Optimization of Wind Turbine Operation by Use of Spinner Anemometer

A prototype spinner anemometer was developed from a standard scientific sonic anemometer with specially designed 1D sonic sensors. A model spinner anemometer was tested in wind tunnel with two sensor head configurations. The tests showed that the sonic sensors responded with a high influence factor on yaw errors, and that the sensors responded with sinusoidal behaviour to rotation of the spinner.

The tests also revealed a significant "sensor head flow distortion effect" from the classic sensor heads. A full CFD analysis of the model spinner anemometer was made. The results showed that the calculations were almost insensitive to rotation and to wind speeds. For all flow angles up to 60º the azimuth variation was a pure sinus.

The shape of the responses was found to be described with a simple function that over one revolution decreases the average value with a cosine to the flow angle and increases the amplitude with a sine to the flow angle. The relation can be used in a conversion algorithm for the spinner anemometer. Field measurements were made with the model spinner anemometer.

The spinner anemometer measurements were compared to measurements from a standard 3D sonic anemometer. The results showed time traces of the two instruments that were very similar. The prototype spinner anemometer was mounted on a 3.6MW prototype wind turbine. Statistics of the yaw error showed an average of about 10°.

The average flow inclination angle was about 1°. The spinner anemometer measurements were correlated with wind speed and wind direction from a free meteorology mast. The results showed that the gain factor of the yaw error was only 0.80, which indicates that the yaw error measurements were overestimated with the use of the K factors from the CFD analysis.

The wind speed at the free mast ahead of the wind turbine was compared to the wind speed measured by the spinner anemometer while the wind turbine was yawing while idling or stopped. The results showed that average wind speeds compared well within about 1%. The power of the 3,6MW wind turbine was analysed for varying yaw error but though the yaw error varied with more than ±20° during the period it was not possible to verify an expected cosine squared variation of power with yaw angle.

The reason is that there was not sufficient data at all relevant wind speeds at the different yaw angles.