High-fidelity Neutronics And Thermal-hydraulics Coupling Method Based On Anderson Acceleration

The coupling calculation between neutronics and thermal-hydraulics can predict the nuclear reactor phenomena more realistically and further provide an important means and basis for the design and safety analysis of the reactor core.With the increasing demand for high resolution and high precision,as well as the development of high-performance computing platform and massively parallel technology,‘high fidelity’ has become the development trend of the coupling neutronics with thermal hydraulics.The high-fidelity neutronics and thermalhydraulics coupling system has the characteristics of high complexity,strong nonlinearity and massive calculation,which leads to the challenges of high difficulty and huge computational cost.However,due to the limitations of poor stability and slow convergence rate,the most widely used traditional Picard method can not meet the above challenges.Therefore,a more general coupling method is urgently needed to solve the coupling problem simply,stably,accurately and efficiently.The Anderson acceleration method is a classical algorithm for accelerating the convergence of fixed-point iteration.The remarkable advantages in stability,convergence rate and simplicity make Anderson acceleration have a bright prospect.To apply Anderson acceleration to the coupling calculation,it is necessary to consider the coupling characteristics for further study.For the potential numerical oscillation of Anderson acceleration,a simple but effective restart strategy is adopted to enhance the stability.Aiming at solving the problem of convergence caused by ‘over-solving’,the partial-convergence technique is introduced.Faced with the problem of the high computational cost attributed to high-fidelity neutronics,the Anderson acceleration is combined with the mature multilevel acceleration theory to greatly improve the computational efficiency.Based on the above method,an Anderson acceleration framework for high-fidelity neutronics and thermal-hydraulics coupling calculation is established,and then the steady-state and transient coupling modules in HNET program are developed.Subsequently,the VERA Problem 6 and NEACRP transient benchmark are simulated for the verification of method and modules.Based on the above coupling modules,the steady-state and transient performance of Anderson acceleration is tested with different types and sizes of typical pressurized water reactor problems,and the calculation accuracy and convergence performance are evaluated respectively.The steady-state results show that the Anderson acceleration method has good computational accuracy and is more stable and efficient than Picard iteration.Furthermore,its computing time is about 20% shorter than that of Picard with relaxation factor.The transient results show that the Anderson acceleration has a computing accuracy higher than that of OS method,but its transient convergence performance is sensitive to the change degree of core state.The convergence behavior of the Anderson acceleration is equivalent to or better than that of the Picard iteration only when the dynamic response of the core is strong.In summary,Anderson acceleration method has strong practicability for steady-state coupling calculation,and has a certain application potential for transient coupling problems.