Due to the required accuracy special attention has to be paid
on the following points for wind measurements necessary for
the planning of wind energy converters:
Use of suitable measuring instruments for wind direction and calibrated anemometer.
Avoidance of oblique wind hitting on the measuring instruments and lee by the measuring installation.
Choice of a suitable measuring site nearby the planned wind farm, if possible in an obstacle-free area.
Correct installation of the anemometer in suitable measuring heights.
The best results can be expected at a wind measurement in hub
height of the planned wind energy converter. Considering the
nowadays large hub heights of wind energy converters this is
very complex and costly. Besides the hub height of the
converter is often not yet fixed at the beginning of the
measurements.
Therefore, wind measurements are executed in lower heights.
The wind velocity in hub height will be computed afterwards.
Therefore, it is necessary to determine the wind velocity in
at least two different heights. According to the measurements
one may draw conclusions about the area situation (roughness)
and is able to reflect and determine a height profile of the
wind velocity of the site.
To receive reliable results, the wind measurements should be
observed for a longer period, at least for one year. A
decisive point in the processing is the long-term correlation.
While on-site measurement data usually is collected for a
comparatively short period of time, the knowledge of the
long-term wind regime is critical to completing a wind and
yield study. Usually this is achieved by projecting the data
over the desired long-term period using an adjacent long-term
data source and using MCP (Measure Correlate Predict) methods.
AL-PRO developed a procedure that clearly outperforms common
linear MCP approaches and typically leads to a far higher
precision of the long-term projection. This method works on
basis of neural computer networks that are able to recognize
complex and non-linear relations between the datasets. In
addition, up to 4 long-term data sources can be used for the
projection simultaneously.
(Long
Term Correlation of wind measurements using neural networks)
