Page 226 - Wind Energy Handbook

200 WIND-TURBINE PERFORMANCE

of Equation (4.12). This can, however, be misleading as it assumes that the wind
turbine maintains a constant C P across its output range and this is far from true.
Christensen and Dragt (1983) discuss this issue in detail and conclude that such
simple corrections should not be made. An example calculated in the same report
does show that a turbulence intensity of 13 percent could give rise to an error in the
predicted energy yield of a particular wind turbine of 1.6 percent. Higher turbu-
lence levels can be encountered and could give rise to errors of 5 to 15 percent. If
the proposed site has a turbulence intensity similar to that on which the wind
turbine was tested then no need for correction will arise. What is needed then is a
record of the turbulence present on the site during the testing, and this is indeed
specified in the international standard. Specifically, plots of mean wind speed and
turbulence intensity as a function of wind direction should be presented for each
data set selected.

4.10 Aerodynamic Performance Assessment

There is often a need for assessing the aerodynamic or instantaneous performance
of a wind turbine. Detailed features, such as the stall characteristic, will tend to be
smoothed out by the 10 min averaging employed in power performance assess-
ment. Consequently only short averaging periods can be used, but this introduces a
further limitation of the method of bins. It has been shown that poor correlation
between power and wind speed results in a systematic distortion of the binned
relationship (Christensen and Dragt, 1986; Dragt, 1983) and shorter averaging times
result in poorer correlation. The effect is to rotate the power curve about the point
where the wind speed probability is highest as shown in Figure 4.24. This can be
understood as follows. Consider a short gust of high wind at the anemometer. The
likelihood is that at this instant, the wind speed at the turbine will be lower (i.e.
nearer the mean), and consequently the power output measured will be less than
would have been expected from the power curve with the wind speed as measured
at the anemometer. This will bias the measured power curve down at higher than
average wind speeds. A similar argument shows that the curve will be biased
upwards at lower than average wind speeds, as illustrated in Figure 4.24.
Dragt (1983) has developed a formula for correcting the measured wind speed

based on the statistics of the sample data set. The corrected wind speed, U ,is
given by:

U ¼ U (1 r)(U U) (4:13)
Here U is the measured wind speed, U is the sample mean, and r is the correlation
coefficient between power and wind speed. This correction was derived on the
basis of a normal distribution for the wind speed variations, and care should be
taken to check that this is applicable to the sample data set before carrying out the
correction. It could be that data are collected so as to cover the wind-speed range
evenly. If this has been done effectively then no systematic distortion should occur.
Another approach to reducing the averaging time whilst maintaining a high