Comparison and analysis of different selection strategies of genetic algorithms for fuel reloading optimization of Thorium-based HTGRs

The nuclear fuel cycle cost can be effectively reduced through fuel reloading optimization. Genetic algorithm (GA) is a classic optimization algorithm that is widely applied in fuel reloading optimization. In the GA, selection is a key operator. However, few studies have compared and analyzed different selection strategies. In this study, 1/6 core of thorium-based block-type high temperature gas-cooled reactor was considered as an example, and ten different selection strategies were compared and analyzed. Five of these strategies were the roulette wheel and proportionate selection, tournament selection, uniform sorting, exponential sorting, and deterministic selection, whereas the other five were the aforementioned selection strategies combined with the truncation selection strategy. These ten different selection strategies were evaluated for single-objective and multi-objective problems. In single-objective optimization problems, the effective neutron multiplication factor was selected as the only optimization objective, whereas in multi-objective optimization problems, the effective neutron multiplication factor and power peak factor were considered as optimization objectives. The results indicated that exponential sorting was the best selection strategy for single-objective optimization problems, whereas hybrid truncation exponential sorting was the best selection strategy for multi-objective optimization problems.