Moment Matching: A New Optimization-Based Sampling Scheme for Uncertainty Quantification of Reactor-Physics Analysis

Because of the complexity of the nuclear reactor system, traditional statistical sampling methods, such as random sampling and Latin hypercube sampling, often lead to unstable uncertainty quantification results of the reactor physics analysis. In order to make the analysis results robust, traditional sampling methods require a large number of samples, which brings a huge computation cost. For this reason, this paper proposes a new sampling scheme based on the moment matching method to generate efficient samples for the uncertainty quantification of reactor physics calculations. A linear programming model is established to minimize the deviations of the first- and second-order moments. The generated samples can better reflect the statistical characteristics of the real distribution than classical sampling methods. A series of numerical experiments is carried out to demonstrate the superiority of the proposed moment matching sampling method, which can quickly provide more reliable uncertainty quantification results with a small sample size.