In-core Optimization of Pressurised Water Reactor Reload Design via Multi-objective Tabu Search

Abstract

Periodically, nuclear Pressurized Water Reactors need to have a proportion of their fuel removed, new fuel added, and the remaining pattern reloaded in such a way as to yield the desired balance of operational considerations. This loading pattern then remains in the reactor until the next reloading event. The subtleties in the calculations of physical properties and the high degree of sensitivity to changes make this a highly complex combinatorial optimization problem. The methods that have historically been used to make the decisions about nuclear reactor loading pattern optimization are increasingly supplemented by computational methods. This work assessed one such method’s ability to optimize multiple objectives simultaneously – multi-objective Tabu Search. It was statistically analysed in comparison to other common leading methods – notably the Genetic Algorithm. It was tested on real reactor models using realistic data provided by a utility. The Tabu Search was first tuned via sensitivity studies to ensure a fair comparison in that both algorithms are near optimally configured. The focus of the work is light water reactors, both standard and small modular size, and it will not look at other reactor types. The objectives chosen reflect a range of the possible calculations. The principal aim is to establish whether the single- and then multi-objective Tabu Search can produce comparable, better, or worse optimization sets than its main competitor, the Genetic Algorithm, when applied to this loading pattern optimization problem. It was found that, although the Tabu Search outperformed the current industry standard algorithms for single-objective runs, the multi-objective results, although comparable, were more mixed. This work discovered that the Tabu Search for the in-core loading pattern optimization is still effective when single objective searches are not restricted, for example, by generalized perturbation theory. The set up, for both single objectives and multi-objective problems, is robust in terms of the choice of configuration. However, on multi-objective search spaces the inherent discontinuities in the search space mean that the confusion in which direction along the search space to traverse means that the population based methods still out perform the method.