Close to Critical
2003-08-17 10:00 pmIn a deep sense, systems near their critical point are very similar. The critical point is the value of some controlling parameter at which the system makes a phase change, from ordered (predictable) to disordered (chaotic).
On one side of the critical point, the system has a finite number of stable states; no matter what the initial conditions, the system will end up in one of them. On the other side there is no stable state at all; the system is chaotic, and two sets of initial conditions that are arbitrarily close together will evolve to be arbitrarily far apart, usually in a short time.
Near the critical point, things get unpredictable in a strangely predictable way: the system is mostly stable, and any small instabilities that appear will eventually be damped out, but they may take an arbitrarily long time and involve an arbitrary amount of the system. A pile of sand is a good example; adding a few more grains can start an avalanche, which may be either tiny or huge.
Near the critical point, things are interesting -- often in the sense of the ancient Chinese curse.
It seems reasonably clear that the US's electrical grid is pretty close to critical.
One might think that this is an undesirable situation, but getting back to the stable part of parameter space would require adding a sizeable margin of safety to the system: overengineering it, adding more capacity. The problem is that extra capacity costs money, and in order to be effective it has to sit idle almost all the time. As a wild guess, you'd need about 1/3 more capacity than you normally use (that's the amount you need in an ethernet or a hash table, and as I mentioned systems tend to behave similarly as they approach criticality).
So what actually happens is that the margin of safety gets shaved, for economic reasons, until the level of instability starts to affect the bottom line. This is a feedback loop that, perversely, guarantees that the system will always be close enough to critical that occasional large-scale events -- like blackouts -- remain possible.
On one side of the critical point, the system has a finite number of stable states; no matter what the initial conditions, the system will end up in one of them. On the other side there is no stable state at all; the system is chaotic, and two sets of initial conditions that are arbitrarily close together will evolve to be arbitrarily far apart, usually in a short time.
Near the critical point, things get unpredictable in a strangely predictable way: the system is mostly stable, and any small instabilities that appear will eventually be damped out, but they may take an arbitrarily long time and involve an arbitrary amount of the system. A pile of sand is a good example; adding a few more grains can start an avalanche, which may be either tiny or huge.
Near the critical point, things are interesting -- often in the sense of the ancient Chinese curse.
It seems reasonably clear that the US's electrical grid is pretty close to critical.
One might think that this is an undesirable situation, but getting back to the stable part of parameter space would require adding a sizeable margin of safety to the system: overengineering it, adding more capacity. The problem is that extra capacity costs money, and in order to be effective it has to sit idle almost all the time. As a wild guess, you'd need about 1/3 more capacity than you normally use (that's the amount you need in an ethernet or a hash table, and as I mentioned systems tend to behave similarly as they approach criticality).
So what actually happens is that the margin of safety gets shaved, for economic reasons, until the level of instability starts to affect the bottom line. This is a feedback loop that, perversely, guarantees that the system will always be close enough to critical that occasional large-scale events -- like blackouts -- remain possible.
no subject
Date: 2003-08-18 11:15 am (UTC)no subject
Date: 2003-08-18 07:33 pm (UTC)The sort of thing that happens in one of these massive failures is that you get a brief surge someplace. So a breaker pops open to prevent melting the transmission line (it's been known to happen). Now the power plant on the input end of that line sees a sudden drop in its load, and throttles back. Meanwhile any power plant on the output end is suddenly straining to deliver power into a dead short. So it shuts down, hopefully before damaging itself. Tens of megawatts change direction in milliseconds.
Sure, if you can use local generation to break the grid into neighborhood-sized chunks, you might be able to cut down on the number of major incidents. It would probably be at the expense of either making the control problem even more complex, and/or increasing the number of little local failures (due, for example, to the large number of people unplugging their cars at 7:40am every weekday).
no subject
Date: 2003-08-18 07:55 pm (UTC)no subject
Date: 2003-08-18 08:23 pm (UTC)