Reinforcement-Learning-Based Proactive Control for Enabling Power Grid Resilience to Wildfire

Salah Uddin Kadir, Subir Majumder, Anurag K. Srivastava, Ajay Dev Chhokra, Himanshu Neema, Abhishek Dubey, Aron Laszka

IEEE Transactions on Industrial Informatics 2024

PDF DOI

Why This Matters

Traditional power system operations rely on myopic load dispatch and post-event remedial actions, which are insufficient for managing cascading failures during wildfires. This work is innovative because it combines machine learning with physics-based wildfire propagation models to enable proactive, anticipatory decision-making. Rather than reacting after failures occur, the approach allows operators to preemptively adjust power flows and resources, representing a significant advance in how autonomous systems can handle extreme events in critical infrastructure.

What We Did

This work develops a reinforcement learning-based proactive control approach for power grid resilience during wildfire events. The researchers model the power system and wildfire propagation using a detailed integrated testbed that captures both the spatial dynamics of fire spread across a geographical grid and the operational constraints of the power network. A deep reinforcement learning agent is trained to make real-time decisions about generator control and load management during extreme weather events, coordinating with multiple microgrids and transmission systems to minimize load shedding and outages.

Key Results

The proposed approach successfully reduces power loss through increased power flow rerouting during wildfire events compared to baseline myopic control policies. The integrated testbed demonstrates that the RL-based controller can provide decision support to operators while maintaining computational tractability. Testing on a realistic IEEE power system mapped to geographical terrain shows that proactive control achieves substantial improvements in reducing load outages and providing resilience, with the ability to be deployed in real-time alongside human operators.

Full Abstract

Cite This Paper

@article{proactivewildfire,
  author = {Kadir, Salah Uddin and Majumder, Subir and Srivastava, Anurag K. and Chhokra, Ajay Dev and Neema, Himanshu and Dubey, Abhishek and Laszka, Aron},
  journal = {IEEE Transactions on Industrial Informatics},
  title = {Reinforcement-Learning-Based Proactive Control for Enabling Power Grid Resilience to Wildfire},
  year = {2024},
  issn = {1941-0050},
  month = {jan},
  number = {1},
  pages = {795-805},
  volume = {20},
  abstract = {Industrial electric power grid operation subject to an extreme event requires decision making by human operators under stressful conditions. Decision making using system data informatics under adverse dynamic events, especially if forecasted, should be supplemented by intelligent proactive control. Power transmission system operation during wildfires requires resiliency-driven proactive control for load shedding, line switching, and resource allocation considering the dynamics of the wildfire and failure propagation to minimize the impact on the system. However, the possible number of line and load switching in an extensive industrial system during an event make the traditional prediction-driven and stochastic approaches computationally intractable, leading operators to often use preplanned or greedy algorithms. In this article, we model and solve the proactive control problem as a Markov decision process (MDP) and introduce an integrated testbed for spatiotemporal wildfire propagation and proactive power-system operation. Our approach allows the controller to provide setpoints for all generation fleets in the power grid. We evaluate our approach utilizing the IEEE test system mapped onto a hypothetical terrain. Our results show that the proposed approach can help the operator to reduce load outage during an extreme event. It reduces power flow through lines that are to be de-energized and adjusts the load demand by increasing power flow through other lines.},
  contribution = {colab},
  doi = {10.1109/TII.2023.3263500},
  keywords = {power grid resilience, reinforcement learning, wildfire propagation, proactive control, critical infrastructure, machine learning, emergency response, optimization},
  month_numeric = {1}
}
Quick Info
Year 2024
DOI 10.1109/TII.2023.3263500
Keywords
power grid resilience reinforcement learning wildfire propagation proactive control critical infrastructure machine learning emergency response optimization
Research Areas
energy emergency scalable AI CPS ML for CPS
Search Tags

Reinforcement, Learning, Proactive, Control, Enabling, Power, Grid, Resilience, Wildfire, power grid resilience, reinforcement learning, wildfire propagation, proactive control, critical infrastructure, machine learning, emergency response, optimization, energy, emergency, scalable AI, CPS, ML for CPS, 2024, Kadir, Majumder, Srivastava, Chhokra, Neema, Dubey, Laszka