Calculations for Fusion Reactors are now faster. AI is to thank for this.

 Fusion reactors are well suited to meet our potential power needs in a secure and sustainable way. Researchers may use numerical models to gain insight into the behaviour of the fusion plasma as well as the efficacy of reactor design and operation. To model the large number of plasma interactions, however, a number of specialised models are needed, which are too slow to provide data on reactor design and operation.

Aaron Ho of Eindhoven University of Technology's Science and Technology of Nuclear Fusion group in the department of Applied Physics investigated the use of machine learning methods to speed up numerical simulations.

The ultimate aim of fusion reactor research is to achieve a net power gain in a cost-effective manner. Big, complicated devices have been built to achieve this purpose, but as these devices become more complex, it becomes increasingly necessary to take a predict-first approach to their operation. This improves operating efficiency while also protecting the system from serious harm.

To simulate such a system, models must be able to capture all of the related phenomena in a fusion device, be accurate enough to make reliable design decisions, and be fast enough to find workable solutions quickly.

Aaron Ho used a model focused on neural networks to create a model to fulfil these requirements for his PhD study. This method helps a model to maintain both speed and accuracy at the expense of data collection. QuaLiKiz, a reduced-order turbulence model that predicts plasma transport quantities induced by microturbulence, was used to test the numerical method. In tokamak plasma systems, this phenomenon is the dominant transport mechanism. Unfortunately, in current tokamak plasma modelling, its measurement is also the limiting speed factor.

The model was then used in an optimization exercise using the coupled device on a plasma ramp-up scenario as a proof-of-principle to assess its efficacy outside of the training results. This research deepened our understanding of the physics underlying the experimental findings and demonstrated the value of quick, accurate, and detailed plasma models.

Finally, Ho proposes that the model be expanded to include additional applications such as controller or experimental design. He also suggests that the methodology be extended to other physics models, as the turbulent transport predictions are no longer the limiting factor. This would increase the integrated model's applicability in iterative applications and allow for the validation efforts needed to move it closer to being a genuinely predictive model.

*PC-GOOGLE