turbulence intensity from TrubSim vs Bladed

[Jian.Li]

I compared the turbulence intensity at hub center with TurbSim Kaimal and Bladed Kaimal. It seems that the Bladed’s longitudinal statistical TI is closer to the “Characteristic value of hub turbulence intensity” defined in IEC 61400 ed.3
For example, the longitudinal statistical TI from TurbSim is 19.0494%, and 21.684695% from Bladed, comparing the 21.9556% by IEC61400 ed.3 for class A at 9m/s.
I tried with different random seeds. Bladed is only affected a little while TurbSim varies obviously.
Is this TI varying in TurbSim accepted?

[Bonnie.Jonkman]

The output TI is a topic we have discussed briefly in the TurbSim User’s Guide (for v1.50, see page 8 under parameter ScaleIEC) and at the TurbSim Workshop held in 2008. Here is what Neil Kelley (now retired) wrote about the problem in 2007:

@Neil Kelley

These results are indicative of the limitations of the Monte Carlo stochastic simulation… The level of turbulent variance included in a TurbSim 10-minute simulation represents a truncation of the frequency range under the spectral curve defined by the IEC Kaimal model. This is due to the length of the record at the long wavelength (low-frequency) end and by the length of the time step on the short wavelength (high-frequency end). The result is usually less total variance (turbulent kinetic energy) in the simulation which is modulated by the random phase variate distribution associated with a given simulation realization…

We can modify the code to ensure that there is a match at least in the TI values as our European colleagues do in their codes. The downside is that this amounts to an artificial level shift (upwards) in the overall turbulence level only for the frequency range defined by the record length and time step specified in the simulation. It does not add any energy at turbulent wavelengths (eddies) that are longer or shorter than the truncation limits set by the record length and time step. Thus there will be more energy within those limits than actually defined by the spectral model but it will meet the target TI value. This is the same thing as specifying a slightly higher value than the target TI in order to achieve a simulated value one closer to it.

The following slide from the 2008 TurbSim Workshop shows the spectra and variance in TI due to random phase variate distribution that Neil mentions:

We have added the ScaleIEC parameter in the TurbSim input file so that the output TI matches the requested values. I personally don’t like to use this feature: It seems like doing numerical integration where you know the analytical solution, and then when you see the numerical and analytical answers don’t exactly match, you scale the function you are trying to integrate instead of changing the numerical method. However, some people are very adamant that the simulated TI be exactly a certain value.

[Jian.Li]

Thanks for your reply Bonnie.
It looks to me this is an dilemma with the coherence structure and turbulence intense for every grid point.
When I use non IEC model and Coherent Turbulence Scaling Parameters together, can I write the turbulence and coherent structure composed wind to .wnd file? Or only the turbulent wind will be written to .wnd file?

[Bonnie.Jonkman]

The coherent structures are not part of the .wnd file. They are superimposed on the turbulent winds in the .wnd file when the files are read in AeroDyn/InflowWind.

[Jian.Li]

Thanks for the information.