University of Virginia computer science professor Ferdinando Fioretto has been selected to lead one of the first research projects launched through the U.S. Department of Energy’s ambitious Genesis Mission, a national effort designed to combine artificial intelligence, high-performance computing and scientific expertise in order to accelerate breakthroughs on some of the most difficult challenges facing the United States. Fioretto’s project, titled “From Static to Adaptive Grids: Autonomous Topology Control at Scale,” focuses on a problem that is becoming increasingly urgent as electricity systems grow more complex: how to make power grids respond intelligently, rapidly and safely to changing conditions.
Modern electric grids were largely designed around predictable patterns of generation and consumption. Power plants supplied electricity through networks whose structures changed relatively infrequently, while grid operators relied on established procedures, forecasts and human decision-making to maintain balance. That model is being tested by the rapid expansion of renewable energy, the electrification of transportation, extreme weather and increasingly volatile patterns of demand. Solar and wind generation can fluctuate with weather conditions, while electric vehicles, data centers and industrial facilities are creating new and sometimes highly concentrated loads. Fioretto’s team will investigate whether artificial intelligence can help grids adapt their configurations in real time rather than forcing operators to work within largely fixed network arrangements.
The project centers on power grid topology control, a technical process that involves changing the operational structure of a network by opening or closing selected switches and circuit breakers. These actions can redirect electricity flows, reduce congestion, isolate damaged equipment or improve the use of available generation. In principle, topology control allows operators to reshape the grid in response to evolving conditions. In practice, however, the number of possible configurations is enormous, and every change must satisfy strict physical and operational constraints. Voltage levels, line capacities, frequency stability and the balance between electricity production and consumption all need to remain within safe limits. A decision that appears beneficial in one part of the network could create instability or overload elsewhere.
Artificial intelligence could provide a new way to manage this complexity. Fioretto’s group will develop AI-driven methods capable of analyzing large volumes of grid data, identifying promising network configurations and recommending or executing actions under carefully defined safety requirements. The systems may draw on machine learning, optimization and knowledge-based reasoning to evaluate possible decisions much faster than conventional approaches. Rather than treating the grid as a static collection of wires and substations, the research aims to represent it as an adaptive system whose structure can change as conditions evolve. Such a system could continuously assess demand, generation, equipment status and weather-related risks before selecting a configuration that improves resilience or efficiency.
The challenge is not simply to make an algorithm produce fast answers. Electric grids are safety-critical infrastructure, and an AI system must operate within physical laws, regulatory requirements and clearly defined operational boundaries. A model that performs well on historical data could still fail when confronted with an unusual combination of events, such as a heat wave, a sudden loss of generation and a transmission bottleneck. For that reason, the project is expected to emphasize scalable optimization, constraint-aware learning and methods that can provide reliable decisions across a wide range of operating conditions. High-performance computing will be essential because large networks can involve thousands of components and an immense number of possible switching combinations.
The Genesis Mission is intended to provide the computational and scientific environment needed for projects of this scale. Researchers participating in the initiative will have access to the Genesis Mission Platform, which brings together advanced AI models, AI-agent frameworks and high-performance computing resources from the Department of Energy’s national laboratory system and industry partners. These resources could allow Fioretto’s team to train and test models using realistic grid simulations, explore scenarios that would be difficult to study on conventional computers and connect algorithmic advances with expertise from power systems engineers and energy researchers. The broader goal is to create AI-enabled scientific workflows in which models do more than analyze information: they help researchers formulate, test and refine solutions to complex real-world problems.
The Department of Energy selected 278 projects from what it described as the largest response to a funding opportunity in the agency’s history. The portfolio includes teams from national laboratories, universities, companies and nonprofit organizations in all 50 states, with research spanning energy systems, advanced computing, materials, nuclear energy and engineering. Fioretto’s selection places the University of Virginia among a national network of researchers working to demonstrate how AI can be integrated into scientific and engineering applications. The initiative reflects a growing shift in the role assigned to artificial intelligence. Instead of focusing exclusively on language, images or automated office tasks, researchers are increasingly applying AI to physical infrastructure, where decisions must be explainable, verifiable and compatible with complex scientific constraints.
For the electric grid, the potential impact is substantial. More adaptive topology control could help operators route power around damaged or congested equipment, make better use of renewable generation and reduce the need for costly infrastructure upgrades in some situations. It could also improve the grid’s ability to recover from disruptions caused by storms, wildfires, cyber incidents or equipment failures. These benefits will depend on whether AI systems can demonstrate consistent performance, communicate their reasoning to human operators and remain dependable when conditions fall outside the data used during training. Fioretto’s project therefore addresses both an engineering problem and a central question in trustworthy AI: how can autonomous systems make high-stakes decisions while remaining subject to human oversight and rigorous physical safeguards?
The University of Virginia’s Department of Computer Science chair, Sandhya Dwarkadas, said Fioretto’s selection reflects the strength of UVA Engineering’s leadership in artificial intelligence and its commitment to nationally important challenges. His research illustrates how advances in computer science can be connected to infrastructure on which daily life depends. As electricity demand rises and energy systems become more decentralized, the ability to adapt may become as important as the ability to generate power. By combining machine learning with optimization, high-performance computing and power systems expertise, the project seeks to move the grid toward a future in which its topology is no longer treated as fixed, but as an intelligent and carefully controlled resource. If successful, the work could offer a blueprint for using AI to make one of the world’s largest engineered systems more efficient, resilient and responsive.
Subject of Research: AI-driven autonomous topology control for adaptive electric power grids
Article Title: AI Project Aims to Transform Static Power Grids Into Adaptive, Autonomous Networks
Keywords: artificial intelligence, machine learning, power grids, topology control, autonomous systems, high-performance computing, energy resilience, renewable energy, smart grids, Genesis Mission
