Overview

This project will provide wildfire managers and utility operators with an AI-driven wildfire risk model to prevent wildfires caused by power lines. The model advances the state-of-the-art by integrating data on fire-ignition electrical faults with high-resolution environmental information, including fire weather, vegetation, fuel conditions, and topography, sourced from in-situ and NASA space-based sensors. Ultimately, the project will offer a turn-key solution, enabling near real-time assessments of electrical grid features that may pose an increased wildfire risk.

Science Area

Failed or faulty power lines are among the most common causes of severe wildfires, including the 2018 Camp Fire – the deadliest, most expensive wildfire in California’s history. Effective wildfire risk assessment from powerlines requires superior models to capture the dynamics and mechanisms of electrical ignition, along with the various factors contributing to fires triggered by electrical faults. Novel AI-based tools are also needed to detect fire-ignition electrical faults and update high-resolution wildfire risk maps in near real-time. This project fills this technology gap, improving our ability to detect hazardous pre-fire conditions before a burn occurs.

Technology

The team will conduct ignition tests using a medium-voltage rig connected to a grid emulator. They will create a comprehensive database on the dynamics of electrical faults, which will inform an AI-based power line situational awareness tool for estimating ignition probabilities. The team will use high-resolution multispectral data from PlanetScope CubeSats to refine fuel models and will develop new AI-based fire danger indices, integrating spatial data on fuel types, vegetation biomass, and greenness to enhance the accuracy of electrical wildfire risk assessments. Finally, the team will test their software during a field campaign in a fire-prone area near power lines.

Advancements

• State-of-the-art ML techniques will equip wildfire managers with an AI-based powerline situational awareness tool for diagnosing and predicting fire-ignition electrical faults and quantifying ignition probability.
• Powerline ignition experiments will create a database that helps wildfire managers and power system operators identify the conditions, dynamics, and signatures associated with electrical ignition.
• Deep learning algorithms will enable an advanced fire danger model, utilizing high-resolution spatial and temporal multispectral data from PlanetScope CubeSats and a ML-based fuel property model.

Principal Investigator

Hamid Nazaripouya is an Assistant Professor of Electrical and Computer Engineering at Oklahoma State University (OSU). His research lies in the broad area of power systems and power electronics. In particular, he is interested in enhancing automation, stability, and resilience of power systems by creating remote sensing platform, advanced data analytics, and control systems for cyber-physical energy systems as well as integration of renewable energy resources, energy storage, power electronics, and smart grid technologies.

Nazaripouya is assisted by Hamed Gholizadeh, Haejun Park, Nick Shumaker, and Jia Yang (OSU); Drew Daily (Oklahoma Department of Agriculture); and Reginald Freeman (Freeman Group, Inc.).