Large-scale data analytics for resilient recovery services from power failures

Abstract

The large-scale data enable analytics on failure events of different severity induced by weather disruptions. The data analytics systematically studies recovery services under different severity of failure impact. A recovery scaling law is developed through unsupervised learning, showing capabilities and limitations of recovery as well as promise for enhancement and generalization to multiple states.

Resilience of renewable power systems under climate risks

“…It highlights the risks of failure propagation within grids during extreme events, due to the connected network topology and network dynamics. In terms of recovery from these outages, data analysis also exhibits a similar 10-90 scaling law pattern, where about 10% of disrupted customers account for nearly 90% of total customer interruption hours due to delayed restoration 44,45 . Moreover, as the severity of weather-induced events escalates from moderate to extreme levels, the effectiveness of rapid recovery degrades by nearly 30%, with more customers suffering from prolonged disruptions 45 .…”

“…In terms of recovery from these outages, data analysis also exhibits a similar 10-90 scaling law pattern, where about 10% of disrupted customers account for nearly 90% of total customer interruption hours due to delayed restoration 44,45 . Moreover, as the severity of weather-induced events escalates from moderate to extreme levels, the effectiveness of rapid recovery degrades by nearly 30%, with more customers suffering from prolonged disruptions 45 . These scaling law patterns with strong nonlinearity consistently indicate that extreme weather events significantly exacerbate the vulnerability of power systems and underscore the urgent need to focus more on the effects of these less-frequent but extreme events on system resilience.…”