American Nuclear Society

Home / Store / Journals / Electronic Articles / Nuclear Technology / Volume 141 / Number 2 / Pages 122-141

Multicycle Optimization of Advanced Gas-Cooled Reactor Loading Patterns Using Genetic Algorithms

A. Kemal Ziver, Jonathan N. Carter, Christopher C. Pain, Cassiano R. E. de Oliveira, Antony J. H. Goddard, Richard S. Overton

Nuclear Technology / Volume 141 / Number 2 / Pages 122-141

February 2003


Member Price:$27.00
Member Savings:$3.00

A genetic algorithm (GA)-based optimizer (GAOPT) has been developed for in-core fuel management of advanced gas-cooled reactors (AGRs) at HINKLEY B and HARTLEPOOL, which employ on-load and off-load refueling, respectively. The optimizer has been linked to the reactor analysis code PANTHER for the automated evaluation of loading patterns in a two-dimensional geometry, which is collapsed from the three-dimensional reactor model. GAOPT uses a directed stochastic (Monte Carlo) algorithm to generate initial population members, within predetermined constraints, for use in GAs, which apply the standard genetic operators: selection by tournament, crossover, and mutation. The GAOPT is able to generate and optimize loading patterns for successive reactor cycles (multicycle) within acceptable CPU times even on single-processor systems. The algorithm allows radial shuffling of fuel assemblies in a multicycle refueling optimization, which is constructed to aid long-term core management planning decisions. This paper presents the application of the GA-based optimization to two AGR stations, which apply different in-core management operational rules. Results obtained from the testing of GAOPT are discussed.

Questions or comments about the site? Contact the ANS Webmaster.