Wildfires pose an escalating threat to electrical grid infrastructure and reliability, particularly in fire-prone regions where rapid operational decisions must account for uncertain and evolving conditions. A new research framework addresses this challenge by integrating decision-dependent uncertainty into grid operations planning.
The proposed model represents a departure from traditional optimization approaches that treat uncertainties as independent variables. Instead, it recognizes that operational decisions made early in a wildfire event—such as implementing Public Safety Power Shutoff (PSPS) protocols—fundamentally alter the probabilities and parameters of subsequent wildfire impacts. This decision-dependent uncertainty reflects real-world dynamics where preventive grid de-energization affects fire behavior, customer impacts, and system recovery needs.
The framework operates across multiple decision stages, enabling utilities to adjust strategies as wildfire scenarios evolve. Preventive measures include load shedding and equipment disconnections before fires reach critical infrastructure. Corrective measures address real-time contingencies as threats materialize. Mathematical decomposition algorithms solve the complex optimization problem efficiently, making the approach computationally practical for operational use.
Testing on the IEEE 30-bus test system demonstrates that ignoring decision-dependent uncertainty leads to suboptimal operational plans. By accounting for how PSPS decisions influence future wildfire probabilities, the framework generates more resilient strategies that better protect both grid assets and customer service. The approach balances competing objectives: minimizing customer outages while reducing wildfire-related equipment damage and cascade failures.
As climate change intensifies wildfire seasons, utilities require decision-support tools that reflect the complex interdependencies between grid operations and environmental hazards. This framework provides a mathematical foundation for more robust operational planning under extreme weather events. Future work may extend the approach to larger network topologies and incorporate real-time weather data integration for dynamic implementation.
