Enhancement of Spatial Treatment in OCEON-P code with Regard to Fidelity and Loading Pattern Type Determination

Show full item record

Title: Enhancement of Spatial Treatment in OCEON-P code with Regard to Fidelity and Loading Pattern Type Determination
Author: Mathur, Arvind Kumar
Advisors: Paul J Turinsky, Committee Chair
Pierre Gremaud, Committee Member
Dmitriy Y Anistratov, Committee Member
Abstract: This thesis summarizes the work done to improve the fidelity of the FLAC core simulator utilized in the OCEON-P code. The OCEON-P (O-P) code was developed to perform multi-cycle optimization for PWRs. Given a planning horizon, the O-P code determines the family of cycling schemes such that the levelized fuel cycle cost is minimized within constraints. The determination of a cycling scheme involves both the fresh fuel selection and burnt fuel re-insertion strategy. The fresh fuel attributes that are determined by O-P are the number, enrichment(s), and burnable poison content, i.e. type and loading. The code consists of three segments: the optimization engine, the core simulator, and the economics engine. The optimization engine is based on a Biased Integer Monte Carlo stochastic optimization method. Two core simulators are available, the linear reactivity model (LRM) and the FLAC (a FLARE derivative) nodal model. A carrying charge model is employed for the economics engine. In the original spatial treatment of the FLAC model, it was assumed that an assembly was coupled only to its four face-adjacent assemblies. Further investigation with the NESTLE code indicated that significant coupling also occurred to the assemblies that were diagonally adjacent to it and ignoring this coupling could lead to erroneous results. As part of this study, the mathematical model and the code of the FLAC simulator were modified to take into account this coupling of an assembly with its diagonally adjacent neighbors. This led to a significant improvement in the performance of the code. Certain other features were also added to improve the capabilities of the code. The EOC reactivity ranking maps were generalized to allow them to be a function of the number of fresh fuel assemblies, in addition to loading pattern (LP) type. LP type was made an optimization parameter. The addition of LP type optimization capability allows O-P to determine the associated optimum batch power shares, which can be provided as constraints to the FORMOSA-P code enabling capture of multi-cycle optimization effects in this single-cycle LP optimization code. The use of the rather simplistic LRM core simulator during optimization was a serious drawback of the O-P code. The computational capability of computers at the time the code was written had required this limitation. The vast improvements in the capabilities of modern computers suggested the logical way out. The problem was resolved by adding the option to use the higher fidelity FLAC core simulator during the optimization. The optimization starts using the LRM simulator in order to generate cycle-dependent region size biasing parameters which steer the search towards feasible and economically attractive cycling schemes. Once these parameters are well estimated, the FLAC core simulator takes over for the rest of the optimization. The modified FLAC core simulator was extensively tested against the NESTLE code, which solves the few-group neutron diffusion equation using the nodal expansion method (NEM). Westinghouse PWR core models for three successive reload cycles were set up in both codes. Initially each assembly was treated as a separate node in FLAC and the beginning of cycle (BOC) relative assembly power distribution and the cycle burnups over one cycle of operation compared on an individual assembly basis for the three consecutive cycles. The EOC k-effective values were also compared. The NESTLE and O-P codes were run in various modes to capture the effects of fission products, thermal hydraulic feedback and of using a 2-D vis-a-vis a 3-D core model. Subsequently, the input to the FLAC model was modified by lumping all assemblies with the same enrichment, burnable poison loading and burnup history into a single spatial node called a FLAC unique batch. The assemblies in a FLAC unique batch were separated further into unique subbatches depending on whether they were located in the core interior or located with the same number of surfaces on the reflector. This spatial nodalization is the default treatment used by O-P. The tests were then repeated to bring out the impact of using FLAC unique sub-batch spatial nodalization on cycle batch incremental burnups and EOC k-effective in the OCEON-P code. Finally a set of tests was run to confirm the fidelity of the FLAC predictions over multiple cycles. The EOC burnups from one cycle, as predicted by OCEON-P code, were carried forward to the next cycle and the cycle batch power shares compared against the NESTLE code predictions. This was done using the default and single assembly FLAC unique sub-batch nodalization. After a review of the results, it was concluded that the errors EOC k-effective predictions of the FLAC core simulator are unacceptably high when the default FLAC unique sub-batch spatial nodalization is used and hence FLAC should be utilized with a finer spatial nodalization, perhaps in the single assembly nodalization mode which produces sufficiently accurate results. Testing was also carried out to verify that the other features added to the code were working satisfactorily.
Date: 2003-08-01
Degree: MS
Discipline: Nuclear Engineering
URI: http://www.lib.ncsu.edu/resolver/1840.16/871

Files in this item

Files Size Format View
etd.pdf 362.7Kb PDF View/Open

This item appears in the following Collection(s)

Show full item record