ENTERPRISE AI ANALYSIS
Modeling Adversarial Wildfires for Power Grid Disruption
This research introduces a novel mixed-integer conic program to model adversarial wildfires that specifically target power grid infrastructure, ensuring realistic fire spread dynamics while identifying worst-case outage scenarios. Our approach enhances power grid resilience planning by integrating robust optimization with empirical fire spread models, offering a unique methodology for characterizing worst-case fire behavior and its impact on critical infrastructure.
Executive Impact & Key Findings
Our analysis uncovers critical vulnerabilities and quantifies the severe economic and operational impacts of strategically targeted wildfires on power infrastructure.
0
Estimated Wildfire Costs for California Grid (2001-2016)
0
Customers Affected by 2017 Thomas Fire Outages
0
Increase in Load Shed for Park Fire with Sequential Outages
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
The Core Optimization Problem
Our work formulates a novel mixed-integer conic program (MICP) designed to characterize an adversarial wildfire. This model optimizes over uncertain wind velocities, ignition points, and fire spread sequences to identify scenarios that cause the most severe impact on power grid infrastructure. By focusing on worst-case outcomes, the model provides crucial insights for robust resilience planning.
Realistic Fire Spread Dynamics
The model incorporates realistic fire spread dynamics based on the Rothermel fire spread model (1972), adapted for two-dimensional scenarios. We introduce wind-assisted fire spread sets and novel convex relaxations, including a second-order conic (SOC) relaxation for inner products over Euclidean balls, to ensure computational tractability while maintaining model accuracy for fire propagation paths.
Security-Constrained Optimal Power Flow
To assess wildfire impact on power delivery, we integrate our adversarial wildfire model with a security-constrained direct-current (DC) optimal power flow (OPF) model. This integrated framework minimizes generation costs and unmet demand, considering element outages identified by the wildfire model. It enables robust operational decisions and encourages resilient grid configurations against realistic contingencies.
California Wildfire Case Studies
We validate our model using experimental settings aligned with recent large California wildfires: the 2024 Park fire, and the 2025 Eaton and Palisades fires. These scenarios leverage a detailed 3,928-bus power system model calibrated with real-world geographical and temporal data, demonstrating the model's ability to characterize minimum time-to-outage and maximum load shed under realistic conditions.
Our Novel Contributions to Wildfire & Power Grid Resilience
Formulate novel MICP for adversarial wildfire impact
→
Propose 2D wind-assisted fire spread set & conic relaxations
→
Design experimental settings aligned with California wildfires & calibrate model
→
Apply adversarial model to characterize min time-to-outage & max load shed
→
Use min time-to-outage for contingency screening in security-constrained OPF
$700M+
Estimated cost for California grid due to wildfires (2001-2016)
| Parameter Relation | B << 1 | B ~ 1 | B >> 1 |
|---|---|---|---|
| C·WB ~ 1 | Inner Product (S◈) | Inner Product (S◈) | Ball (S◦) |
| C·WB >> 1 | Inner Product (S◈) | Ball (S◦) | Ball (S◦) |
The choice of convex relaxation for the Rothermel spread set (S*) depends on the relationship between the exponent B, the coefficient C, and the maximum wind speed W. For high wind influence (CWB >> 1), the ball spread set (S°) is generally preferred, while for balanced influence (CWB ~ 1) or low B, the inner product spread set (S◈) offers better accuracy.
Real-World Impact: California Wildfire Scenarios
We designed experimental settings to mimic real California wildfires: the 2024 Park fire, and the 2025 Eaton and Palisades fires. These scenarios leverage a 3,928-bus power flow model and calibrated wildfire spread dynamics to assess actual grid resilience. This includes using hourly temperature, renewable generation availability, and hydroelectric data to simulate power grid behavior during these events.
Impact: Our model identifies critical outage sequences and minimum time-to-outage, providing data for resilient power grid operation and contingency planning against these specific, high-impact events. For instance, the Park fire simulations showed up to a 21x increase in load shed when considering sequential outages, highlighting the importance of dynamic modeling. Optimal wind vectors and ignition points were identified to maximize harm, demonstrating the model's adversarial capability and informing proactive mitigation strategies.
21x
Increase in load shed for Park fire with sequential outages
Calculate Your Potential ROI
Estimate the efficiency gains and cost savings your enterprise could achieve by implementing AI-driven resilience strategies.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A typical journey to integrating AI-driven insights for enhanced power grid resilience.
Discovery & Strategy
Comprehensive assessment of existing infrastructure, data sources, and operational challenges. Define specific resilience objectives and an AI integration strategy tailored to your utility's needs.
Data Integration & Model Calibration
Consolidate and prepare disparate data (weather, fuel, grid topology). Calibrate and validate wildfire spread and power flow models using historical and real-time data for accuracy.
Adversarial Model Deployment
Integrate the adversarial wildfire model to simulate worst-case scenarios, identify critical contingencies, and determine optimal operational responses and hardening investments.
Testing & Validation
Rigorous testing of the AI system's performance in simulated environments. Validate its predictions against a range of disruption scenarios and stress tests.
Continuous Optimization & Support
Ongoing monitoring, model refinement, and integration of new data for continuous improvement. Provide training and support for your operational teams.
Ready to Enhance Your Grid's Resilience?
Our experts are ready to discuss how AI-driven adversarial modeling can protect your critical power infrastructure from unpredictable wildfires.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat