Boiling Water Reactor In-Core Fuel Management through Parallel Simulated Annealing in FORMOSA-B

Boiling Water Reactor;BWR;Optimization;Fuel Management;Nuclear;Simulated Annealing;Parallel Computing

Hays, Ross D ; Robert White, Committee Member,Paul J. Turinsky, Committee Chair,Edward Davis, Committee Member,Hays, Ross D ; Robert White ; Committee Member ; Paul J. Turinsky ; Committee Chair ; Edward Davis ; Committee Member

University:North Carolina State University

关键词: Boiling Water Reactor;    BWR;    Optimization;    Fuel Management;    Nuclear;    Simulated Annealing;    Parallel Computing;   

Others  :  https://repository.lib.ncsu.edu/bitstream/handle/1840.16/2746/etd.pdf?sequence=1&isAllowed=y

美国|英语

来源: null

【 摘 要 】

A commercial nuclear power plant with a boiling water reactor willutilize between 368 and 800+ individual fuel assemblies to generatesteam for 18 to 24 months between refueling outages.The compositionand reactivity of each fuel assembly will vary due to variations ininitial enrichment, burnable poison loading and irradiation conditionsin the core.These variations pose a challenge to the engineers whomust design subsequent reloads because only one quarter to one half ofthe fuel will be replaced at a time.One of the challenges is todetermine the optimum layout of the fuel within the core in order toget the highest value from the fuel without violating any safety oroperational limits.The FORMOSA-B program wasdeveloped to automatically find a family of optimum loading patterns by combining a robust, accurate 3-D coresimulator with a simulated annealing loading pattern search.Otherfeatures have been added to allow the program to rapidly computeshutdown margins and optimize control rod programmingthrough the application of heuristic rules.Onedrawback of the FORMOSA-B program is that long run-times,sometimes exceeding a week, arerequired to generate and evaluate the large numbers of solutionsrequired by the simulated annealing algorithm.The rising popularityand availability of parallel computing and computational clustersprovides a possible solution to the problem of long run-times.Tothis end, a parallel simulated annealing capability has been developedfor the FORMOSA-B program. The parallel simulated annealing driver utilizes standardized MessagePassing Interface routines to divide the individual Markov searchchains of the simulated annealing algorithm among a large number ofprocessors.By evaluating multiple loading patterns concurrently, runtimes are significantly reduced.In testing with a 368-assembly BWR/4model, parallel speedup factors exceeding 32 were observed with 48processors.Parallel efficiencies are calculated to be in the rangeof 68% to 95% when correcting for hardware variations and CRP updatefrequency. Further testing was performed to investigate theeffects on the annealing performance of the Control Rod Programmingupdate frequency, Markov chain length versus parallelization width andsolution downselect method.