Page 59 - Wind Energy Handbook

WIND-SPEED PREDICTION AND FORECASTING 33

These transient events are deterministic gusts intended to represent the extreme
turbulent variations which would be expected to occur at the specified return
period. They are not intended to occur in addition to the normal turbulence
described previously. Such deterministic coherent gusts, however, have little basis
in terms of actual measured or theoretical wind characteristics, and are likely to be
superseded in future standards by something more closely related to the actual
characteristics of extreme turbulence.

2.9 Wind-speed Prediction and Forecasting

Because of the variable nature of the wind resource, the ability to forecast wind
speed some time ahead is often valuable. Such forecasts fall broadly into two
categories: predicting short-term turbulent variations over a time-scale of seconds
to minutes ahead, which may be useful for assisting with the operational control of
wind turbines or wind farms, and longer-term forecasts over periods of a few hours
or days, which may be useful for planning the deployment of other power stations
on the network.
Short-term forecasts necessarily rely on statistical techniques for extrapolating the
recent past, whereas the longer-term forecasts can make use of meteorological
methods. A combination of meteorological and statistical forecasts can give very
useful predictions of wind farm power output.

2.9.1 Statistical methods

The simplest statistical prediction is known as a ‘persistence’ forecast: the predic-
tion is set equal to the last available measurement. In other words the last measured
value is assumed to persist into the future without any change:

^ y y k ¼ y k 1

where y k 1 is the measured value at step k 1 and ^ y k is the prediction for the next
y
step.
A more sophisticated prediction might be some linear combination of the last n
measured values, i.e.,

X
n
^ y y k ¼ a i y k i
i¼1
This is known as an nth order autoregressive model, or AR(n). We can now define
the prediction error at step k by
y
e k ¼ ^ y k y k
and then use the recent prediction errors to improve the prediction: