Operation of power systems - Wind Power Integration: Connection and System Operational Aspects - page 188

inter-annual variability is low - in Europe the distribution tends to be normal with a
standard deviation of 6 per cent (EWEA, 2005).
Actually determining a figure for the capacity credit of wind generation,
i.e. the ability to displace an equivalent amount of 100 per cent firm capacity, has
proven to be controversial and subject to interpretation. Anecdotal evidence has
indicated that wind output can be low during periods of high system demand,
although others have suggested that across a sufficiently large region (Great
Britain/Europe) it will always be windy somewhere. Over a 30 year period, based
on wind speed measurements from over 60 locations across the United Kingdom,
no single hour was identified during which the wind speed at every location fell
below 4 m/s (turbine cut-in speed). On average, there was only 1 hour per year
when over 90 per cent of the United Kingdom experienced low wind speeds, and
only 1 hour in every 5 years when this would have occurred during the winter
higher demand period (ECI, 2005).
A variety of approaches have been proposed to calculate the capacity credit,
based mainly on loss of load expectation (LOLE) and loss of load probability
(LOLP) methods, with wind generation alternatively modelled statistically or using
time series data. For a particular area of interest, a wide range of capacity credit
figures can often be quoted, dependent on assumptions of system reliability,
seasonal wind regime, distribution of onshore/offshore sites, degree of system
interconnection, etc. Time series approaches can be particularly affected by coin-
cident weather/demand patterns, particularly during peak demand periods. Sensi-
tivity analysis, in the form of time shifting wind production relative to the demand
in 24 hour steps, improves robustness.
Following a probabilistic approach, the availability of each generating unit is
determined, based on past operational performance. Such an approach cannot be
extended directly to individual wind turbines, since power production will be
affected by the wind regime, which in turn results in the output of neighbouring
turbines being to some extent correlated. Consequently, geographical dispersion
and the smoothing effects of aggregation need to be recognised. Some approaches
define a dispersion coefficient (Voorspools and d'Haeseleer, 2006), which ranges
between a value of 1 (the output of all turbines are perfectly correlated - no dis-
persion) and 0 (total wind power output is constant - infinite dispersion) dependent
on the wind penetration level. For Ireland, as an example, a dispersion coefficient
of 0.33 has been adopted, resulting in a capacity credit of approximately 30 per cent
at a wind penetration level of 10 per cent (ESBNG, 2004b). Also considering the
Ireland system, Figure 5.29 illustrates the estimated capacity credit of wind gen-
eration for a 2020 scenario (Doherty et al. , 2006). It can be seen that for low wind
penetrations the capacity credit is approximately 40 per cent, exceeding the
assumed capacity factor of 35 per cent. This follows from the weak correlation of
wind power with system demand in Ireland, such that it tends to be windier during
peak demand periods (see Figures 5.9 and 5.10). As would be expected intuitively,
both reduced dispersion of wind farm sites and increased wind penetration cause
the capacity credit to fall. So, for example, with an installed capacity of 3,500 MW,
wind's capacity credit falls to approximately 20 per cent.