[en] Highlights: • A genetic algorithm for the automatic cross-section energy grouping is proposed. • The algorithm effectively finds appropriate energy structures for each test system. • Results show a strong relation between meshing and system spectrum. • Energy structure effect on the results is larger for heterogeneous systems. • Usage of inappropriate energy structures may lead to large results discrepancies. - Abstract: The generation of multigroup neutron cross-section libraries is a key issue of the multigroup transport calculations in reactor physics. The correct choice of the boundaries of the energy groups, in particular, is decisive for obtaining reliable results. Knowledge of the reactor physics, general and specific of the studied reactor, along with long and refined analyses are required for finding out a reasonable energy structure, which is specific for the considered reactor and might be unsuitable for other systems. The genetic algorithm presented in this work aims to choose the most appropriate energy structure for the considered system to collapse a fine multigroup library into a few-groups one, usable for transient transport calculations. The user is free to choose the number of energy groups of the final library, which is in direct relation with the precision required and the time available for the simulation. The methodology is coupled with SIMMER-III code and applied to 3 reactor systems: ESNII+ ASTRID, ESFR and MSFR. The results show that the algorithm can find representative energy structures, providing accurate results on the multiplication factor. The results of each test are analyzed, showing how different compositions, geometries and neutron spectra guide the algorithm choices, so demonstrating the effectiveness of the method.