Quantification and Minimization the Uncertainty Propagation of Monte Carlo Cross Section Generation for HTR Applications

Abstract

The High Temperature Reactor (HTR) is one of the six Generation IV advanced reactor designs. HTRs have become more and more attractive for power generation applications among both commercial companies and research institutes because of its intrinsically safety and modular design features. For several reasons, the modeling and simulation of HTRs can be much more challenging than traditional Light Water Reactors which rely primarily on deterministic methods to analyze the coupled neutronics and temperature fluid fields. In the research here, hybrid methods combining stochastic Monte Carlo methods and deterministic methods were applied to HTR simulations. In particular, Monte Carlo methods are an attractive alternative to overcome the complexities in the modeling and analysis of the considerable heterogeneity of small modular reactors such as the HTR which use TRISO fuel. In the research here, the Monte Carlo code Serpent was used for few-group neutron cross section generation and the deterministic codes AGREE and SAM were used for neutronics and thermo-fluids transient simulation using the cross sections generated by Serpent.

The stochastic nature of Monte Carlo processes has the potential to introduce additional statistical uncertainties. Quantifying this uncertainty was a principal focus of the research performed here. In order to quantify the uncertainties introduced from Monte Carlo cross section generation, two different methods were investigated, including an analytic perturbation-based method and a stochastic probabilistic method using the industry standard code Dakota. The ultimate objective was to quantify the uncertainty for a model of an advanced HTR, the Kairos Power FHR (KP-FHR), which is a novel advanced reactor technology that leverages TRISO fuel in pebble form combined with a low-pressure fluoride salt coolant. However, the research methods developed here were first demonstrated using an experimental HTR reactor, the HTR-10, which became an international IAEA benchmark and is currently used world-wide to validate computer codes used in the safety analysis of small modular advanced HTRs such as the FHR. Specifically, the four benchmark problems of the HTR-10 were modeled. Good agreement of the deterministic and probabilistic was demonstrated which provided confidence in then applying the stochastic methods to the FHR for both steady-state and coupled neutron and temperature fluid field transient analysis. Parametric studies were also performed to investigate the factors that may affect the uncertainty quantification. The results for the FHR equilibrium core problem showed that the uncertainty of the k-eff and local power introduced from the Monte Carlo generated cross section was small if sufficient number of histories were used to generate the neutron cross sections in Serpent. However, the uncertainty of some important safety parameters, e.g., the region-wise reactivity coefficients, was very large if insufficient numbers of histories were used. A ramp reactivity insertion simulated using the industry standard SAM code with coefficients generated through Serpent/AGREE. Despite the uncertainty of some neutronics parameters being larger, the uncertainty of the important thermo-fluids parameters, such as the maximum coolant and fuel temperature, was still small. For this phase of the research the Dakota code was also coupled with the SAM code to propagate the uncertainties of Serpent/AGREE.

The results of this research demonstrated the use of innovative uncertainty quantification methods and results which provide guidance to reactor analysts on the numbers of histories necessary to minimize the contribution of the uncertainty introduced from Monte Carlo into the prediction of HTR reactor safety performance.