power failure causes and effects

Fred Heutte aoxomoxoa at sunlightdata.com
Fri Aug 15 12:23:10 UTC 2003


It appears that this was the largest power outage on record,
in a variety of respects (geographic reach, number of grid
line-miles, megawatts of capacity, number of affected
customers, etc.).  

Despite all the noise already arising about the "antiquated"
American grid, it's important to recognize how stable and
reliable it generally is.

(did Bill Richardson really say "third world"?  now there's
someone who doesn't know anything about power 
engineering here *or* there)

The reason is simple: it gets down to good engineering
practiced over 100 years.  Strogatz and Watts did an 
interesting "small world" analysis of the western power
grid about six years ago.  A more recent paper by Motter
and Lai compares cascading failures of the Internet at
the AS level and the western power grid -- side by side!

http://chaos1.la.asu.edu/~yclai/papers/PRE_02_ML_3.pdf

Some of this work is now being transmuted into practical
form at EPRI and other transmission research places:

http://www.epri.com/programHigh.asp?objid=261741

The emergent properties of the power grid and the net
are similar for quite obvious reasons.  Whether consciously
done or not, good design requires hierarchical ordering,
loose coupling of regional systems, self-healing and "immune
response" mechanisms, and so forth.  These are then discoverable
at the mathematical level, which is what the "small world"
approach is all about.  

Neglect of infrastructure certainly raises the prospects for
cascading failures.  But the causality associated with one
single incident -- even one as widespread as msblast or
the August 14 power failure, is not easily attributable to
any single element, even if an initial failure point can be
identified.

Fred




More information about the NANOG mailing list