Power Spectrum Density Discrepancies in TurbSim

Wind & Water Computer-Aided Engineering Software Tools
1

Hi everyone,

I’m using TurbSim (OpenFAST v4.1.2) to generate low- and high-resolution .bts wind files for FAST.Farm. In the low-resolution TurbSim input, I specified IECKAI as the turbulence model.

As validation, I extracted the hub-center velocity time series from the resulting Low.bts file and computed a power spectral density using Welch’s method. I then compared that PSD against the theoretical Kaimal spectrum used to generate the turbulence field.

The comparison shows noticeable disagreement starting at very low frequencies (below roughly 0.02-0.03 Hz). One explanation for discrepancies at higher frequencies is that the low-resolution field was generated with a timestep of approximately 0.9 s, corresponding to a Nyquist frequency of about 0.56 Hz. Therefore I would not expect the low-resolution signal to contain meaningful information much above 0.5 Hz.

However, what puzzles me is that the same low-frequency spectral discrepancy also appears in the high-resolution turbulence boxes. These are generated using TurbSim’s TIMESR option from a hub-height time series extracted from the low-resolution field, and the high-resolution boxes use a timestep of 0.05 s (Nyquist ≈ 10 Hz). The imposed time series and the resulting high-resolution hub-center signal correlate reasonably well (correlation ≈ 0.89), and the spectral shape appears to carry the same low-frequency deviation through the pipeline.

image
image734×513 76.3 KB

I’d like to ask:

  1. Is this level of disagreement between a Welch PSD and the target Kaimal spectrum normal for TurbSim-generated fields?
  2. When validating TurbSim output against a Kaimal spectrum, is there a recommended normalization or spectral formulation that should be used?

Thank you in advance!

2

Dear @Dalia.Sebat,

The TIMESR feature of TurbSim derives the power spectra of the target wind field from the user-provided time series, so, if the time series has a PSD that is filtered out beyond 0.5 Hz, so will the TIMESR-derived wind field, even if that wind field is generated with a smaller time step. So, when generating a high-resolution domain from a low-resolution domain with the TIMESR feature of TurbSim, either the low-resolution domain must use a smaller-than-desired time step (matching that of the high-resolution domain, to increase its frequency content) or higher-frequency content (following the desired spectra) must be added to the time series (including increasing its time resolution) before the TIMESR TurbSim analysis.

Regarding your specific questions:

  1. Your PSD matches the target spectrum as I expect it should, given the time resolution you have used.
  2. I believe you have normalized the PSD consistently with the target spectrum given the good match you are showing over the frequency band you are using.

Best regards,

1 Like