Advertisement

Advertisement

A New Strategy for Maintenance: Artificial intelligence provides benefits for utility operations

By Chuck Ross | Jan 15, 2025
A New Strategy for Maintenance: Artificial intelligence provides benefits for utility operations
Generative A.I., which can create new content by learning patterns from existing data, is energy-intensive, especially for images or videos. But this isn’t a one-way relationship for electric utilities supporting these operations—it turns out, they’re becoming big customers of A.I.

Power-demand projections are going through the roof for data centers focused on artificial intelligence (A.I.) expected to come online in the next 5–10 years. Specifically, generative A.I., which can create new content by learning patterns from existing data, is energy-intensive, especially for images or videos. But this isn’t a one-way relationship for electric utilities supporting these operations—it turns out, they’re becoming big customers. The move can help companies improve operations that could aid efforts to boost clean-energy supplies and reliability.

Goldman Sachs researchers estimate data center power demand will grow by 160% between 2024 and 2030, driven primarily by A.I. Finance, healthcare and marketing are easy targets for A.I. applications that can crunch huge data sets to identify patterns and solutions to complex problems. Increasingly, utilities are also turning to A.I. approaches to address labor-intensive tasks.

Supporting nuclear relicensing

The U.S. nuclear fleet is being given a new life, thanks to the demand for emissions-free energy to power new data centers. In California, the two operating units at the Diablo Canyon nuclear plant were scheduled to shut down in 2024 and 2025, respectively. But the state’s rapidly increasing power demand, and aggressive clean-energy targets, have extended operations by at least five years while owner Pacific Gas & Electric (PG&E) completes the paperwork for a 20-year extension.

Document retrieval is a major challenge for PG&E in maintaining these 40-year-old units—it’s not like nuclear power plants come with a single, handy user manual. Over the decades, many original paper documents have been transferred to microfiche, and then to digital PDFs, with the entire collection now totaling more than a million pages. Indexing is poor, so finding just the records needed to address specific operating issues can mean major work.

The utility has turned to software developer Atomic Canyon, San Luis Obispo, Calif., which trained the Fermi A.I. models used in its Neutron Enterprise software specifically to meet nuclear plant needs, through work with the Department of Energy’s Oak Ridge National Laboratory. Atomic Canyon has spent the last year training its software on the nuclear plant’s documents, which could be spread across multiple Diablo Canyon databases. As a result, workers can access critical information in seconds rather than hours or days. Eventually, the A.I. technology could be used to help with more complex tasks, such as scheduling maintenance, which is complicated in nuclear plants by interconnected systems.

Reducing interconnection gridlock

Getting new clean-energy resources connected to the grid is a slow process. According to the Lawrence Berkeley National Laboratory’s 2024 “Queued Up” report, the process from application to completion typically took almost five years in 2023, up from three years in 2015. The biggest challenge for utilities and grid operators is figuring out if their wires are up to handling more generation resources, which is difficult given the hundreds of applications they might be handling at once.

Several companies are offering A.I.-driven software to ease applicants through the approval process, including error-­checking applications and importing data into grid-modeling platforms that evaluate a new project’s effect on surrounding equipment. It also automatically updates data as other projects join or drop out of the interconnection queue, and it speeds up the process of determining what networks might need upgrading and where new substations might be required—information that’s critical to allocating interconnection costs.

Informing vegetation management

According to researchers with Accenture, New York, electric utilities spend $6 billion to $8 billion every year on maintenance for overhead power lines. Vegetation management is a critical element in these efforts. Traditionally, it’s also been highly labor intensive, requiring on-site visits by personnel able to identify specific risks.

New A.I. tools paired with rapidly developing light detection and ranging technology are helping to automate this process. Using aerial surveillance and photography, which now can be performed using autonomous drones, utilities can identify numbers and species of trees that could be encroaching on their lines. Combining that data with information on historic weather patterns and forecasts, including precipitation levels and other site-specific conditions, these companies also can target management efforts on priority areas most likely to face problems in extreme weather events. This approach could end up being more effective in preventing tree-related fires or damage than traditional approaches cycling inspection crews through a service territory on a 3- or 5-year schedule. 

stock.adobe.com / Lakindu

About The Author

ROSS has covered building and energy technologies and electric-utility business issues for more than 25 years. Contact him at [email protected].

 

Advertisement

Advertisement

Advertisement

Advertisement

featured Video

;

Advantages of Advertising with ELECTRICAL CONTRACTOR in 2025

Learn about the benefits of advertising with Electrical Contractor Media Group in 2025. 

Advertisement

Related Articles

Advertisement