Statistical Estimation of Cascading Blackout Size and Propagation with Branching Processes

Cascading blackouts on a bulk power transmission system are potentially catastrophic events with a large impact on society characterized by a sequence of line trips and load shed. One spectacular example of a cascading blackout is the August 2003 blackout in North America that affected 50 million people. We propose using branching processes to model the line tripping and load shed behavior of cascading blackouts. We use an estimator of the offspring mean, λ, to fit simulated blackout data to the branching process model. The parameter λ is a measure of cascade propagation, and helps us to estimate how likely large blackouts are. We compute distributions of blackout size and total number of line failures from the model and match them against simulated data. The match with simulated data suggests that the branching process model captures important aspects of the cascade phenomenon. The line failures and load shed in cascades are seen to have similar λ, meaning they propagate at the same rate. The branching process model is an efficient way to estimate line failure and load shed probability distributions.