Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Real-Time Risky Fault-Chain Search using Time-Varying Graph RNNs

Anmol Dwivedi, Ali Tajer

TL;DR

The paper addresses the challenge of real-time risk assessment for cascading outages in power grids facing climate-driven extreme events. It formulates the risky fault-chain search as a partially observable Markov decision process and solves it with a time-varying Graph Recurrent Neural Network (GRNN) that leverages grid topology and sequential observations via a Graph Recurrent Q-Network (GRQN). The approach captures spatio-temporal dependencies while maintaining scalability through latent-state summarization and parameter sharing, enabling efficient discovery of high-risk fault chains. Empirical results on IEEE-39 and IEEE-118 bus systems show improved accumulated risk (TLL) and reduced regret compared to baselines, under both unbounded and bounded compute budgets, suggesting practical potential for real-time grid resilience analytics and planning. The authors also provide public code, facilitating reproducibility and deployment in real-world settings.

Abstract

This paper introduces a data-driven graphical framework for the real-time search of risky cascading fault chains (FCs) in power-grids, crucial for enhancing grid resiliency in the face of climate change. As extreme weather events driven by climate change increase, identifying risky FCs becomes crucial for mitigating cascading failures and ensuring grid stability. However, the complexity of the spatio-temporal dependencies among grid components and the exponential growth of the search space with system size pose significant challenges to modeling and risky FC search. To tackle this, we model the search process as a partially observable Markov decision process (POMDP), which is subsequently solved via a time-varying graph recurrent neural network (GRNN). This approach captures the spatial and temporal structure induced by the system's topology and dynamics, while efficiently summarizing the system's history in the GRNN's latent space, enabling scalable and effective identification of risky FCs.

Real-Time Risky Fault-Chain Search using Time-Varying Graph RNNs

TL;DR

The paper addresses the challenge of real-time risk assessment for cascading outages in power grids facing climate-driven extreme events. It formulates the risky fault-chain search as a partially observable Markov decision process and solves it with a time-varying Graph Recurrent Neural Network (GRNN) that leverages grid topology and sequential observations via a Graph Recurrent Q-Network (GRQN). The approach captures spatio-temporal dependencies while maintaining scalability through latent-state summarization and parameter sharing, enabling efficient discovery of high-risk fault chains. Empirical results on IEEE-39 and IEEE-118 bus systems show improved accumulated risk (TLL) and reduced regret compared to baselines, under both unbounded and bounded compute budgets, suggesting practical potential for real-time grid resilience analytics and planning. The authors also provide public code, facilitating reproducibility and deployment in real-world settings.

Abstract

This paper introduces a data-driven graphical framework for the real-time search of risky cascading fault chains (FCs) in power-grids, crucial for enhancing grid resiliency in the face of climate change. As extreme weather events driven by climate change increase, identifying risky FCs becomes crucial for mitigating cascading failures and ensuring grid stability. However, the complexity of the spatio-temporal dependencies among grid components and the exponential growth of the search space with system size pose significant challenges to modeling and risky FC search. To tackle this, we model the search process as a partially observable Markov decision process (POMDP), which is subsequently solved via a time-varying graph recurrent neural network (GRNN). This approach captures the spatial and temporal structure induced by the system's topology and dynamics, while efficiently summarizing the system's history in the GRNN's latent space, enabling scalable and effective identification of risky FCs.

Paper Structure

This paper contains 31 sections, 16 equations, 4 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Graphical Risky Fault-Chain Search Framework
  4. Experiments
  5. Conclusion
  6. Appendix
  7. POMDP Modeling
  8. State Space $\mathcal{S}$:
  9. Action Space $\mathcal{A}$:
  10. Transition Kernel ${\mathbb {P}}$:
  11. Reward Dynamics $\mathcal{R}$:
  12. Time-Varying GRNN Model
  13. Time-Varying GCNNs:
  14. Time-Varying GRNNs:
  15. Architecture, Training, Algorithm Initialization
  16. ...and 16 more sections

Figures (4)

  • Figure 1: Performance comparison for the IEEE-$39$ bus system under unbounded computational budget.
  • Figure 2: An agent decision process rendering a FC $\mathcal{V}^*_{s} = \langle \ell^{1}_{1}, \ell^{2}_{2}, \ell^{1}_{3}\rangle$.
  • Figure 3: Graph recurrent $Q$-network $({\sf GRQN})$ architecture.
  • Figure 4: ${\sf Regret}(s)$ versus $s$ for the IEEE-$118$ bus system.