Systems Seminar
Cascading Failure, Criticality, and the Risk of Large Blackouts
Prof. Ian Dobson
UW ECE Department
Abstract
What is the risk of large blackouts such as the spectacular cascading
blackout in August that cut power to 50 million North Americans?
Data on North American blackouts suggest that large blackouts,
although rare, occur more frequently than expected.
Indeed the empirical probability distribution of blackout sizes shows a
power tail similar to that observed in other complex systems near criticality.
Since component failures in power systems tend to be dependent on
component loading and can cause further failures in a cascading fashion,
we examine a probabilistic model that simply represents these general
characteristics.
The model also describes the number of customers arriving during the
initial busy period of a queue.
The distribution of the number of failures is a saturating
quasibinomial distribution.
At a critical loading the model produces power tails in the
distribution of the number of failed components.
The model can be approximated as a saturating Galton-Watson branching
process and the critical loading corresponds to the usual branching process
criticality. This suggests a way to quantify the propagation of cascading
failures and the proximity to a high risk of cascading failure.
This research is joint work with Ben Carreras at Oak Ridge National
Laboratory, Tennessee and David Newman at the University of Alaska.
Time and Place: Wed., Oct. 29, at 3:30 pm in 4610 Engr. Hall.
SYSTEMS SEMINAR WEB PAGE:
http://www.cae.wisc.edu/~gubner/seminar/