Abstract
A many-reactor power plant with a shared used fuel pool presents significant challenges and a novel opportunity for fuel management optimization. We develop a Python package and interface it with optimization software and core modeling software to automate exploration of the design space for the multireactor multicycle problem. A genetic algorithm is used to search the core reload design space for a two-reactor system, with and without used fuel sharing. With equal computational effort, we find that the fuel-sharing strategy slightly lowers cost.
Acknowledgments
The authors wish to thank the engineers in the Nuclear Fuels Analysis Group at NuScale for their mentorship. Thanks also to Paul J. Turinsky for introducing B. L. to the topic of nuclear fuel cycle and providing feedback on the early work of this project. Finally, thanks are due an anonymous reviewer for improving the quality of this work by asking important questions about the repeatability of the optimization parameters.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Notes
a After the initial work on our project, NuScale Power announced a new module powered at 250 MW(thermal).
b The cycle length (lower limit) constraint is similar with signs changed.
c Specific keywords and metaparameters are the following: population_size = 100, max_evaluations = 4000, seed = 123 (and incremented), max_iterations = 1000, fitness_type = linear_rank, replacement_type: chc = 10, mutation_type = offset_uniform, mutation_rate = 1.0, crossover_type = uniform, and crossover_rate = 0.8
d Kropaczek has shown this may improve overall performance by about 1% (CitationRef. 14).