Analyzing Cascading Failures and Blackouts Using Utility Outage Data

Historical utility data for cascading failure and large blackouts is foundational for understanding and quantifying blackouts. This chapter surveys some of the main ideas in obtaining and exploiting the patterns in this data, beyond the useful lessons that can be learned from each particular blackout. Historical data on blackout size shows a heavy tailed distribution that implies that large blackouts are both rare and will occasionally occur, and that their risk is substantial. Detailed outage data is routinely collected by utilities and can be processed into cascading events or weather-related events. Metrics for these events can then be readily obtained. Almost all of the research on cascading is based on models and simulation, despite the promising and emerging opportunities to also learn from utility outage data. Some of these opportunities are outlined to encourage further work on real data: As well as its obvious key use to ground models and simulations in reality, the historical outage data can enable better contingency lists, be transformed into influence or interaction graphs, replace simulation by sampling, and describe typical blackouts and their restoration with Poisson processes.