D’Souza Pens Science Perspective on Cascading Failures

Raissa D’Souza, professor of computer science and mechanical and aerospace engineering at UC Davis, recently wrote perspective article for Science about cascading failures that arise from the reorganization of flows on a network, such as in electric power grids, supply chains and transportation networks. Photo: ISTOCK.com/YangPhoto

By Aditi Risbud Bartl

DAVIS, Calif.; Nov. 20, 2017 – In last week’s issue of Science, Raissa D’Souza, professor of computer science and mechanical and aerospace engineering at UC Davis, wrote a perspective article about cascading failures that arise from the reorganization of flows on a network, such as in electric power grids, supply chains and transportation networks.

Researchers are studying ways to prevent electric power outages, particularly those that span large geographic regions impacting massive numbers of people. Such outages are usually due to cascading events in which a localized outage puts strain on another part of the network, causing it to go down and creating a “domino effect” of widespread outages.

In her Science perspective, D’Souza reviews the challenges researchers face in studying cascading failures using existing models for such events. Although theoretical mathematical models provide a big picture understanding, they miss out on specific details. In comparison, more detailed computer simulation models neglect secondary effects and are computationally expensive.

In the study D’Souza’s perspective accompanies, researchers at Northwestern University bridged the divide between these models using a simulation for analyzing the vulnerability of a continent-wide power system that captures the underlying physics of power flow, many standard operating practices and cascading failures caused by power lines overheating.

To create an accurate assessment of the North American power grid, which includes the U.S. and most of southern Canada, the Northwestern researchers obtained data from the U.S. Federal Energy Regulatory Commission from 2008 to 2013. These data were used to create a model showing the entire energy network of more than 100,000 power lines.

Their results showed about eleven percent of all transmission lines in the network were vulnerable to an overload failure due to cascading events, and 85 percent of all primary failures (ones that could propagate a cascading event) were located in just 20 percent of those parts of the network deemed vulnerable, and were biased to be closer to large urban areas.

These findings suggest that putting more effort into reducing vulnerabilities in just a few parts of the network could dramatically reduce the chances of any part of the network experiencing a major cascading event, and could be used by electrical planners to prevent future network outages. Pinpointing which elements are actually the vulnerable ones varies dramatically with the state and operating conditions of the grid.

“These insights for the power grid should have wider-reaching implications,” D’Souza noted in the perspective piece. “A comprehensive theory of cascading failures across domains will require consideration of which quantities are conserved in the load-shedding process. The flow could be made of electrons in the power grid, packets on the internet, people on an airline network, or goods in supply chains, leading to vastly different constraints. Understanding how the different constraints and costs lead to different classes of cascading failures will be a key to eventually designing control interventions for avoidance and rapid recovery.”

Read the full perspective article here, and learn more about D’Souza’s research at UC Davis here.