Page 172 - Fluid Power Engineering

W ind Resource Assessment 145

and standard deviation of σ. The standard deviation also called the
standard error is an estimate of the uncertainty of the prediction. An-
other estimate of uncertainty is the correlation between YM and f (X)
time series. Both estimates of uncertainty are provided by tools at the
end of prediction using any of the methods listed above. A prediction
method with lowest σ and the highest correlation is a candidate for
prediction. Care must be exercised to ensure that a candidate method
is not chosen just because it has the lowest standard deviation. For
example, if the correlation post prediction (between f (X) and X)is
lower than correlation pre-prediction (YM and X) for the measure-
ment period, then the prediction method may not be valid.
To improve prediction, the following changes to the prediction
method must be evaluated for better predictability:

Filtering out of wind speed data below 3 m/s. This data is not

important and it can skew the linear regression.
If the long-term reference data is hourly and measured data

is 10-min, then using hourly average of measured data versus
point estimates may yield a higher correlation.

After the transfer functions have been determined, predicted time se-
ries is computed by adding the residual term in Eq. (6.3) using the
Monte Carlo method. The predicted time series is then used to com-
pute annual energy production, as described below.

Annual Energy Computations
There are two common metrics used to compute energy production.
The first is power density

N
1 3
Power density = ρ i v i (7-10)
2
i=1
where ρ i , v i are air density and predicted wind speed in time period
i; each time period is of length t hours, and N is the number of
data points in the predicted long-term wind speed time series. Unit
of measure of power density is: kilowatt per square meter.

N
i=1 PC(v i )
Annual energy production (AEP) of specific turbine =
N/8760
(7-11)
where PC(ν t ) is the power curve of the chosen turbine, (for example
see Fig. 2-8).