The development of a novel adaptive genetic algorithm for the optimization of fuel cycle length

Abstract

The report presents the novel genetic algorithm (GA) developed to improve fuel cycle performance and in-core fuel management process. The primary purpose was to develop GA dedicated to solving core loading pattern (LP) problem with predefined core operation constraints. Particular focus was put on maximizing the length of the fuel cycle, but presented solutions allow to limit or control the magnitude of excess reactivity, population and type of fuel assemblies, fissile material mass and inventory of burnable absorbers. GA was implemented in a new computational framework coupled with PARCS3.2 core simulator. The Pressurized Water Reactor (PWR) based on the MIT BEAVRS Benchmark was used as the demonstration case, and developed tools were applied to improve the first fuel cycle. Several test simulations were performed, including population size, a number of generations, mutation level and other aspects of GA. New variance control method, was proposed and tested. It was compared with the standard roulette and rank method, showing improved convergence. Four test cases were studied with different constraints put on the fuel assemblies population, mass of fissile materials, burnable absorbers and limiting excess reactivity. The obtained results show that the new algorithm works properly and can allow designing efficient fuel loading pattern with increased fuel cycle length.