Sandia and DOE's AI-Based Protective Relay — Isolating Wildfire-Igniting Grid Faults 100 Times Faster (An Early-Stage Case)

Wildfires in the United States increasingly start when faults in ageing electrical equipment — a downed line, a failing connector — arc and throw sparks into dry vegetation. The U.S. Department of Energy's CESER office and Sandia National Laboratories are developing an AI-based protective relay to detect and isolate such faults before they can ignite a fire.

Detect and isolate grid faults fast enough to prevent them from igniting wildfires, as part of Sandia's wider Wildfire Electric Grid Resilience programme.

The programme's 'Instant Arc Detection' component analyses high-speed sensor data to locate and isolate faults. The wider programme also includes PyroKit (wildfire risk-mapping and Public Safety Power Shutoff planning, now in customer discovery with utilities), SMOKEWISE (modelling how wildfire smoke reduces solar generation), and a publicly hosted Grid Event Signature Library of anonymised voltage and current data for other utilities and researchers to train their own fire-detection models. Utility partner Public Service Company of New Mexico (PNM) is named, with field testing on utility distribution systems still in the planning stage as of August 2025.

In prototype testing on a microgrid, the Instant Arc Detection system located and isolated a fault approximately 100 times faster than conventional protective equipment, with detection achieved in under four milliseconds.

This is an early-stage federal research programme rather than a proven operational system: the headline performance figures come from laboratory and prototype testing, not a completed field deployment, and real-world fault-detection accuracy under actual grid conditions remains to be demonstrated.