AI poised to become integral to utility grid maintenance, reliability

Artificial intelligence will increasingly be integrated into utility operations, assisting with forecasting maintenance and improving grid reliability, experts from Microsoft Corp., eSmart Systems AS and CreatorsAGI said Sept. 12.

"Over the next two to five years, you are really going to see a mass adoption of AI and AI becoming part of the utility industry's DNA," Bilal Khursheed, worldwide power and utilities executive director at Microsoft, said during a Sept. 12 webinar hosted by Utility Dive.

The panelists acknowledged that while utility operations have long relied on predictive maintenance models and forecasts, AI's ability to sift through massive amounts of data could enable better preparation for infrastructure maintenance, repair and vegetation management.

Erik Asberg, chief technology officer of eSmart, a Norwegian company that provides AI services to utilities, said AI can be used to listen to the sound of turbines to detect any signs of failure.

"Because now AI systems are able to analyze images and data, now people are analyzing satellite data to manage vegetation better. This has led to some utilities cutting outages by half, because you know where overgrowth is," Joseph Sirosh, CEO of CreatorsAGI, said in the webinar. CreatorsAGI works with companies and individuals to build AI assistants.

Sirosh, who previously served in AI-focused roles at Microsoft and Amazon.com Inc., pointed to a program by San Diego Gas & Electric Co. as an example of how a utility can utilize AI to improve reliability. SDG&E uses an AI modeling platform to help make decisions regarding wildfire mitigation practices, such as prioritizing the hardening of strategically important power lines in high-risk areas.

"Imagine where you have a workforce that comes into work and instead of rummaging through piles of data, both structured and unstructured, you could just come in and ask a copilot a question in terms of what equipment has the highest likelihood of failure in a particular region," Khursheed said.

Energy transition

AI can also be used to assist in maximizing the value of new renewable energy resources, according to the panelists.

"AI will help us to understand where it's best to deploy a battery or solar panels. It will also help you to know when it's best to charge or discharge a battery," Åsberg said. AI experts at CERAWeek by S&P Global in March also asserted that AI can be used as a tool to accelerate the pace of the energy transition, pointing to how some oil refiners have turned to the technology as a way to meet carbon reduction goals and track emissions.

However, utilities have also warned about a massive spike in demand resulting from the explosion of AI and datacenters. In January, the International Energy Agency said power demand from cryptocurrencies, AI and datacenters will double within the next three years.

"Because of the power of those large language models, we are also now able to build small language models: models that are smaller but still serve a purpose because the energy consumption and the energy need for those large models are still quite high," Asberg said.

In July, US Sen. Joe Manchin (I-W.Va.) and Sen. Lisa Murkowski (R-Alaska) introduced a bill that would require the US Department of Energy to study the potential power grid impacts of AI. Some companies within the AI sector have pushed back against claims that growth in AI computing is resulting in power drain, arguing that innovation by the industry will lead to energy efficiency improvements.

Sirosh said innovation in materials, semiconductors and energy storage could lead to "huge efficiency improvements" in AI.