PLANT DESIGN AND ENERGY PRODUCTION ESTIMATION - Wind Resource Assessment: A Practical Guide to Developing a Wind Project - page 246

be problematic where winds are heavily influenced by the terrain and land surface
properties. For example, winds channeled through a mountain pass may exhibit very
different directions at different points within and outside the pass. Similarly, as winds
transition from offshore to onshore, and vice versa, there is frequently a change in
direction caused by the contrasting thermal properties and roughness of the land and
ocean surfaces.
To handle such situations, most programs include a “directional shift” or similar
option, which makes a guess, based on the modeled variation in directional frequencies,
about how the observed frequencies differ between the mast and a given point. Again,
there is little evidence to say whether this adjustment improves accuracy, or by how
much, and its effect is likely to be very site-dependent.

16.7.2 Using Multiple Masts
As noted in Chapter 13, most wind energy projects employ more than one mast. This
poses the practical challenge of combining the information from the various masts in
estimating the energy production.
One common approach is to divide the project area into sections, each of which is
assigned to one mast. The sections may be defined by distance (i.e., the closest mast is
assumed to “dominate” the area), or they may be defined by some other criterion, such
as topographic similarity (e.g., ridgetop sections are assigned to ridgetop masts). The
simulation for each mast is done separately from the others, and the wind resource
for any turbines within a section is extrapolated from whichever mast that section is
assigned to.
This approach is pragmatic, but it can be awkward when, as often happens, there is
a discontinuity in the predicted wind resource where two sections meet. The resulting
energy production estimate can change abruptly (and unrealistically) when a turbine is
moved from one side of the dividing line to the other. This problem is a reflection of
the inaccuracies of wind flow modeling, for if the models were perfect, the predicted
wind resource at a point would be the same (within the measurement uncertainty at
the masts) no matter which mast was used to initialize or adjust the model.
A more esthetically pleasing, if not necessarily more accurate, approach is to
smoothly blend the predicted wind resource from the different masts. This method
adopts the assumption that every mast offers at least some useful information about
the wind resource at any point, and that the weighted average of several estimates
should be more reliable than any single estimate alone. The challenge is to determine
a suitable method of weighting. A relatively simple blending technique is to weight
each mast's prediction according to the inverse of the squared distance to that mast.
Extending Equation 16.1,

j = 1

M

v pij
d pj + C

=

v pij (16.6)

j = 1

M

1
d pj + C