compared to the simulation length. To minimize this sensitivity, the fatigue-counting algorithm should be processed with all of the simulations from each bin concatenated into one dataset instead of processed separately.

I hope that helps.

Best regards,

[Simon.Wiedemann]

Thank you for clarification!
Is there in Turbsim an option to calculate periodic wind files?

Best regards,

Simon

[Jason.Jonkman]

Dear Simon,

The ability to generate periodic wind files was introduced in TurbSim v2.00 and newer, available from here: nwtc.nrel.gov/alphas. Simply set TurbSim input UsableTime to the string, “ALL” and run TurbSim to generate periodic wind data files useable by FAST.

Best regards,

[Dawn.Ward]

Hi Jason,

I have been reading a paper that you are a co-author of, ref OMAE 2013-11397 ‘Simulation length requirements in the loads analysis of offshore floating wind turbines’.

Within that it states that 10 minute periodic wind files were used to generate the simulations, some 6hr, 3hr, 1hr, 20min and 10mins.
It also states that there are 10 wind seeds for each 100 min wind bin.

Also in the information you provide in this thread it says:

[i]

[This periodic wind data can be successively repeated for simulations involving combined wind and wave excitation longer than 10 minutes using wave data based on the total simulation length (up to 6 hours).]

[/i]

However I am a little confused on how using a periodic wind file for say a 1 hour simulation can achieve this? Is it not the same as running 6 x the same 10 min simulation, such that you only have one representative 10 min wind file? Am I missing something fundamental in the way the periodic wind file works or in the requirement for 10 minute wind seed representations for load analysis?

Thanks in advance for your help,
Regards,
Dawn

[Jason.Jonkman]

Dear Dawn,

Sorry, but I’m not sure I understand your question.

Running a 1-hour simulation with periodic 10-minute wind data means that there are 6 successive repetitions of the same 10-minute wind inflow, but other model parameters are likely different during the 1-hour simulation, including the incident wave elevation and structural response. For floating wind turbines, which tend to have platform natural frequencies that are very low in some modes of motion, there may only be a few oscillation cycles of these modes in a 10-minute simulation, but many more oscillations in longer simulations (hence the desire to check if the simulation length makes a difference versus simply running more 10-minute simulations).

Best regards,

[Dawn.Ward]

Hi Jason,

Sorry, I worded that very badly. My question was what do we gain from a longer simulation when we are just repeating the same 10min wind profile, which you answered perfectly

Many thanks,
Dawn