Differences in speed values between observed wind climate and WTG at the same position

[Andrés]

Hi all, 

I am performing some checks between the observed wind climate (.tab) and a virtual WTG, which i have located in the same position as the met mast.

I understand that when inserting the .tab, the mean wind speed ("U") can slightly differ from the mean speed obtained from the measured values in the met mast. However, when i compare the mean wind speed ("U") from the observed wind climate, and the speed value in the virtual WTG, those values also differ, even though that both sources are located at the same position and height. Does someone understand why do these values are different? 

I can't attach screenshots in here, but the speed i am getting in the virtual WTG is of 6.23 m/s, while the mean wind speed ("U") from the observed wind climate is of 6.39 m/s (both emergent and fitted).

Thanks a lot!

 

[Morten]

In WAsP, the WTG object refer to the wind turbine generator, specifying power and thrust-coefficient curves, so I guess that you are actually referring to a turbine site placed at the met mast position. In this way you are making a self-prediction test, and wonder about discrepancies in the predicted mean wind speed. 

There are many ways to fit a Weibull distribution and in WAsP the fit is designed to match the mean of the cube of the observed wind speeds and the probability of winds higher than the empirical mean wind speed. This usually results in a good fit to data in the high-wind range but a less accurate fit at lower wind speeds. This is a deliberate choice, as wind power production mostly depends on accurate modelling in the high wind-speed range. However, unless the observed wind speed distribution is a perfect Weibull distribution, there is no guarantee that the fitted distribution will match the mean wind speed of observation. This is the main reason for the discrepancy. 

In addition, there can be small errors in sector-wise frequencies and Weibull distributions, due to a rotation of the wind rose when converting the observed wind climate to the generalized wind climate and a reverse rotation back to a (self-)predicted wind climate. These rotations shifts probability mass between sectors, and the finite sector width leads to discretization errors. This type of error increases for wind climates with a large variation between neighboring sectors. 

 

[Andrés]

Thank you Morten for the detailed answer, i was effectively referring to a turbine site placed at the met mast position as you mentioned. Some other questions related to your answer:

• What do you consider as a high-wind range and lower wind speeds for determining when the model provides a good fit?
• Do you recommend applying any corrections to the results of the model, specially when the model is less accurate at lower wind speeds, maybe comparing theoretical production at the met mast position using the real measured data-power curve v/s WAsP results? 

Thanks again!

[Morten]

In the context of fitting a Weibull distribution to an empirical distribution, I would call the wind speeds above the mean wind speed the high wind speed range. Try to open an observed wind climate in WAsP or WACA and compare the two distributions for high wind speeds and for low ones, and I think you will see what I mean. 

I do not recommend any correction to the model. We should remember that the AEP is found by an integral of the mean wind speed probability distribution and power curve, so we are not concerned about PDF model errors at wind speeds with zero or very low production.