Page 163 - Fluid Power Engineering

W ind Resource Assessment 137

Regression Method Equation
Linear regression Y = a 1 X + a 2
Linear regression through (0,0) Y = a 1 X
2
Quadratic regression Y = a 1 X + a 2 X + a 3
2
Quadratic regression through (0,0) Y = a 1 X + a 2 X
Source: WindPRO.
TABLE 7-8 Regression Methods Commonly Used

Wind direction is normally predicted using linear regression of
the form:

(7-4)
Y = X + a 2
The regression methods are applied sector-by-sector. Figure 7-10a
and b graphically illustrate the use of linear regression in two sec-
tors. The straight line is the prediction. Error is plotted in Fig. 7-11a
and b.

Weibull Parameter Scaling
In this method, the Weibull parameters A and k are computed for
wind speed in each of the 12 or 16 wind direction sectors for both the
onsite measured data and reference data. In this simple method, the
following is used to scale the parameters: 9

long short
long short
λ = λ . λ λ (7-5)
site site ref ref
long
where λ is predicted parameters A or k for onsite wind data. The
site
quantity in parenthesis is the correction factor (see Table 7-9).
The frequency of each sector also needs to be computed by nor-
malizing the sum to 100%. 9

short N short
f f
long sitei long sitei long
f = f f (7-6)
sitei short refi short refi
f f
refi i=1 refi
long
where f is the prediction of onsite frequency in sector i, N is the
sitei
number of sectors.
This method is appropriate only when:
Both time series are primarily Weibull distributions, and the

scaling of the two parameters is not very large.