Operation of power systems - Wind Power Integration: Connection and System Operational Aspects - page 141

each stage to ensure that the demand is met at minimal cost, while observing the
minimum up and down times of each unit. A variety of mathematical techniques
are available for solution, including the use of neural networks and genetic
algorithms. However, the standard approaches are based on mixed integer pro-
gramming or the Lagrangian relaxation technique, whereby a Lagrange function
(incorporating the multi-stage cost and weighted loading and unit limit constraints)
is optimised subject to the unknown multipliers (Wood and Wollenberg, 1996).
Of course, in order that unit commitment can be achieved successfully, it is
essential that accurate forecasts of the system demand for the 24-hour period are
available. Many factors affect the system demand, as outlined earlier, but these
are reasonably well understood and it is generally possible to forecast the demand with a
1-2 per cent error over the required time horizon. The most critical periods tend to be the
morning rise, the daily peaks and the overnight trough. Forecasting is achieved using a
combination of data trending and analysis of cyclical variations from historical load
profiles, and/or construction of an overall demand profile from a sampling of individual
load sectors. Weather forecast data, awareness of major events, including dramatic
developments in popular TV programmes, and system operator judgement may also
have an important role to play in fine-tuning the final predicted demand curve.
So far, against a predicted demand profile, unit commitment has been
performed using Lagrangian relaxation techniques, or otherwise, to ensure that
sufficient generation plant has been committed to the system at the appropriate
time. Since units can only be committed in discrete steps, spare generation capacity
will normally be available to cope with errors in the forecasted demand. However, it
is clearly not possible to predict the demand perfectly at each instant in time, so the
task remains to ensure that a continuous balance is obtained between demand and
generation. As an illustration of the difficulties that can be encountered, Figure 5.3

35

34

33

32

31

18:30 19:00 19:30 20:00 20:30 21:00 21:30 22:00 22:30

Time (h)

Figure 5.3 World Cup 2006 final - Italy versus France (National Grid)