Two U.K. public sector entities - the U.K. Atomic Energy Authority (UKAEA) and the Science and Technology Facilities Council's Hartree Centre (STF) – are collaborating with US-headquartered technology firm IBM to design future experimental fusion power plants.
The partnership intends to unite fusion scientists and Artificial Intelligence (AI) experts from the three organizations to achieve transformative breakthroughs in applying AI to fusion power plant designs and experimental facility operations. The collaboration will combine the Hartree Centre and IBM's expertise in AI and high-performance computing, with UKAEA’s data and modelling capabilities. They will create a ‘frontier’ or ‘foundation model’ capable of learning and underpinning the fundamental dynamics of experimental fusion data. The UKAEA is the U.K.'s national organization responsible for researching and delivering fusion energy. It will provide program requirements, domain expertise and selected data from its JET and MAST-U machines.
IBM said, “Our approach to-date for designing these complex machines has been one of ‘test-based design’ - ie an iterative approach of ‘learning by doing’. Unfortunately, measured against the demanding timeline for decarbonizing and transitioning economies into the Net-Zero era, test-based design for fusion has now become too slow and too expensive.”
IBM added, “It is essential therefore that the fusion sector adopts the latest digital technologies to accelerate and de-risk the delivery of commercial fusion power - for operations and for plant design. In short, we must move the dial which represents how we design complex strongly coupled fusion systems away from test-based design and towards the digital world of simulation and ‘data centric’ engineering.”
The new collaboration is expected to develop foundation models that can learn the underpinning dynamics of the UKAEA's fusion plasma/plant experimental data. This will allow the generation of new information and new capabilities that will feed into various applications, including training downstream models for simulation and/or prediction. Utilizing these techniques, the models will ‘learn’ from past experiments. Ideally, these models will evolve ‘incrementally’ whereby they will ingest live experimental data.
Rob Akers is the UKAEA Director of Computing Programs. He said, “I am delighted that we are joining forces with IBM and STFC's Hartree Centre to work on our ambitious program aiming to deliver commercial fusion in the 2040s by exploiting the transformative power of Artificial Intelligence. IBM's expertise in complex systems engineering and supercomputing and the Hartree Centre's expertise in democratizing high-performance computing and AI into the engineering sector, combined with UKAEA's leading research and development in fusion energy will be a powerful force for progress in this hugely important field.”
Vassil Alexandrov is the Chief Science Officer at national computing center, STFC Hartree Centre. He said, “I am really very pleased that, thanks to our well-established collaborations with both IBM and UKAEA, we can now come together to address a key grand challenge and advance state-of-the-art in modelling and simulation of fusion powerplants, thereby supporting the UK's ambition to become a global leader in clean energy innovation.”
Juan Bernabe-Moreno is the Director of IBM Research Europe, U.K. and Ireland. He commented, “I am especially excited to see our team exploring together with the UKAEA and the Hartree Centre experts how we can use generative AI technologies to approach one of the most challenging problems of our time. It is certainly a testament to the kind of research we are driving in the UK for the greater good.”