Application of adjoint nased node optimization method to nuclear thermal-hydraulic system analysis code

Abstract

In the field of nuclear engineering, 1-D Thermal-Hydraulic system analysis code is used for analysis of a large scale system such as a nuclear power plant. Recently, the Best Estimate plus Uncertainty (BEPU) approach is widely used instead of a conservative approach. Among the uncertainties, user effects occur because there are many degrees of freedom such as choosing physical model options and nodalization. Previously, errors and uncertainties related to spatial discretization in the process of nodalization have been considered to be less dominant than other user effects due to their difficulty in quantification. However, in recent years, there are some results that they are comparable to other uncertainties such as uncertainties due to physical models. Previous studies have been conducted by evaluating and optimizing user effects by performing iterative calculations on many occasions using genetic algorithms or parallel computation techniques. Since these methods consume a large amount of resources for computation, this thesis suggests a method of consuming less resources. For this purpose, the user effect associated with node is evaluated and the adjoint method that analyzes the sensitivity for many parameters efficiently is applied to nuclear thermal-hydraulic system analysis code for node optimization.