A comparative study on multiobjective metaheuristics for solving constrained in-core fuel management optimisation problems

• Published in: Computers & operations research 2016-11, Vol.75, p.174-190
• Main Authors: Schlünz, E.B., Bokov, P.M., van Vuuren, J.H.
• Format: Article
• Language: English
• Subjects: Algorithms; Comparative studies; Constraint handling; Constraints; Genetic algorithms; Handling; Heuristic; Heuristic methods; In-core fuel management optimisation; Mathematical models; Metaheuristics; Multiobjective optimisation; Nuclear reactor; Nuclear reactors; Optimization; Penalty function; Problem solving; Search algorithms; Statistical analysis
• ISSN: 0305-0548; 1873-765X; 0305-0548
• DOI: 10.1016/j.cor.2016.06.001

In this paper, the topic of constrained multiobjective in-core fuel management optimisation (MICFMO) using metaheuristics is considered. Several modern and state-of-the-art metaheuristics from different classes, including evolutionary algorithms, local search algorithms, swarm intelligence algorithms, a probabilistic model-based algorithm and a harmony search algorithm, are compared in order to determine which approach is the most suitable in the context of constrained MICFMO. A test suite of 16 optimisation problem instances, based on the SAFARI-1 nuclear research reactor, has been established for the comparative study. The suite is partitioned into three classes, each consisting of problem instances having a different number of objectives, but subject to the same stringent constraint set. The effectiveness of a multiplicative penalty function constraint handling technique is also compared with the constrained-domination technique from the literature. The different optimisation approaches are compared in a nonparametric statistical analysis. The analysis reveals that multiplicative penalty function constraint handling is a competitive alternative to constrained-domination, and seems to be particularly effective in the context of bi-objective optimisation problems. In terms of the metaheuristic solution comparison, it is found that the nondominated sorting genetic algorithm II (NSGA-II), the Pareto ant colony optimisation (P-ACO) algorithm and the multiobjective optimisation using cross-entropy method (MOOCEM) are generally the best-performing metaheuristics across all three problem classes, along with the multiobjective variable neighbourhood search (MOVNS) in the bi-objective problem class. Furthermore, the practical relevance of the metaheuristic results is demonstrated by comparing the solutions thus obtained to the current SAFARI-1 reload configuration design approach. •Considers the constrained multiobjective ICFMO problem for nuclear research reactor.•Comparative study using eight metaheuristics and two constraint handling techniques.•Nonparametric statistical analyses on results across several problem instances.•New constraint handling technique found to be competitive to the existing technique.•NSGA-II, P-ACO and MOOCEM generally found to be the best-performing metaheuristics.