Thea Energy collaborates with AI companies to develop stellarator digital twin

Thea Energy has announced it is working with Nvidia and Synopsys to develop a digital twin of its stellarator fusion power plant concept, called Helios.

The team, which also includes Argonne National Laboratory and Princeton Plasma Physics Laboratory, will "analyze and scale vast datasets, rapidly evolve Thea Energy's plant designs, and stress-test system operation in a workflow that outpaces traditional tools," according to the company.

Last December, Thea announced that it had completed Helios's preconceptual design, which includes software-controlled magnet coils that are adaptable to real-world conditions. David Gates, cofounder and chief technology officer of Thea, said this collaboration expands the company's AI applications to include multifaceted device modeling.

"With the Helios digital twin, we can shorten development cycles and essentially run the system before we even put a shovel in the ground. We are committed to expanding this ecosystem further with partners that share our vision for building the most reliable, scalable, and maintainable fusion power plants," said Gates.

Nvidia will be integrating Thea Energy's models, codes, and real-world data into a digital twin platform using Nvidia Omniverse libraries, which Thea Energy said will allow it to analyze power plant performance in real time. Synopsys will be integrating data into a Ansys simulation-driven, AI-accelerated framework to develop Thea's breeding blanket system.

Steve Pytel, senior vice president of product management at Synopsys, said, "We are developing the Helios digital twin to be more than a model. It will be an analytical tool with the potential to further derisk Thea Energy's path to fusion on the grid. Our high-fidelity simulation solutions reveal complex interactions with photorealistic detail, enabling engineers to predict behavior and evaluate system performance."

According to Thea Energy, PPPL will supply key knowledge spanning plasma modeling and high-fidelity codes required to simulate complex plasma behaviors under power-plant-relevant conditions. Argonne will contribute data to be integrated into the Helios digital twin and will provide expertise in neutronics analysis and blanket design.

"We look to bridge the gap between large-scale physics simulations and practical engineering iteration and for the first time, complex datasets required to train AI surrogate models on real-world blanket design challenges can feed into the Helios digital twin and provide rapid feedback. This process fundamentally alters the workflow for major power plant components, accelerating the deployment of fusion energy," said John Tramm, a computational scientist at Argonne.

Background: Thea Energy (formerly Princeton Stellarators) is one of eight companies that was chosen by the Department of Energy in its inaugural round for the Milestone-Based Fusion Development Program in 2023. Thea is developing a large-scale demonstration facility, Eos, as a step toward its fusion power plant concept, Helios. The company has said Eos is scheduled to be operational by 2030, and Helios is scheduled to operate in the 2030s.

According to the DOE, Thea's stellarator concept was made more commercially viable with the development of a magnet design that does not require the "highly precise spaghetti-like coils" previous stellarator designs relied on.

"We've reinvented the stellarator with no more wiggly coils," said Gates. "We can replace twisting coils with flat coils and vary their currents. So, we've taken the complexity out of the coils and put it into the control system."

Thea calls its magnet technology planar shaping coils, and in May the company announced it had successfully completed its first full-size, full-current, and full-field test of the technology.

The company has received six DOE Innovation Network for Fusion Energy (INFUSE) awards.

Genesis mission: Thea emphasized that the collaboration aligns with the DOE's Genesis Mission, which has been driving a national effort to accelerate the use of AI in scientific research.

According to the company, "Thea Energy is using AI to bridge simulation and real-world system operational data, closing critical gaps to build the Helios digital twin faster and with orders of magnitude less capital compared to traditional models."