Sources of self-prediction error

[jfcorbett]

I'd like to know what the sources of self-prediction error are. Here is what I think they are.

1. Directional rebinning. Due to turning, some numbers are pushed back and forth between adjacent sectors during the wind atlas methodology up-and-down process. This causes some "blurring" between sectors.

2. Interpolation of wind atlas statistics (especially if mast height is not among the standard heights)

Am I right, and is there anything else? Does Weibull fitting have anything to do with it?


(A similar question was asked previously on this forum but got no reply. I therefore sent the above question to WASPSUPPORT@risoe.dk yesterday... will post reply if and when it comes.)

[Mkevin]

different w'tetha' or different oreography

[jfcorbett]

I'm not sure what you mean by w'theta' ?

Since only one vector map is used in self-prediction, a different Oreography should by definition not be an issue.

[jfcorbett]

I received the following reply from none other than Ib Troen himself.

It is true that prediction and self-prediction are influenced by the truncation mechanisms mentioned. In the "up" transformation the initial (observed) speed and directional bins are redistributed and re-binned. Also in the "down" branch, when using the atlas tables to arrive at site distributions, there is a re-binning by sectors due to wind turning (orography in particular), and a renormalization. Depending on the site (especially its directional inhomogeneity roughness-wise and orographic effects) this can be seen in self-prediction errors. Also, there are (usually much smaller) effects of the interpolation within the wind atlas tables when predicting between table heights and roughnesses. This effect can be remedied somewhat in case one uses only a narrower range of heights by selecting standard heights in the range.

The Weibull fitting is done at the "down" branch to give the atlas tables in terms of the Weibull parameters. The fitting procedure conserves power density, but not the mean values. Thus mean values may deviate from the original observations because of this.

It’s quite gratifying to see that I was pretty much right in my initial guess!

[Mkevin]

But oreography also is an issue as it was written unfortunately stability of atmosphere (vertical montions of air parcels) was omitted and this for sure has influence upon crossprediction errors.

[jfcorbett]

Perhaps, but it has no influence on self-prediction errors, which is the topic of this thread.

[legrandbleu]

Hi,
I 've had the same issue sometimes very amplified due to site particularities.

It is not however clear to me if those deviations have an impact on final energy yield figures or just mix the mean wind speeds intra-bins and overall.

Thanks,

Nick