Intel and Dell are to take part in the building of a “digital twin” of the UK’s Spherical Tokamak for Energy Production (STEP) prototype fusion power plant.
The UK Atomic Energy Authority (UKAEA) is collaborating with the tech giants and the Cambridge Open Zettascale Lab to produce a simulation of the plant, in an effort to meet the ambitious goal of delivering fusion energy to the UK’s energy network in the 2040s.
The UKAEA plans to make use of the lab’s supercomputer - based on Intel technologies - with 4th generation Intel Xeon processors running on Dell PowerEdge Servers.
The engineering behind such a fusion power plant requires enormous amounts of modelling and simulation, with a single plasma turbulence simulation potentially outputting hundreds of petabytes of data for analysis.
The business value of Dell PowerStore
High-performance storage that can improve performance and reduce operational costs
Thousands of graphics processing unit (GPU) nodes are required for large simulations with the open source Intel Distributed Asynchronous Object Storage (DAOS) object store taking care of the storage needed by high performance computing (HPC) applications.
Intel’s Data Center GPU Max Series chips are being looked at by the UKAEA as part of providing the large step increase in power required to drive the simulations.
During a briefing on Tuesday, Dr Paul Calleja, director of Computing Research at the University of Cambridge, remarked that traditional x86 systems “are not going to get us there” and that GPU technology would provide an order of magnitude increase in performance per watt.
“That’s really what this is about,” he said.
Calleja also called out Intel’s oneAPI, with a nod to SYCL, as “a really interesting programming environment”. The open nature of the standard means that the team has the option to use alternative silicon in the future.
“We can also run those codes on NVIDIA GPUs, and even AMD GPUs with minimal recode,” he explained, “And obviously, some of the coding is required, but it's not extreme.”
As for an operating system, Scientific OpenStack will be used in an effort to make the supercomputers accessible to a broad range of scientists and engineers.
Why do we need all this computing power?
In a word: Time. While it may seem a long away, the 2040s represent a very ambitious timescale considering the historic rate of progress of Fusion energy. As such, there simply isn’t enough time to adopt a fully test-based design process to work out what the power plant should look like.
Head of Advanced Computing at UKAEA, Dr Rob Akers, remarked that “Fusion has long been referred to as an exascale Grand Challenge”.
“We need to exploit the world's largest supercomputers to handle all of this physics and all of this complexity,” he added, “It's an absolutely staggering number of calculations.”
“We'll be able to get to the point where a digital version of STEP starts to emerge before the real plant itself,” he said. ”In the short term, we'll be able to use that digital version … to dramatically reduce the need for real world validation.
“So we'll be able to be a lot more intelligent around which components that we do have to prototype and test. And we'll be able to be a lot more intelligent about where we have to introduce design margin, or design overhead in order to make sure that the power plant has the efficacy to deliver power to the grid.”
In the longer term, there is the potential to turn the design into a full digital twin of the power plant for operational planning and anomaly detection.
What is STEP?
The Spherical Tokamak for Energy Production is a UKAEA programme to demonstrate the ability to generate electricity from fusion. A concept design is expected by 2024 and construction of the plant is expected to be complete “around” 2040.
According to the UKAEA “Fusion power creates nearly four million times more energy for every kilogram of fuel than burning coal, oil, or gas".
Akers described the project as “effectively a moonshot programme”, the goal of which was to prove fusion can be economically viable. It is, however, not expected to be a commercially operating plant at this stage.
Other fusion projects underway globally include ITER, which is expected to become operational at the end of 2025 with the goal of demonstrating a large fusion reactor rather than generating electricity.
