OptiFun: Optimising Fusion with Functional Programming

Research Project, 2022 – 2023

The aim of this project is to combine numeric and symbolic methods to accelerate first-principles simulations and enable optimisation of fusion confinement designs.

As a complement to renewable energy technologies, we need to develop new large-scale carbon-free energy sources, such as fusion. The aim of this project is to combine numeric and symbolic methods to accelerate first-principles simulations and enable optimisation of fusion confinement designs. By combining the interdisciplinary expertise of the Departments of Physics and Computer Science \& Engineering (CSE), we will build a reliable and versatile tool-set for performance optimisation of strongly coupled physical systems with a large parameter space, and apply it to explore reliable fusion designs.

Models of the real world are always approximations, but there is a vast range from very crude (but fast) to very accurate (but slow) simulation models. When searching for optimal designs, it is often necessary to rely heavily on the faster models, because running the slower models could take months. But this can easily lead to designs which are "optimal" only in the crude approximation, but not in the real world. The core idea of this project is to build a tool-set that automatically combines multiple runs of the faster models with occasional runs of the slower models to calibrate and check the correctness. If this works out, we could get the best of both worlds.