Hello!
I’m having issues matching up to the OC3 LC5.2 and 5.3 PSD calculations with my simulations (using another software package). My frequencies look good, but my magnitudes are 2-3 orders of magnitude higher than the results listed in the OC3 results while my time-series results actually show less amplitude. I’ve made sure to match up to the 0.0125 timestep used in 5.2 and 5.3 (can anyone verify that?) but I’m still too high. Can someone point out the method used to determine the PSD in the OC3 results? Was the data normalized against something?
Thanks!
Sean Quallen
Dear Sean,
In OC3 Phase IV, we didn’t prescribe the time step or the post-processing tool. However, in my FAST simulations, I used a time step of 0.0125 s.
A PSD is normally defined such that the integral of the PSD across all frequencies equals the variance of the time series (standard deviation squared); this is how the PSD was used in the OC3 project. My guess is the FFT used by your PSD tool uses a different normalization scheme.
Best regards,
Jason,
Thanks! I’m fairly certain my issue has to do with normalization inside the ‘black-box’ I’m using for my FFT as well. Much appreciated!
Sean Quallen
Jason,
I hope addressing you directly in this forum is acceptable. I found the issue with my PSD calculation and I’m now matching up reasonably. I do have one (hopefully final!) question about the NREL OC3 5.2/5.3 data. Did you resample this data to get the bins presented (0.005, 0.01, 0.015…)? Did you reshape the resulting PSD to fit into that? My frequency bins don’t match up evenly to those presented so I’m concerned about presenting those.
Thanks!
Sean Quallen
Dear Sean,
In OC3 Phase IV, we only suggested a resolution and range of the PSD – the exact frequency bin size and max frequency were not prescribed and everyone submitted their results there own way. If your PSD has a much finer resolution and is “noisy” as a result, you could bin the data into larger frequency bins to “smooth” the result.
Best regards,