Multi-Configuration Optimization of the Multi-group Energy Structure for Lead Fast Reactor simulations

The identification of an appropriate energy group structure is a key step in accurately collapsing group constants for full-core transient and multiphysics analyses, as it directly impacts both the accuracy and computational efficiency of multi-group calculations. Genetic Algorithms (GAs) have proven to be effective tools for solving such high-dimensional, non-linear optimization problems. By emulating the Darwinian principle of natural selection, they aim to identify a nearly-optimal, coarser energy grid through physics-informed objective (fitness) functions. In this study, physics-based fitness functions associated with relevant full-core parameters were implemented in the optimization framework, extending previous works in literature. The GA was applied to simultaneously optimize the energy group structure for three different core configurations of the NEA Lead-cooled Fast Reactor (LFR) benchmark, using the neutronic module of the FRENETIC code coupled with Serpent 2 Monte Carlo simulations. The results indicate that the constraints imposed by the simultaneous optimization reduce the solution space, leading to similar optimized structures across configurations, with variations observed only for a few energy boundaries.