Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

power-grid physics



Hi --

I've been trying to figure out the basic
physics that made the blackout possible.

My main source of information has been
-- training manuals, for training powerplant and
grid operators. These are downloadable from various
utilities.
-- NERC standards, downloadable from NERC.

I've had somewhat limited success figuring out
what's going on. So the following are offered
as preliminary observations and hypotheses for
your consideration.

The first thing that jumps out is that the
documents do not spend much time talking about
power in terms that physicists would recognize
such as voltage * current. Mostly they talk
about _frequency_ measurements as a proxy for
measurements of power imbalance. If a generator
faces a sudden increase in load, it will slow
down (dropping below 60 Hz). This will be
detected (typically by a flyball governor!) and
this leads to the opening of a valve on a steam
turbine to bring things back to equilibrium.

At first glance you might think there is nothing
but the rotational kinetic energy of the generating
machinery to keep things running while this
feedback loop is making adjustments. This is
kinda scary, because the loop has distinctly
finite bandwidth, and the rotational KE is verrrry
small compared to the perturbation caused by the
connection or disconnection of a tie-line to/from
another part of the grid.

At second glance, the utility relies on the fact
that a goodly part of its customer load consists
of rotating machinery, which has rotational KE
of its own. The utility relies on this to keep
the system stable. Wow. If a small part of the
system becomes an island, cut off from the rest
of the grid, it becomes significantly less stable.

Also note that as the customers become more
modern, more of their energy goes into lighting
and computers and semiconductor-controlled motors
which provide no inertia to the electrical system,
making things less stable year by year. (I don't
know how significant this is; I see no evidence
that the industry has even considered the question.)

============

Another interesting bit of physics concerns phasing.

When I was in high school, I imagined that the
power grid operated as a giant phaselock loop.
That is, I assumed that all the generators ran
in phase. Obviously bad things are going to
happen if you have two AC generators 180 degrees
out of phase.

But guess what: They don't even try to phase-lock
them. They try to keep the frequency everywhere
close to 60Hz, but they don't even try to phaselock
them.

They aren't completely ignorant of phase, but
they don't call it that. Instead they speak of
"reactive loads" and "reactive flows" or more
commonly just "VAr" (volts times amps, reactive).

VAr occurs when the voltage is 90 degrees out of
phase with the current.

The utility gets paid for selling VA(nonreactive)
whereas VAr is nothing but trouble. Reactive
amps cause just as much I^2 R losses in the
transmission equipment, and just as much I^2 R
heating (which is a major limitation on what the
equipment can handle).

Obviously if you have one region running at 60.00Hz
and another region running at 60.03Hz, there are
going to be some really bizarre current-flows on
the tie-line joining those regions. As far as I
can tell, they don't even try to prevent this.
The only feedback loop they usually have (valves
on steam turbines) doesn't provide enough loop
bandwidth to phaselock the loop, so all they
can do is get the frequency "pretty close" and
then throw the switch to connect the tie-line
and hope for the best.

As far as I can tell, the only thing that keeps
these bizarre current flows from becoming nearly
infinite is the self-inductance of the wires in
the tie-lines. It's not at all clear that this
is well controlled.

Some of the more modern tie-lines are DC, with
AC/DC converters at each end. This ought to make
the phasing problem go away. But I don't think
these are very common.

=================================

Another odd thing has to do with load-shedding
versus overload.

The rules for operational reliability and security
sensibly enough say that when faced with an
impending overload situation that cannot be
remedied in any other way, the utility should
drop some fraction of its customers. This is
called load-shedding.

So the question arises: Suppose a given tie-line
trips off. What should you do:
-- shift the load to the remaining tie-line(s), or
-- shed some customers?

The answer depends on whether the remaining tie-lines
can handle the additional load. If they can, you
don't want to drop customers unnecessarily.

The goofy thing is that they don't necessarily
know how much load a tie-line can handle! A
common failure scenario is that a line stretches
due to thermal expansion, due to a combination of
hot weather and I^2 R heating. This makes it sag
into a tree. But the utility doesn't have
high-confidence information about all the trees
in the world. So sometimes they don't know that
a line can't handle the load until they apply
the load and see what happens. Obviously there
is major potential for a ripple effect here.

===========================

The NERC reliability and security standards say
that each powerplant should have blackstart
capability.

Obviously most of the plants in the region affected
by the recent blackout are not in compliance with
these standards.

If there had been good blackstart capability, I
imagine that at least half the customers (including
all the essential services) would have had their
power restored in less than an hour, whereas in
the event many of them had to wait 27 hours as
the restart percolated west to east and north to
south.

New York City does not have enough endogenous
power to operate everything in the city, but it
has more than enough to operate the subways and
the traffic lights.

In the event, though, some NYC powerplants were
among the last of the non-nuclear plants to come
back on line, because they needed to wait for
imported power to be restored before they could
restart. This boggles the mind. Would you
buy a car that couldn't be started except by a
push from your neighbor's car, especially when
your neighbor's car has the same "feature" and
both of you have had reliability problems for
years?

=============================

I am disappointed by the exceedingly superficial
media coverage of these issues. For example,
http://news.google.com/news?q=blackstart
returns only one hit .. from the Centre Daily
Times of State College, PA ... hardly a major
media outlet.