[nSLUG] US Homeland Security
jason at ohm.ath.cx
Thu Aug 19 13:15:09 ADT 2004
> The explanation I read last year made it clear that the power systems
> did not fail all at once. It went down as a series of cascading
> events, and as each sub-grid went down, the automatic response
> of diverting power from the neighbouring sub-grid then caused
> it to have too much burden, and then that sub-grid failed, and so on.
> The dominos analogy fits even better than it has elsewhere.
You're right that it didn't happen *instantaneously* everywhere, but the
timescale over which the "cascade" happened was very short.
Take a look at this:
This is a very comphrensive overview of the blackout. The situation in
Ohio gradually worsened until about 4:08, when it spread to other states.
Look at the time timeline from 4:08 to 4:11. This small 3-minute window is
when 90%+ of the disconnects/failures happened.
At 4:13 they declared the situation static, eg. no more changes.
In any case, the problem was still that when the load got too high,
some of the grids didn't disconnect themselves automatically in time. And
as I said before, some grids did. That's why we still had power in NS.
When this report was published, they weren't sure why the
> People were writing 20 or 30 years ago that the super grid was
> vulnerable to a failure like this. If McGuinty's idea of integrating
> with Quebec and Manitoba had already been in place, then there would
> have been two more dark provinces.
That's quite the claim. Quebec has one of the most advanced power system
networks in the entire world. And it's not up to Ontario to connect to
Quebec if I remember right, it's quite Quebec refusing to connect to
Ontario. I could be wrong about the politics here, I've probably only
heard them second hand.
And to repeat myself, we were connected directly to New England, who were
connected directly to New York and Ontario, and disconnected from them
both. Ontario only lost power in a small region close to the States as
well, most of Ontario (land-wise...) still had power. I think your Quebec
and Manitoba claim is silly.
> The obvious solution is to introduce
> fewer grid load diversions that can take place automatically like this.
> The existing ones or new ones need a component that can decide whether to
> suck up the coming load or disconnect.
They already have them. Just, some didn't work. Some did. Your
"obvious solution" is to REDUCE the amount of single points of failure?
Having a section of the grid go completely down is really really bad. When
the generators went down in this crisis, it's a two or three day process
to bring them back up so they are operational again. And if you split the
power grid into a bunch of different islands, it's really hard to connect
them back up again. Keeping things as connected together is a good thing,
I think. Sychronization is very important!
But since this is a Linux mailing list, no more!
More information about the nSLUG