Typically, utilities send field crews to visually inspect equipment for damage, which can require significant personnel hours, increasing costs, the chance of worker injuries and the possibility of damaging the surrounding environment when driving or hiking through sensitive habitats. Utilities, including SDG&E, also have used helicopters, which also comes at a cost. Therefore, SDG&E sought approval from the Federal Aviation Administration to use drones for multiple use cases, including assessments. It was one of the first utilities to receive approval in 2014.
Since then, the company’s use of drones has expanded and the utility developed its Drone Investigation, Assessment & Repair (DIAR) program. The DIAR program includes the deployment of numerous highly advanced drones to assess tens of thousands of poles, power lines and equipment and capture millions of images, which are then uploaded to a central database for review by qualified linemen. The advantages of drone inspections over ground inspections were quickly realized, including the top-down and close-up views drones provided. This helped the utility to identify additional issues to repair, leading to a more effective inspection program. However, SDG&E also realized drones could be used for more applications.
“I knew they were collecting the images, so the question was, ‘Could we apply computer-vision models to help us be more efficient long term in our maintenance and fire prevention efforts?’” said Gabe Mika, strategic technology and investment manager for SDG&E. “Could we use the drone shots to build machine-learning models, and make this kind of drone program not just real, but something that had real benefits for our customers?”
Intelligent Image Processing
Analyzing millions of images is an arduous task and takes time. By using artificial intelligence, the digital acceleration team in the Information Technology department was brought in to see if the process could be streamlined further. Through intelligent image processing, SDG&E is now using artificial intelligence and machine learning to automatically detect damaged assets that could lead to an ignition.
“First, machine learning models were created to detect unique asset and damage conditions from the drone program imagery, including insulator, transformer, wooden pole and crossarm damage,” Mika explained. “We involved qualified electrical workers to train the machine models to identify specific equipment damage by drawing a bounding box around an example of damaged equipment on an image. Then, that image is used to teach the machine software to automatically identify the same pattern of damaged equipment in future images.”