Fragmentation of outage clusters during the recovery of power distribution grids

Hao Wu, Xiangyi Meng, Michael M. Danziger, Sean P. Cornelius, Hui Tian, Albert-László BarabásiNature Communications 2022 · OpenReview

TL;DR

This paper reveals that power grid recovery isn't random, but follows a universal pattern influenced by "blackout intensity" (peak number of outages) and the recovery status of nearby outages, not the number of affected customers. They propose a novel "cluster-based recovery framework" that models this fragmentation process, offering a promising avenue for speeding up restoration by focusing on mitigating blackout intensity.

What Problem Does This Paper Address?

Understanding how power distribution grids recover from outages, especially large-scale blackouts, has been hampered by a lack of detailed, real-world data. Existing research often focuses on transmission grids or uses simplified models, which don't fully capture the unique tree-like structure and vulnerabilities of distribution networks. This makes it difficult to predict recovery times and develop effective strategies to minimize downtime. The authors aim to bridge this gap by analyzing realistic outage data and developing a theoretical framework to explain and predict recovery behavior.

Methods

The core of the paper's methodological contribution is the development of a cluster-based recovery framework. This framework models the recovery process as a fragmentation process of outage clusters. When a component is repaired, it breaks "dependence links" between outages. As these links are removed, larger outage clusters break down into smaller ones, eventually leading to individual outages being repaired. This process is analogized to depolymerization in materials science, where a polymer chain breaks down.

The authors also employ statistical analysis to identify key factors influencing recovery duration. They fit probability density functions to recovery times and analyze the correlation between blackout intensity and average recovery hours.

Datasets

The study utilizes a comprehensive dataset from three electric companies in the United States: National Grid, Eversource, and Entergy. This dataset covers outages in Massachusetts, New York, and Texas from November 2018 to April 2020, containing a total of 682,733 outages. This dataset is notable for its granularity and scale, including both daily operations and tens of large blackouts (defined as having an intensity > 100 outages).

Key Insights and Ideas

  • Recovery is not random: The paper strongly suggests that power grid recovery follows predictable patterns, not just random chance.
  • Blackout Intensity is King: The peak number of simultaneous outages during a blackout (blackout intensity) is a dominant factor in determining recovery duration. Higher intensity leads to longer recovery times.
  • Neighbors Matter: The recovery time of an outage is significantly influenced by the recovery status of its spatially nearest outages. This highlights the interconnectedness of the distribution grid.
  • Customer Count is Misleading: Surprisingly, the number of customers affected by an outage does not appear to be a significant predictor of its recovery duration. This challenges common assumptions.
  • Universal Recovery Pattern: The study identifies a universal recovery behavior across different utilities, geographical regions, and disruption causes, suggesting a fundamental dynamic at play.
  • Fragmentation as a Model: The cluster-based recovery framework, inspired by polymer fragmentation, provides a powerful analytical tool for understanding and simulating this universal recovery pattern.

Key Results

  • The probability density function of recovery duration for individual outages can be well-fitted by an inverse Weibull distribution.
  • The average repair time for outages increases logarithmically with blackout intensity (N).
  • A strong positive correlation exists between the average recovery duration of an outage and the average recovery time of its nearest neighboring outages.
  • The proposed cluster-based recovery framework accurately reproduces the observed exponential survival function of outages during blackouts, aligning well with real data.
  • The study demonstrates that suppressing blackout intensity is a promising strategy to speed up restoration.

Limitations and Future Work

  • Data Granularity: While the dataset is extensive, the authors note that obtaining high-resolution, long-term tracking data for outages remains a challenge for future research.
  • Topology Independence: The cluster-based model is designed to be topology-independent, which is a strength for generalization but might miss nuances specific to certain grid configurations.
  • Aggregation-Fragmentation Dynamics: The current model focuses on fragmentation. The authors suggest extending it to a more general aggregation-fragmentation form to better reflect the dynamics of failure propagation and restoration occurring simultaneously.
  • Interdependent Networks: The paper hints at extending the concept of interdependence to other distributed networks beyond power grids, suggesting a broader applicability of their findings.
  • Practical Implementation: While the findings suggest suppressing blackout intensity, the paper acknowledges that the "function reflecting the impact of those factors on blackout intensity is of vital importance, yet pretty hard to capture." Future work could focus on developing such functions and translating these insights into actionable engineering strategies.

Summary updated

Abstract

The understanding of recovery processes in power distribution grids is limited by the lack of realistic outage data, especially large-scale blackout datasets. By analyzing data from three electrical companies across the United States, we find that the recovery duration of an outage is connected with the downtime of its nearby outages and blackout intensity (defined as the peak number of outages during a blackout), but is independent of the number of customers affected. We present a cluster-based recovery framework to analytically characterize the dependence between outages, and interpret the dominant role blackout intensity plays in recovery. The recovery of blackouts is not random and has a universal pattern that is independent of the disruption cause, the post-disaster network structure, and the detailed repair strategy. Our study reveals that suppressing blackout intensity is a promising way to speed up restoration. Here the authors, by using data from three electrical companies in the USA, find that the recovery duration of an outage is connected with the downtime of its nearby outages and blackout intensity and present a cluster-based recovery framework.
BibTeX
@inproceedings{wu2022fragmentation,
  author = {Hao Wu and Xiangyi Meng and Michael M. Danziger and Sean P. Cornelius and Hui Tian and Albert-László Barabási},
  title = {{Fragmentation of outage clusters during the recovery of power distribution grids}},
  booktitle = {Nature Communications},
  year = {2022},
  url = {https://openreview.net/forum?id=IOcntPFAoT},
}
Figure