Skip to main content
/ Smart Grid Center > Awards, Scholarships > New DOE CESER Project “Preventing Wildfire Ignition from Powerline Equipment Failures Using ML-Based Classification of Real-Time Electrical Measurements”

New DOE CESER Project “Preventing Wildfire Ignition from Powerline Equipment Failures Using ML-Based Classification of Real-Time Electrical Measurements”

Powerline-caused wildfires can occur when powerline equipment fails, creating ignition mechanisms. These mechanisms include melting devices, burning particles dropping from arcing parts, and lines down due to component failure. Research has shown that many equipment failures slowly develop over days or weeks prior to catastrophic failure that can ignite a fire.

The Office of Cybersecurity, Energy Security, and Emergency (CESER) of Department of Energy (DOE) has awarded a new 3-year research project titled “Preventing Wildfire Ignition from Powerline Equipment Failures Using ML-Based Classification of Real-Time Electrical Measurements” with total cost of nearly $3.2M. The principal investigator (PI) is B. Don Russell, Ph.D., Regents professor and holder of the Harry E. Bovay, Jr. Chair in the Department of Electrical and Computer Engineering (ECE) at Texas A&M University (TAMU), and the lead organization is Texas A&M University Engineering Experiment Station (TEES) and titled “Preventing Wildfire Ignition from Powerline Equipment Failures Using ML-Based Classification of Real-Time Electrical Measurements”. The Co-PIs are Mr. Carl Benner, Dr. Jeff Wischkaemper, and Dr. Karthick Mannivannan (ECE, TAMU).

Prior work by the research team has shown that incipient failure signals of equipment can be detected in electrical measurements. This project will use ML-based techniques to enhance and identification and classification of failing devices. This project leverages more than 25 years of Texas A&M research that has studied powerline-caused wildfire ignitions extensively. The team previously instrumented more than 600 feeders on 20 utilities, developing a database of more than feeder-years of waveform captures measured in the millions of events. This unique dataset will enable ML-software development.

Researchers with the Power System Automation Laboratory (ECE, TAMU), in concert with engineers at Pacific Gas and Electric Company, will join to design and field test new ML-based failure detection. Existing Distribution Fault Anticipation (DFA) monitoring platforms currently installed on 75 feeders at PG&E will be used in the demonstration phase of the project. The new software will be evaluated for overall performance, using real events as the occur on PG&E feeders.

The benefits of this project are numerous. By keeping feeders healthy, the reliability of delivering power to the public can be increased during normal operations. However, the failing devices that cause customer outages are often ignition mechanisms for wildfires. By finding and fixing these failing components, wildfires can be prevented while feeder reliability is improved.

The DOE announcement can be viewed here.