ComEd's grid analytics team has built a computer vision model that analyzes drone images of wooden utility pole tops and assigns a "health score" to rank poles for replacement or reinforcement. The model has already reached 86 percent accuracy compared to human expert agreement. ComEd relies on a network of more than 1.3 million power distribution poles to deliver power across northern Illinois, and the source describes the approach as an emerging technology with the potential to speed up maintenance processes and reduce potential power outages.
"More utility companies are using image detection models to identify damaged equipment," said Joey Martinez, principal data scientist for ComEd's grid analytics team. "Our pole top AI health ranking model goes a step further by giving us a more actionable list of poles to prioritize for restoration, which improves budgeting and scheduling efficiency."
Traditionally, wooden pole inspections are conducted manually from the ground: field crews measure the circumference of the base of the pole, then perform engineering calculations to determine how much remaining strength it has. One valuable indicator of pole health is the amount of decay around the center of the very top of the pole, which is only visible from a bird's eye view, and drones help efficiently gather images of it.
Field crews began capturing drone imagery of the pole tops in 2024, and the grid analytics team completed the computer vision models in early 2025. To train the models, the team input data from thousands of expert image comparisons provided by field engineers, based on the question "which pole is worse?" This creates a scalable AI that thinks like the best pole inspectors. Beyond detecting pole tops that are damaged or decayed, the model assigns the health score used for ranking.
Forensic research has been introduced to the initiative. Six retired poles, identified by the pole top AI health model, were sent to the ComEd forensics lab to be sectioned and analyzed to identify the area of the pole with the most decay. For aging poles with less damage, a C-Truss reinforcement can be installed to increase lifespan, and the forensics research aims to better inform how and when these reinforcements can be used.
The project will be further enhanced as additional pole top imagery is collected and incorporated with forensics insights, which ComEd says will help strengthen the AI models and improve prioritization of pole replacements.