Re: [LEAPSECS] power grids and leap seconds

From: Markus Kuhn <Markus.Kuhn_at_CL.CAM.AC.UK>
Date: Tue, 30 Sep 2003 22:23:47 +0100

Steve Allen wrote on 2003-09-30 18:54 UTC:
> One open question regarding leap seconds has been what to do with
> power grids so that clocks with motors that run synchronously with the
> AC power will experience 86400 * (50 or 60) cycles in a UTC day even
> if there is a leap second. Some reports have indicated that many
> grids adjust their frequencies on the day of a leap second such that
> such clocks will retain sub-second accuracy.

Power grid frequency varies automatically with the difference between
supply and demand and is controlled by opening and closing the steam
valves of the turbines that drive the generators. In the continental
European grid (UCTE), the spec -- which I have on my desk here -- says
that the time-scale defined by the rotating grid phase vector must
remain within +/- 30 seconds of UTC. The UCTE control center in
Laufenbourg, Switzerland sends out instructions every 24 hours to shift
the nominal frequency of 50.0 Hz, if necessary, to either 49.99 or 50.01
Hz (+/- 10 mHz) if the network clock starts to drift away from UTC too
far. For comparison: the measured frequency deviation from the nominal
frequency is not supposed to be more than 40 mHz for more that 10% of
the time and not more than 60 mHz for more than 1% of time.

So leap seconds are taken into account in UCTE grid frequency control,
but in practice these frequency adjustments made to follow a leap second
vanishes in the noise of the frequency offsets caused by normal supply/
demand regulation. With an (atomic) frequency meter connected to a power
socket, you'd probably have to average over a couple of weeks before you
can even recognize that there might have been a leap second.

> In light of recent world events on power grids, I'll offer that this
> might no longer be considered as a relevant issue. The grids should
> just keep atomic frequency. Clock owners appear likely to need to
> reset them much more often than every leap second.

Electricity grid quality requirements differ significantly in different
parts of the world. For example, on 14 August, New York was running on
65 Hz for a while (+5 Hz deviation), before the collapse, whereas in the
UCTE, controlled emergency load-shedding (10%-20% of consumers at each
phase) starts already a 1, 1.3 and 1.6 Hz deviation and power plants
finally separate themselves from the grid at 2.5 Hz deviation of the
nominal frequency (to prevent mechanical damage to equipment). In UTCE,
the design goal is that a sudden catastrophic 3000 MW change in load or
supply within a fraction of a second must not cause a frequency
deviation of more the 800 mHz and that no single failure in the grid
must on its own be able to cause such a change. Unfortunately,
privatization has led to suppliers not taking these specifications as
serious any more everywhere as they used to. Spare power plants and
lines to quickly reroute 3000 MW cost money that managers would rather


Markus Kuhn, Computer Lab, Univ of Cambridge, GB | __oo_O..O_oo__
Received on Tue Sep 30 2003 - 14:25:19 PDT

This archive was generated by hypermail 2.3.0 : Sat Sep 04 2010 - 09:44:54 PDT