Environmental Engineering Reference
In-Depth Information
Chapter 6
Wind power forecasting
6.1
Introduction
The increasing penetration level of wind power into power systems has resulted in
the requirement for accurate, reliable, online wind power forecasting systems. The
typical transmission system operator (TSO) requirements for a forecasting system
have been identified by Schwartz (2000) as follows:
The forecasts should be of wind power output (in MW), rather than wind
speed, with look-ahead times extending out to 48 hours.
●
The forecasts should be available for individual wind farms, for regional
groupings of wind farms and for the total wind power installed in a TSO's area.
●
The forecasts should be accurate and supplied with an associated level of
confidence - clearly dispatchers would tend to be more conservative when
dealing with large forecast uncertainties.
●
The forecast should predict changes in wind power reliably.
●
There should be a good understanding of the meteorological conditions which
would lead to poor quality forecasts.
●
Historical data should be used to improve the forecast over time.
●
Wind power forecasts are used as inputs to the various simulation tools (market
operations, unit commitment, economic dispatch and dynamic security assessment)
available to the power system operators to ensure that economic, efficient and
secure operation of the power system is maintained.
Over the last decade or so there has been considerable activity and progress in
the development of wind power forecasting. However, there is still much scope for
improvement. Two broad strands can be identified in the systems that have been
developed so far: those using predominantly physical modelling techniques; and
those using predominantly statistical modelling techniques. In the physical model-
ling approach the physical atmospheric processes involved are represented, in as far
as this is possible. The statistical modelling approach is based on the time series of
wind farm power measurements which are typically available online. If purely sta-
tistical modelling techniques are used, good forecast results can be achieved for the
short-term look-ahead times (from 0 to 6 hours) only. Beyond this, and especially in
the 12- to 48-hour range of look-ahead times, it is essential to use an input from a
numerical weather prediction (NWP) model if successful results are to be achieved.
Search WWH ::
Custom Search