Utilities that piggyback artificial intelligence (AI) into drone power line inspection could save human eyes hundreds of thousands of looks each year, said Zefan Tang, a senior data scientist at Eversource Energy.
“While drones take images efficiently, manual detection of damaged components in each image still requires laborious work,” Tang said.
But that could change soon. Tang will share specifics during his Sept. 14 session, “Drone and AI-enabled Power Line Damage Inspection,” during the T&D World Live Conference in Sacramento, California.
His remarks will include the potential of AI to inspect damaged components of power lines, Tang said.
“I will discuss how images collected by Eversource Energy’s drone team were sorted by types of damage and were divided into training, validation and test data sets, and then were used to train an AI model,” he said. “Test results provide numerous valuable insights to increase the accuracy of AI-enabled power line damage inspection.”
Tang’s session is crucial to the utility industry because it will show how utilities can save engineer man-hours, plus reduce power outages and the number of affected customers.
AI can work most effectively with utility drones following the scenario below, Tang said.
“After trained drone pilots capture images of critical components,” Tang said, “these images are then manually scrutinized to identify and evaluate any damages, sometimes requiring closer examination for finer details. This results in hundreds of thousands of images that must be reviewed annually — a process that is time-consuming and labor-intensive. In addition, training engineers to spot diverse defects within this vast data set poses a further challenge.”
AI, however, can expedite this process significantly, Tang said. A well-trained AI model automatically can detect certain defects and reduce the need for manual examination. Even when the AI model doesn’t achieve perfect performance, he said, it still saves considerable time and effort for engineers. That means well paid and well educated engineers can focus on reviewing the AI results.
“The introduction of AI models to assist our engineers in inspecting the critical infrastructure enables us to swiftly detect potential problems that may jeopardize system reliability,” Tang said. “This allows us to plan maintenance for necessary repairs and upgrades, ultimately enhancing the overall reliability of our power systems. With these measures, we can withstand the winter storm seasons, leading to a reduction in outages and minimizing the number of impacted customers.”
Attendees at Tang’s session should take away the following:
- The potential of AI to inspect damaged components of power lines.
- How to train an AI model using a GPU device running Microsoft Azure.
- How to improve model performance from different aspects, including architecture, loss function and data labeling.
Tang is part of the Data Innovation Team under the Substation & Transmission Line Engineering organization at Eversource.
“I’m mainly focused on data science and machine learning and their applications in substation, distribution and transmission engineering,” he said. “Specifically, I lead data science solution brainstorming, develop power system problem solving strategies, interface with teams across Eversource to understand data landscape, identify opportunities to leverage data streams from different power system groups, develop robust model evaluation criteria, develop classifiers and tools leveraging machine learning models, help develop web applications and support roadmaps for substation data analytics.”
Tang earned a Ph.D. from Stony Brook University in 2021 and served as a postdoctoral researcher at Brookhaven National Laboratory from 2021 to 2022. There, he delved into research projects in the power systems field.
“I conducted extensive research focused on solving practical power system challenges, using cutting-edge technologies in diverse domains, including software-defined networking, computer networks, cybersecurity, machine learning, quantum computing and quantum communication.”
Tang hired on at Eversource in October 2022. He’s eager to talk about sports and pop culture, too.
“In my spare time, I enjoy playing basketball, hiking and watching movies.”
Meet Tang in Sacramento. Register here for the T&D World Live Conference.
Kristen Wright is a journalist with more than 20 years’ experience covering global utilities, petroleum and policy. She is chief strategist at Kristen Wright Strategic Communications. Reach her at [email protected] and wrightkristenm on LinkedIn.
