Quantifying performance of passive systems in an integrated small modular reactor under uncertainties using multilevel flow modelling and stochastic collocation method

Abstract

Many newly designed advanced nuclear reactors deploy several passive systems to increase the inherent safety. However, uncertainties often exist in these new designs because of lacking sufficient operation experience and experimental data, which leads to the possible deviation on the performance of systems from their prospective values and increases the risk of the plant. Thus, methods should be developed to quantify the influence of uncertainties on the performance of the passive systems properly to improve our understanding of these new designs and provide support for decision-making.Traditionally, the system performance analysis under uncertainties is performed by Monte-Carlo based methods (MCS), which is extremely time-consuming. Thus, to circumvent the limitation of traditional Monte-Carlo based method, a novel hybrid method that combines multilevel flow modelling and stochastic collocation (MFM-SC) is proposed in this paper. The MFM-SC method utilizes the MFM to model the functions of collaborative working passive systems and inference the relevant uncertain parameters according to causal reasoning rules. Then, the stochastic collocation method which combines the strength of Monte Carlo methods and stochastic Galerkin methods is used to create an orthogonal polynomial based surrogate model for the thermal hydraulic model. With the efficiency of the surrogate model in computation, the quantification of the performance of passive systems under uncertainties which requires thousands of repetitive computations can be easily conducted.An example test case is also presented in this paper to analyze the performance of core makeup tanks and the steam drum in an integrated small modular reactor during a station blackout accident under multiple uncertainties using MFM-SC method. The statistical results of the deviation range of the systems’ performance under uncertainties are given according to thousands of experiments using a surrogate model, and the response surfaces are also calculated based on the MFM-SC method to reveal the relation between the uncertain inputs and the systems output separately. The result shows the significant improvement in the computational efficiency of the suggested MFM-SC method, and it can provide essential information for decision-making.