数字化反应堆高保真中子学程序SHARK研发

SHARK程序是由中国核动力研究设计院新近研发的基于全堆芯确定论非均匀输运理论体系的数字化反应堆软件。该软件从多群数据库的截面与共振数据出发，采用改进子群方法刻画有效共振截面的复杂非均匀效应，采用二维/一维或准三维特征线方法开展堆芯层面非均匀输运计算。目前该程序的定态微观问题计算能力已建立完毕。数值结果显示，SHARK程序对于商用压水堆相关基准问题具有良好的计算精度和效率。

Development of Digital Reactor High-fidelity Neutronics Code SHARK

The concept of digital nuclear reactor, which is featured with high-resolution modeling and high fidelity multi physics simulation, has been proposed and developed in recent years with the dramatic development of high performance computers. It has special and important meanings for the design improvement and mechanistic understanding of nuclear energy system. In the digital reactor system, the high fidelity core physics neutronics simulator is a key component. In this paper, a newly developed high fidelity neutronics code SHARK (Simulation based High fidelity Advanced Reactor physics Kit) was introduced. This is a one step heterogeneous transport code developed by Nuclear Power Institute of China (NPIC). It adopts constructive solid geometry (CSG) method to model the problem geometry, which enhances its geometry flexibility including square and hexagonal lattices with different kinds of fuel pellets. In methodlogies, the code is currently characterized by a 45 group neutron library, subgroup self shielding method, and both 2D/1D method of characteristics (MOC) and quasi-3D MOC as transport solvers. For the multi group library, the selected energy group structure was proved to be accurate enough and effective for light water reactor (LWR) applications. The isotopes in the library covered fuels, moderator and coolant, poison absorbers, structural materials and their depletion chain daughters. Reaction types including fission, capture, (n, 2n)/(n, 3n) and decays were considered. For the resonance self shielding, an improved subgroup method based on equivalence cross section tables was utilized. This method kept the advantages of conventional methods in the sense of geometry flexibility and distribution effects, and gained better efficiencies on computational costs. Resonance scattering and interferences were also properly treated. For the heterogeneous transport solvers, 2D/1D MOC was applied with a discrete ordinate (SN) finite difference axial solver. To improve the stability of neutron transport calculations, quasi-3D MOC was also developed, which did not contain a transverse leakage term and gained stability. For acceleration and parallelization, optimally diffusive coarse mesh finite difference method (odCMFD) was used to improve the convergence stability, and an MPI plus OpenMP mixed parallel scheme was implemented. The overall calculation frame was established by the languages of Python and C++. The code used object oriented design concept and had good maintainability and user friendliness. The steady state verification results were described in the paper including some VERA progression benchmarks and macro BEAVRS problem. The preliminary benchmarking results demonstrate the methodology and code developments in terms of reactivities, radial and axial power distributions and control worth. In the future, performance and functionality will be improved continuously.