Meet eGridGPT: How AI copilots could change grid operations

Posted: September 24, 2026

You flip on the lights, start the coffee maker, plug in your laptop, and electricity arrives instantly, exactly when you need it. But behind this simple mechanism is one of the most complex systems humans have created—one that’s under increasing stress.

According to the IEA, global power demand is expected to grow at an average of 3.6% annually through 2030, with electricity demand growing at least 2.5 times as fast as overall energy demand. This adds roughly 50% more demand per year than the previous decade^[1] An increase in distributed energy resources (DERs) and intermittent renewables like wind and solar add volatility and complexity to the grid, and more extreme weather, changing regulations, and data center expansion will add even more.

Energy security has become one of the most pressing issues of our time—one many experts believe technology, particularly AI, can best solve. Of course, AI is a significant draw on the power grid. According to an IEA report, a typical data center requires as much electricity as 100,000 households annually, with the largest currently in the works consuming 20 times as much.^[2]  This is sometimes called the AI-energy paradox, or the AI-energy nexus. Whatever we call it, the future of energy depends on how we approach it.

How the U.S. Department of Energy is using AI to improve grid reliability

In the United States, the Department of Energy’s Genesis Mission brings together the department’s 17 national laboratories, along with industry, academia, philanthropies, and other partners to “harness AI for breakthroughs in energy dominance, discovery science, and national security.”^[3] The mission has grown in scope to include a total of 15 different government agencies, including the Department of Health and Human Services, the Department of Transportation, NASA, and the Department of Defense—all of which are contributing a total of $5 billion for the mission, in addition to the use of their datasets and research facilities. The goal is ambitious: double the productivity and impact of American science and engineering within a decade.^[4]

“Scaling the grid to power the American economy” is one of the mission’s featured challenges. According to the mission’s website, the goal is to use AI to “reduce uncertainty, improve insights, and speed processes in grid planning, interconnection, operations, and security.” Specifically, this includes 20-100 times faster decision-making while improving electricity cost and reliability by at least 10%.^[5]

AVEVA is one of the mission’s industrial partners, as well as a longtime collaborator with the National Laboratory of the Rockies (NLR, formerly NREL), one of the 17 participating national laboratories. Scientists and operators at NLR have been using AVEVA™ PI System™ for years to collect and contextualize operational data. That foundation is now helping researchers explore what AI-enabled grid operations could look like, which has led to the development of eGridGPT (Electric Grid Generative Pretrained Transformer), a generative AI assistant designed to support grid operators.

Watch NRL’s presentation on how eGridGPT can be used to optimize grid operations

Why grid operations are more complex today than in the past

In order to make the decisions necessary to keep power flowing reliably to where it’s needed, a grid operator needs to know how much load there will be at any given moment. They need to schedule the generation to meet this load. And they need to choose the equipment and source of generation. Operators receive lots of data to help them orchestrate all this: operational data from generation, transmission, and distribution systems, as well as forecasting data such as outages, loads, interchange schedules, generation, demand, and weather forecasts^[6]

These decisions—and this data—become exponentially more complex with the addition of variable renewable energy, especially with the rise of prosumers—individuals who both consume and produce power—and DERs such as rooftop solar. Instead of the traditional one-way flow of yesterday’s grid, today’s grid operators must juggle a much more dynamic set of variables.

Read AVEVA’s reporting: Rewiring the power grid

What is eGridGPT?

Based on large language models (LLMs), eGridGPT, the platform developed by NRL, aims to provide decision support for grid operators and improve their situational awareness^[7] An operator can ask eGridGPT a question in natural language or pose a hypothetical scenario: “What is the impact of taking a substation offline during a heat wave?” or “Do you know how much wind generation is currently operating in ERCOT? Is the system operating normally?” And AI can return an answer that captures real-time conditions and operational context, verified with digital-twin simulations, delivered in a way an operator can understand and act on.

But eGridGPT is not just a chatbot. What sets it apart is the rigorous validation of information based on verified real-time data and physics-informed simulations.

Here’s how it works: When an operator submits a question or prompt, eGridGPT analyzes real-time grid information—including SCADA measurements, state estimates, and contingency analyses—to understand current conditions. It generates potential recommendations, tests them through digital-twin simulations, and refines the results for the operator. The result is insight that helps grid operators make faster, better decisions.

The importance of quality data in the AI-grid equation

None of this works without trustworthy, accessible data. The Genesis Mission represents a new model for scientific and technological innovation: connect the data, models, computing and expertise that already exist—and make them work together. AVEVA’s collaboration brings industrial data infrastructure into that equation, helping turn vast volumes of operational information into the contextualized intelligence AI needs to deliver meaningful results.

Read AVEVA’s reporting: What it takes to manage the power grid in a heatwave

Giving grid operators better intelligence

While still in the development stage, researchers are designing eGridGPT as a tool, not a replacement for a human operator. Instead of handing control of the grid to an algorithm, the aim is to give grid operators better intelligence—with the right guardrails around what AI can see, recommend, and do. Critical decisions will still be made by the expert, with transparent data tracing that enables operators to see the reasoning and information behind each suggestion.

As electricity demand continues to rise and energy systems become more distributed and dynamic, AI will undoubtedly play an increasingly important role. But its greatest impact won't come from replacing people—it will come from helping engineers, operators, and planners make better decisions using the information they already have.

^{[1] International Energy Agency. (2026, Feb. 6). “Global electricity demand is set to grow strongly to 2030, underscoring need for investments in grids and flexibility.” https://www.iea.org/news/global-electricity-demand-is-set-to-grow-strongly-to-2030-underscoring-need-for-investments-in-grids-and-flexibility}

^{[2] International Energy Agency. (2025). “Energy and AI: Executive summary.” https://www.iea.org/reports/energy-and-ai/executive-summary}

^{[3] U.S. Department of Energy. (2026). Genesis Mission. https://www.energy.gov/undersecretaryforscience/genesis-mission/genesis-mission}

^{[4] The Washington Post. (2026, July 23). AI Tech Brief: Washington’s AI policy knot. https://www.washingtonpost.com/wp-intelligence/ai-tech-brief/2026/07/23/ai-tech-brief-washingtons-ai-policy-knot/}

^{[5]U.S. Department of Energy. (2026). Scaling the grid to power the American economy. https://www.energy.gov/undersecretaryforscience/genesis-mission/scaling-grid-power-american-economy}

^{[6]Choi et al.. NREL. (May 2024). eGridGPT: Trustworthy AI in the Control Room. chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://docs.nlr.gov/docs/fy24osti/87740.pdf}

^{[7] Global Transmission. (2024, October 10). NREL’s eGridGPT Model: AI-driven solution for modern power grid challenges. https://globaltransmission.info/nercs-egridgpt-model-ai-driven-solution-for-modern-power-grid-challenges/}