Designing energy discretisations for sensitivity studies in lead-cooled fast reactors using a genetic algorithm

n nuclear engineering, sensitivity studies are essential for understanding the impact of nuclear data on simulation uncertainties. Energy-dependent sensitivities are typically scored over suitable energy discretisations. The chosen energy grid affects the accuracy of the scored sensitivity in comparison with the continuous-energy reference. In this work, we explore the use of genetic algorithms as a search optimisation technique to define tailored energy grids for heavy liquid metal-cooled fast systems, such as ALFRED and MYRRHA. We developed a custom routine to achieve energy discretisations that minimise the root mean squared error between reference sensitivity profiles and evaluated sensitivity profiles for the effective multiplication factor 𝑘𝑘eff of 238U and 239 Pu. These improvements were compared with those obtained using ECCO-33, a widely employed energy discretisation for fast systems.