Comparison of heuristic optimization techniques for the enrichment and gadolinia distribution in BWR fuel lattices and decision analysis

Abstract In the present study a comparison of the performance of five heuristic techniques for optimization of combinatorial problems is shown. The techniques are: Ant Colony System, Artificial Neural Networks, Genetic Algorithms, Greedy Search and a hybrid of Path Relinking and Scatter Search. They were applied to obtain an “optimal” enrichment and gadolinia distribution in a fuel lattice of a boiling water reactor. All techniques used the same objective function for qualifying the different distributions created during the optimization process as well as the same initial conditions and restrictions. The parameters included in the objective function are the k -infinite multiplication factor, the maximum local power peaking factor, the average enrichment and the average gadolinia concentration of the lattice. The CASMO-4 code was used to obtain the neutronic parameters. The criteria for qualifying the optimization techniques include also the evaluation of the best lattice with burnup and the number of evaluations of the objective function needed to obtain the best solution. In conclusion all techniques obtain similar results, but there are methods that found better solutions faster than others. A decision analysis tool based on the Position Vector of Minimum Regret was applied to aggregate the criteria in order to rank the solutions according to three functions: neutronic grade at 0 burnup, neutronic grade with burnup and global cost which aggregates the computing time in the decision. According to the results Greedy Search found the best lattice in terms of the neutronic grade at 0 burnup and also with burnup. However, Greedy Search is unpredictable because its results are diverse and in average is the worst method in terms of the global cost which aggregates also the computing time to the final evaluation. Genetic Algorithms and Path Relinking coupled to Scatter Search have the best results in terms of global cost.

[1]  David E. Goldberg,et al.  Genetic Algorithms in Search Optimization and Machine Learning , 1988 .

[2]  Alejandro Castillo,et al.  Nuclear Fuel Lattice Optimization Using Neural Networks and a Fuzzy Logic System , 2009 .

[3]  Victor J. Rayward-Smith,et al.  Modern Heuristic Search Methods , 1996 .

[4]  Enrique Mérida Casermeiro,et al.  An Efficient Multivalued Hopfield Network for the Traveling Salesman Problem , 2001, Neural Processing Letters.

[5]  Juan José Ortiz,et al.  Fuel lattice design in a boiling water reactor using an ant-colony-based system , 2011 .

[6]  Chaung Lin,et al.  Automatic fuel lattice design in a boiling water reactor using a particle swarm optimization algorithm and local search , 2012 .

[7]  David A. Pelta,et al.  GreeNN: A hybrid method for the coupled optimization of the axial and radial design of BWR fuel assemblies , 2010 .

[8]  F. Glover.  Scatter search and path relinking , 1999 .

[9]  Raúl Perusquía,et al.  Fuel lattice design with Path Relinking in BWR’s , 2011 .

[10]  Guy L. Curry,et al.  Optimization of MOX enrichment distributions in typical LWR assemblies using a simplex method-based algorithm , 2002 .

[11]  Goldberg,et al.  Genetic algorithms , 1993, Robust Control Systems with Genetic Algorithms.

[12]  G. Ivan Maldonado,et al.  An Application of Linear Superposition to Estimating Lattice-Physics Parameters , 2001 .

[13]  Juan-Luis François,et al.  Optimization of BWR fuel lattice enrichment and gadolinia distribution using genetic algorithms and knowledge , 2007 .

[14]  Juan-Luis François.  A practical optimization procedure for radial BWR fuel lattice design using tabu search with a multiobjective function , 2003 .

[15]  Luca Maria Gambardella,et al.  Ant colony system: a cooperative learning approach to the traveling salesman problem , 1997, IEEE Trans. Evol. Comput..