GreeNN: A hybrid method for the coupled optimization of the axial and radial design of BWR fuel assemblies

摘要

Radial and axial optimization of the fuel assembly in a boiling water reactor are usually solved as independent problems, despite they are highly related. In this work we propose GreeNN, a hybrid system composed by a simple greedy search technique and a neural network that allows approaching the solution of both problems in a coupled way. Firstly, GreeNN performs the radial optimization of the fuel assembly (minimizing the Local Power Peaking Factor according to a 2D simulation) and then, the obtained fuel lattice is added to a fuel lattices inventory. This inventory is used to solve the axial opti-mization of the fuel assembly where a 3D core simulator is used to make a Haling calculation at the end of the cycle and to estimate the generated energy. The method proceeds iteratively, with the aim of decreasing the uranium enrichment of the designed fuel lattices in the radial stage while keeping the energy requirements. GreeNN system was applied to design the fuel lattices for an equilibrium cycle of 18 months. The fuel assembly's performance proposed by GreeNN system was better than the reference case, without jeopardizing the reactor safety.