air density correction of power curves

[windmueller]

Today I compared the results of our air density correction and WAsP 12 (IEC extrapolation method) for different power curves. As we don't resample the values at original wind speed it seems okay that there are some differences in power below rated power for a calculated power curve.
But when I compared two measured power curves there was a change in rated power. Rated power was higher (in one case 10 kW and in the other case 17 kW) than in the original power curve (pitch) after air density correction with WAsP 12.

According to the "Proposal of an Improved Power Curve Correction" Approach 2 is used for pitch controlled turbines and is the IEC61400-12-approach. How ca it be possible that the rated power ist not the same after air density correction for pitch controlled turbines (for measured power curves)?

[Morten]

Hi Stefan,


The extrapolation is done in two steps. First we correct the wind speeds of the reference points of the original power curve. For a decrease in air density, this shifts the power curve to the right, but we take care of not moving the point at the cut-out wind speed.
We then interpolate the shifted power curve and find new values at wind speeds with regular spacing. I have seen one case where this interpolation failed because the interpolation function was forced into unrealistic undulations. The problem was an unconventional additional reference point near cut out inserted by the user. It did not use to disturb WAsP but fooled the air density correction method. If you like me to investigate the problem then please send a WAsP workspace file illustrating the problem to waspsupport@dtu.dk.


Best regards,
Morten

[Morten]

A note for third-party readers:


The assumption in my first reply was wrong. The real problem is that WAsP applies an automatic routine to detect whether a turbine is pitch or stall regulated. It simply consider a turbine to be pitch-regulated when the power curve has at least three points equal to the maximum. Unfortunately, this routing fails for measured power curves with a bit of noise in the power-curve data. It should not be a problem with power curves supplied by manufactures.