Characteristics of the Wind - Wind Turbine Technology: Fundamental Concepts of Wind Turbine Engineering - page 528

Physical Modeling
Physical modeling consists of obtaining velocity, direction, and turbulence measure-
ments over a scale model of the selected terrain in a wind or water tunnel. Data from the
tunnel model may be acquired in sufficient detail to determine the spatial variation in the
wind flow field. One wind tunnel study of a wind power station site [Davis 1984] was able
to accurately predict the sign of the change in local wind speed from the average ( i.e.
whether a specific turbine site would receive more or less wind than a reference site) and
it also did well in defining areas with low speeds. However, the magnitude of wind speed
variations caused by complex terrain was underestimated. Testing of a contoured model
of the Kahuku Point area on the north coast of Oahu (Fig. 8-2) produced wind speeds about
5 percent higher than those measured at 18 field anemometer data stations [Chien et al.
1980]. The opposite result was obtained with a physical model of the region of the Rakaia
Gorge in New Zealand [Meroney et al. 1978].
Physical modeling of the flow field over a planned wind power station can be very
useful in the final stages of wind turbine siting. It can be used to locate individual
machines and wind monitoring equipment to within a small area. It does, however, require
large specialized facilities [Hunt and Fernholz 1975, Riley and Delisi 1977], and only
neutral stability conditions are easily simulated in a wind tunnel. References for more
detailed descriptions of the theoretical principles of physical modeling of the atmospheric
boundary layer are Snyder [1972], and Hiester and Pennell [1981].
Numerical Modeling
Numerical flow modeling uses digitized and gridded terrain elevations as the lower
boundary of the model and produces vector and contour plots of predicted local wind
speeds and directions for specified boundary conditions. Complex terrain in a 2-km square
section of a wind power station was numerically modeled using nodal points on a 40-by-40
grid [Wegley and Barnard 1986]. For 8 different flow conditions, values of RMS error
between calculated and observed data at 28 short-term anemometer stations ranged from 5
percent to 18 percent, and errors were less than 8 percent for six of these flow conditions. A
numerical model of a 1.6-by-3.7 km wind power station with a 30.5-m grid spacing [Veenhui-
zen and Lin 1987b] accurately simulated the combined energy production of 93 turbines and
gave the undisturbed wind flow at 18 measuring stations within an RMS error of 7 percent.
The accuracy of numerical flow models is significantly improved when measurements
from a few locations are used to adjust the direction of the initial flow and the model
parameters that simulate the enhancement or suppression of vertical motion due to thermal
stratification. Measurements for this purpose can be made over a very short time period
when atmospheric conditions are climatologically representative of the power-producing
winds. The few measurements recommended to help determine the spatial variability of the
wind over an area should not be confused with measurements made to estimate the
long-term average of the wind speed. Under most conditions, a minimum of one year of
data is required to estimate a long-term mean wind speed to an accuracy of 10 percent with a
confidence level of 90 percent [Corotis 1977].