At this year’s Energy Drone & Robotics Summit, Clint Palermo and Joey Martinez took the stage to present a fascinating case study on how ComEd is revolutionizing grid analytics through the use of drone inspections and AI-driven computer vision. This initiative showcases the intersection of technology and utility management, aimed at automating and optimizing the inspection of vast power grid networks.

The Challenge: Managing Grid Infrastructure

ComEd, the largest electric utility provider in northern Illinois, serves over 4.2 million customers and oversees 35,000 miles of overhead power lines. Managing and maintaining this enormous infrastructure, especially in a time of extreme weather, regulatory demands, and increasing reliability standards, poses a monumental challenge.

One of the most pressing issues ComEd faces is the complexity of gathering and processing massive amounts of data. Their network includes over 1.2 million utility poles, and collecting data—be it images or measurements—often results in scattered repositories across various teams and departments. This lack of centralized data storage and accessibility hampered their ability to efficiently review images and make informed maintenance decisions.

The Solution: Building a Centralized AI-Powered Platform

In response to these challenges, ComEd embarked on a project to build a platform designed to centralize and optimize data collection. By integrating drones, cameras, and AI algorithms, they created a system that accelerates the inspection process, reduces manual labor, and enhances grid reliability.

This new platform brings all image data under one roof, making it accessible to all employees, regardless of their department. As Joey Martinez explained, “We built a platform for creating more efficiency and effectiveness in data collection, enabling AI to help us with inspections and deliver insights much faster.”

How It Works: AI-Driven Inspections

The inspection process follows a comprehensive pipeline where data is ingested, stored, and prepped for analysis. ComEd uses drones and cameras to capture images of power poles, then labels these images to train their AI algorithms. To date, they have labeled over 11,000 images, creating a robust AI capable of detecting various components and potential defects on power poles.

The AI operates through a multi-step process. Initially, it determines whether a pole is present in an image. Once confirmed, it zooms in on specific components, such as crossarms, insulators, and lightning arresters, which are critical to maintaining uninterrupted service. Since these are the components most likely to fail, the AI focuses on detecting potential issues like damage or wear.

Impressive Accuracy and Continuous Improvement

The AI system has proven highly accurate, with a detection accuracy rate ranging from 68% to 90%, depending on the component. Common elements like crossarms and insulators achieve higher detection rates, while less frequently encountered items, such as lightning arresters, have lower but still respectable accuracy rates.

Crucially, the platform is designed to continuously improve. Maintenance inspectors regularly review AI outputs, correcting any inaccuracies. These corrections are then fed back into the system to retrain the AI, enhancing its accuracy over time.

Expanding the Use Cases: Vegetation Management and Third-Party Audits

In addition to inspecting power poles, ComEd is expanding the platform’s capabilities to tackle other challenges. For instance, the utility uses LiDAR data to monitor vegetation near power lines, ensuring that tree limbs don’t interfere with infrastructure. 

They are also working on automating third-party audits for attachments on utility poles from companies like AT&T—a process that currently requires expensive physical audits. Martinez said, “We hire a contract company to go and physically count how many lines are on a pole. That's a five-year cycle project, and it's very expensive now. We are improving to make a new AI case where we can count the number of lines and automate that system instead of having to contract that out.”

The Future of AI in Utility Management

As Martinez highlighted during the presentation, while ComEd is making great strides in automating inspections and defect detection, they are not planning to fully remove human oversight. “We will always have a human in the loop to verify whether the AI got it right,” he said. The goal is to enhance efficiency without sacrificing safety or reliability.

With ongoing developments, ComEd aims to automate more aspects of asset defect detection and continue refining its AI models. From improving outage prevention to reducing maintenance costs, this case study illustrates the transformative potential of integrating AI and drone technology into grid management.

Want More?

You can watch this full session (and other Energy Drone & Robotics Summit sessions) on-demand. 

And make sure you have next year’s summit on your calendar. Join us June 16-18, 2025 in Houston, Texas, for the latest in emerging tech in energy.