Wind Turbine Performance Measurements by Means of Dynamic Data Analysis

The state of the art power performance measurement method refers to the IEC61400-12-1 standard from 2005 [1]. A method for faster power curves was proposed by researchers at Oldenburg university in 2004. The method was called Langevin power curve method and advantages was claimed to be that power curves could be made faster with 1Hz dataset.

In the FastWind project the Langevin power curve method was used on real power curve measurement datasets with the purpose to evaluate the method for practical use. A practical guide to application of the method to real power curve measurement data was made. The study showed that the method has a range of parameter settings that the user must consider.

Additionally to the wind speed binning power binning is needed but power binning size is not specified. Determination of drift in each bin is described with a general formula but in practice several additional tools have been developed by authors to try to make the drift field and fixed point determination more robust.

A sensitivity analysis with nacelle lidar data showed drift determination was not very dependent on the time steps applied, leading to use of time steps of 2-3 points for each dataset. Power bin size should be fixed. Data averaging with 5 sec data was more distinct for determination of the fixed points than 2 and 1 sec data.

With the nacelle lidar the Langevin method seemed to produce a power curve that was comparable to the IEC power curve. Analysis of the Langevin method with spinner anemometer data showed that fixed points were very sensitive to bin size and to requirement of minimum amount of data in each bin. The Langevin method failed to produce acceptable robust power curves comparable to the IEC power curve.

Simple binned averaging of data with shorter time averages gave better results than the Langevin power curve method.