Keeping the lights on in GB

In this post I propose and demonstrate a simple rule for determining how much dispatchable (non-wind) power is required to be available to the GB grid to “keep the lights on” in coming winters.

Background

Recent years in the area covered by the GB electricity grid have seen major changes in the mix of generators, with some types (wind and solar) expanding rapidly, other types diminishing rapidly (coal and oil), and some remaining mostly stable (gas and nuclear), though some gas plant has been mothballed. The rapid decline of online dispatchable generation has raised concerns of a shortfall of supply at the times of peak demand, which occur in the early evening on cold working days in winter.

If consumers of electricity wish to consume a total of X GW in an early evening then that will only be possible if the sum of available wind and non-wind power is at least X. The current 14.3 GW of wind power (metered plus embedded) reduces the demands on the non-wind generators, and the critical question is by how much at the times of peak winter demand. This post provides the answer to that question, in the form of a simple rule, for the weathers of the most recent nine winters.

Minimum Dispatchable Power

I have examined wind power and demand data for the most recent nine winters, and for each of these winters the highest dispatchable power used follows closely a simple rule, which can be illustrated directly by showing an example. The following figure shows, for the winter weather of 2008/09, with 2016/17 wind capacity, the daily peak consumption (red dots) and the wind-reduced demands on non-wind generators (solid red):

The rule in words is:

The amount of dispatchable power available should be at least 103% of the maximum expected consumption, giving 5% headroom to deal with unexpected losses of power, demands higher than expected, and abnormally low wind power. In effect, wind power at 2016/17 levels can be relied upon, given the capacity margin, to reduce the highest peak demand by 2%, sometimes shifting the day of highest demand, as it did in the example shown above. The substantial amount of wind power capacity in 2016/17 means that the day of highest non-wind demand will always be a day of very low wind power, so that the highest non-wind demand reduces very slowly with increasing wind capacity.

Note in the figure above that wind power is substantial on the majority of days, and on those days it contributes to system security by catering for substantial losses of dispatchable power. However, there was a calm cold day that limited the overall benefit of wind power to around 2% of peak demand, with a very similar benefit found for all nine winters examined.

Derivation

This part gives the details that have so far been skimmed in the above. The details have been covered already in other posts on this blog, and readers may wish to go to the next section, which demonstrates the proposed rule for all nine winters examined.

Consumption is defined here as the sum of demand (on all metered generators) and embedded wind, both of which are contained in files that can be downloaded from a National Grid (NG) website (given on the NATIONAL GRID DATA page of this blog) . This sum is not the total consumption, because some of the total is met by other distributed generators, but is the closest to it that can be achieved from the data examined.

The data for consumption at the times of daily peak demand were then “demodulated” to remove (approximately) the effect of reduced demand at weekends (including Fridays) and holidays. The demodulation was done by temperature-weighted interpolation across weekends and holiday periods, plus extrapolation at typically 500 MW per degree C when a temperature falls outside the range covered by nearby (in time) working days. The temperature used was the mean of the Central England Temperature (HadCET).

Estimated data for embedded wind is available in the NG demand files for all years back to 2007/08, and is scaled-up to 2016/17 levels using ratios of capacities, plus a small amount of manual scaling to get consistent levels for all past winters. Data for metered wind was downloaded from the gridwatch website ( http://www.gridwatch.templar.co.uk/ ), available there from 2011/12, and was also scaled-up to 2016/17 levels, but to levels observed in 2016, not using ratios of capacities. Metered wind before 2011 was estimated by suitable scaling-up of embedded wind data, again to levels observed in 2016.

Subtracting total wind power (metered plus embedded) from consumption gives the demands on metered non-wind generators. The highest such demand in each winter examined was found to be around 98% of peak consumption, a figure that reflects the differences in demand between breezy and calm cold days, and the wind speeds of those calm cold days.

Demonstration

The following set of nine figures demonstrate the rule for dispatchable power for the weathers of each of the most recent nine winters. Each figure shows what would happen in 2016/17 if there is a repeat of the weather of the winter in question, assuming that variations in consumption remain the same. In each case wind power has been scaled-up to 2016/17 levels.

Discussion

Note that the 103% part of the minimum dispatchable power rule is determined mainly by the reliability of generation and transmission equipment, and by the accuracy of demand forecasts, both beyond the scope of this blog. It is the 98% part of the rule that is being established here via the data shown above for recent winters.

The 98% figure reflects the differences in demand between breezy and calm cold days, and the wind speeds of those calm cold days.