A genetic algorithm to optimize the multi-group structure for the neutronic analyses of the ARC fusion reactor

In the framework of the modeling of fusion reactors with deterministic neutronic codes, the choice of an appropriate energy grid for the generation of the multigroup nuclear properties is essential. In this work, a Genetic Algorithm is employed to optimize the energy grid employed in the nemoFoam multiphysics code to reproduce the results provided by the Monte Carlo code Serpent in terms of neutron flux, neutron power deposition and Tritium Breeding Ratio for the Affordable, Robust and Compact (ARC) fusion reactor. Different runs of the Genetic Algorithm are performed, with the aim of optimizing not only the quantities of interest separately, but also trying to combine them thanks to the definition of appropriate fitness functions. The optimization is performed starting from a pre-defined 86 groups energy grid, over which the nuclear properties and the reference quantities are evaluated with Serpent. The results show that it is not straightforward to optimize at the same time the energy grid for different quantities and that, in general, coarse energy grids are able to provide good results in nemoFoam for what concerns the ARC reactor, allowing to alleviate the computational burden of the neutronic evaluation too.