Wind resource and site assessment - Wind Power Generation and Wind Turbine Design - page 63

nor seasonal infl uences from vegetation and fi nally there should be no long-term
equipment trends. However the quality of re-analysis data still depends on the
quality of the input data, with the result that in sparsely instrumented regions, the
re-analysis data can still suffer from defi ciencies in individual instrumentation and
data coverage. The same care should be applied to the use of re-analysis data, as
with ground-based data. The NCEP/NCAR data is available in the form of pres-
sure and surface wind data in a 2.5° grid corresponding to a spacing of approxi-
mately 250 km. The data consists of values of wind speed and direction for four
instantaneous values per day (every 6 hours).
The statistical method for long-term correcting data is called Measure-
Correlate-Predict or MCP. This method is based on the assumption that the
short- and long-term data sets are correlated. This correlation can be established
in different ways depending on the data quality and the comparability of the two
wind climates.
3.1.2 Regression method
If one mast is correlated with a second on-site mast, a linear regression either
omnidirectional or by wind direction sectors (typically 30°) might be best suited.
For the concurrent period, the wind speeds are plotted versus each other and a
linear regression based on the least-square fi t is established. This relationship is
used to extend the shorter data set with synthetic data based on the longer data
set. The regression coeffi cient is a measure of the quality of the correlation.
R 2 should not be less than 70%. The same method might be appropriate for a
short-term measurement on-site and a reference station in some distance if the
orography and the wind roses are closely related. If the wind roses vary, care has
to be taken since the wind rose of the reference station will be transferred to the
site by applying a linear regression, which can have a signifi cant impact on the
layout as well as the energy yield. Another inherent problem of this methodol-
ogy is the decreasing temporal correlation between the site and the reference
station with increasing distance. The introduction of averaging of the two data
sets can improve the correlation.
3.1.3 Energy index method
Rather than transposing the wind distribution from the reference station the energy
index method determines a correction factor for the short-term data. For the con-
current period, the energy level of the reference data set is determined and com-
pared with the long-term energy. The resulting ratio is then applied as correction to
the short-term on-site data set. The correlation is best proven comparing monthly
mean wind speeds of the two data sets.
This method has the main advantage that the on-site measured wind rose is not
altered. The energy index method is particularly suited for NCEP/NCAR data
since the low temporal resolution of NCEP/NCAR prohibits the use of the more
detailed regression method. However, care has to be taken since NCEP/NCAR
data represents only geostrophic wind and does not refl ect local wind climates like
wind tunnel effects across a mountain pass or thermal effects.