A hybrid system for optimizing enrichment and gadolinia distributions in BWR fuel lattices

A new approach for optimizing the enrichment and gadolinia distribution in a typical BWR fuel lattice is presented. It takes advantage of potentialities of two different methodologies: the Ant-Colony-System (ACS) metaheuristic technique and S15 a method based on knowledge. The S15 method was programmed in such a way that can reduce the objective function systematically in a few iterations. In this way, a human expert was essentially substituted by the system. S15 takes advantage of the ACS's ability for exploring the searching space. Reciprocally, ACS can use the directed search generated by S15. The system was tested using typical fuel lattices from both transition and equilibrium operation cycles in a BWR. The lattice physics code CASMO-4 was used to generate the parameters of the objective function. As a result of the hybridization process of the aforementioned methodologies, the obtained results overcame those obtained using different heuristic techniques applied to the same problem. All imposed goals or requirements were improved or fulfilled: on average, the obtained local power peaking factor is 7.57% lower than the reference one, the obtained value for the initial average U-235 enrichment is, on average, 1.12% lower than the reference one, the precision in obtaining the k-infinity's requirement was improved by about 33%, and the gadolinia content was fulfilled in all the cases. Finally, the needed total wall time for obtaining a suitable solution was substantially reduced using the proposed system.