Advancing grid intelligence: Argonne hosts workshop on AI foundation models

As society’s reliance on electricity deepens, artificial intelligence (AI) is reshaping how we manage power grids and optimize energy sources. A recent workshop hosted by the U.S. Department of Energy’s Argonne National Laboratory brought together leading experts from national labs, universities, government agencies and industry to explore the transformative potential of AI foundation models for electric grids.

The three-day Foundational Models for Electric Grid workshop — organized by Argonne researchers Kibaek Kim, Emil M. Constantinescu and Adrian Maldonado — marked the third event in an evolving series. In partnership with industry leaders IBM, Hydro-Québec and the National Rural Electric Cooperative Association (NRECA), the event strengthened the collaboration and innovation needed for advancing smarter, more adaptive grid systems.

“We are laying the foundation for a future where AI-driven models will be an integral part of how we manage and optimize our power grids.” — Emil M. Constantinescu

Kim, a computational mathematician, noted the growing interest.

“We’ve seen attendance grow from about 25 participants at our first workshop to well over 100 at this latest session,” he said. ​“This surge in interest reflects the field’s rapid advancement and the urgent need for AI solutions in electric grid management.”

From forecasting to outage prevention: Workshop prioritizes collaboration, real-world application

The workshop emphasized the importance of multidisciplinary collaboration. Through technical sessions, panel discussions, live demonstrations and structured networking, participants exchanged best practices and forward-looking insights.

Industry experts showcased AI-driven advancements ranging from sophisticated forecasting models to automated distribution management systems, demonstrating how AI can dramatically enhance grid performance and resilience.

Argonne’s Valerie Taylor, director of the Mathematics and Computer Science division, and Henry Huang, director of the Energy Systems and Infrastructure Assessment division, delivered keynote addresses that reinforced Argonne’s leadership at the intersection of AI and energy research.

Huang highlighted the urgency of modernizing grids across all scales, underscoring how advanced analytics and AI are critical for building a smarter, more resilient and efficient energy infrastructure.

A major focus of the workshop was the application of foundation models, which are AI systems pre-trained on vast, diverse datasets and fine-tuned to address specific grid challenges. Maldonado, an assistant energy systems scientist, explained their versatility.

“Our foundation model is an AI engine trained on extensive datasets covering various power grid functions,” he said. ​“It’s designed to handle everything from forecasting to operations, making it a comprehensive solution for modern grid management.”

Constantinescu, a senior computational mathematician, further detailed the benefits of AI models.

“These models can detect subtle signals that traditional methods often miss, helping us predict and prevent outages before they cause significant disruptions,” he said.

Building smarter, safer grids with privacy conscious AI

Looking forward, participants identified priorities such as scaling AI deployment, shaping regulatory frameworks and enhancing cross-sector collaboration.

A key innovation discussed was privacy-preserving federated learning (PPFL), a method that enables foundation models to be trained on energy sector data while safeguarding sensitive information, such as smart meter data connected to consumers.

Kim emphasized the significance of PPFL especially regarding distributed energy resources (DER), which are small-scale units of local power generation or storage that are located close to where electricity is used, like homes, businesses or industrial sites. Instead of relying entirely on large, centralized power plants and long transmission lines, DERs allow for energy to be produced, stored and sometimes even consumed right at or near the point of use.

“As we introduce more distributed energy resources to the grid — such as natural gas generators and geothermal — these AI systems will help us manage increasingly complex operations with greater precision,” he said. ​“The models we’re developing today will evolve alongside the grid itself.”

Constantinescu stressed the workshop’s practical focus. ​“Our goal is not just theoretical. We are actively working on integrating foundation models into operational workflows, ensuring they can be used effectively in real-world power systems,” he said.

Argonne remains at the forefront of advancing AI-driven energy resilience. Insights from this workshop will drive future research initiatives and strengthen industry collaborations, ensuring that tomorrow’s power systems are secure, efficient and adaptive.

“There are two major challenges we are addressing — technological limitations in grid modeling, and broader resilience issues facing modern power systems,” concluded Constantinescu. ​“This is only the beginning. We are laying the foundation for a future where AI-driven models will be an integral part of how we manage and optimize our power grids.”