Use of Meta-Heuristics for Design of Fuel Loading Pattern in Light Water Reactors Comprising Some Radial and Axial Heterogeneities

摘要

The third generation nuclear core should be attractive from the nuclear fuel cycle management aspect. The main goal is to achieve a good conversion of fertile isotopes into fissile isotopes while respecting safety constraints. Preliminary studies have shown the interest of the core loaded with fertile and heterogeneous fissile assemblies. The purpose of this work is to find out an optimized loading pattern and axial repartition of fissile and fertile elements in regards to the following criteria: power peak and void coefficient minimization. Due to the huge number of possible combinations, we have replaced the global optimization by two simple forms: first optimization of the fertile / fissile repartitions on an assembly in 1D representation, and secondly optimization of the core's loading pattern in 3D representation. The evaluations of the criteria have been done using the neutron physics transport solver MINOS integrated in the new lattice-core APOLLO3 code. The research algorithms used in this study are of an evolutionary algorithm, a "Max Min ant system colony algorithm" and a particle swarm adapted to our multi criteria approach. These algorithms are distributed using the heterogeneous island's method. A second level of parallelism has been introduced inside each island. Therefore we could increase the number of evaluations while respecting the CPU time limitation of the batch queues available. There are several solutions that stand out from the simulations. Some of them are completely original and have proved to be relevant a posteriori. This method could then be used as a decision support tool for the exploitation reactors and also for the design of new reactors.