Research Of GPU-accelerated Core Transient Analysis Method For Pebble-bed Fluoride Salt Cooled High Temperature Reactor

Pebble-bed fluoride-salt cooled high temperature reactor(PB-FHR)is a kind of novel nuclear energy systems,combining techniques of several advanced reactors such as coated particle(TRISO)sphere fuel and molten-salt coolant to achieve better safety and economy performance.Core transient analysis of reactor core is significant for both design and safety issues,but no specialized core transient analysis code for PB-FHR has been developed so far.Core transient analysis is a time-space neutron-dynamics/thermal-hydraulics coupling simulation issue,consuming much time.GPU is a massive parallel-computing device that can efficiently accelerate numerical computing speed.The aim of this research is to build physical/mathematic models of core transient analysis based on the neutron-dynamics and thermal-hydraulics characters of PB-FHR and develop a GPU-accelerated core transient analysis code that can analyze three-dimensional core transient behaviors for PB-FHR.In this paper,a neutron-dynamics model based on time-dependent multi-group neutron diffusion theory was built.Structural mesh system of three dimensional cylindrical coordinate was adopted.Fine mesh finite volume method and implicit time-integral method were adopted for model discretization and solution.A thermal hydraulic model based on porous media theory was built.The thermal dispersion mechanism in the pebble bed was considered by adopting macroscopic turbulence models of porous media.A local thermal non-equilibrium model was built to simulate heat transfer between fuels and coolants.A double heterogeneous heat transfer model for fuel element was built to simulate coated particle temperature.Auxiliary thermal-hydraulic models and a pebble bed/side reflector coupling heat transfer model was built.The SIMPLEC(Semi-Implicit Method for Pressure Linked Equations-Consistent)algorithm was adopted to solve thermal hydraulics model.A method based on pseudo-material method and lagrange interpolation was proposed to calculate the macroscopic cross sections in different temperature.GPU-parallelized iterative algorithms for solving large seven-diagonal linear equations were studied.The conjugate gradient algorithm(CG)and bi-conjugate gradient stabilized(BICGSTAB)algorithm were implemented on GPU.Four efficient equation preconditioning algorithms,including Neumann polynomials(POLYN),incomplete Cholesky factorization(IC0),incomplete LU factorization(ILU0)and geometric-algebraic multigrid method(GAMG),were also implemented on GPU.A GPU-accelerated core transient analysis code were designed and developed for PB-FHR.The core transient analysis code was verified by comparing results with a neutron dynamics benchmark problem and commercial computational fluid dynamics code FLUENT.The good agreement of results indicated the physical/mathematic models and numerical methods were correct and rational.The detail acceleration performance of GPU solvers for neutron-dynamics and thermal hydraulics models was analyzed to prove the efficiency of GPU acceleration and find the fastest solving scheme.The analysis results for neutron dynamics model indicated the GPU solver ‘POLYN-CG' had the highest parallelization and can provide the largest speedup ratio of 21.65 times in the four GPU solvers,GPU solver ‘GAMG-CG' had faster convergence speed but lower speed up ratio(13.8).When grid quantity is smaller than 20000,GPU solvers are not efficient.When grid quantity was between about 20000 and 3 million,the GPU solver ‘POLYN-CG' had faster computing speed.When grid quantity was larger than about 3 million,the GPU solver ‘GAMG-CG' had faster computing speed.The analysis results for thermal hydraulics models model indicated the solving scheme of solving the pressure correction equation and coolant temperature equation with GPU GAMG solver and solving other equations with GPU POLYN solver has the highest computing speed with a speed up ratio of 8.39 times.A PB-FHR geometry model was built,referring to the experimental PB-FHR designed by Shanghai Institute of Applied Physics(SINAP).Static and transient cases was simulated to analyze PB-FHR core's coupling characters of neutron-dynamics and thermal-hydraulics and prove the calculation results of developed code in this paper were rational.The results of static simulation indicated control rods had significant influence on neutron flux and power distribution's shape.The porosity ratio and porous media resistance had significant influence on pressure and velocity.Significant temperature gradient existed between fluoride salt coolant,sphere fuel surface,sphere fuel center and TRISO particle.Fluoride salt had significant influence on sphere fuel surface temperature.Power density had significant influence on sphere fuel center and TRISO particle temperature.The transient cases of single control rod moving,inlet temperature changing and inlet mass flow changing were simulated.The results indicated single control rod moving caused significant local disturbance of power density distribution and temperature.The thermal feedback could effectively control the changing speed of power but had time lag.The delayed neutron had influence on power change.Inlet temperature changing caused significant change of fluoride salt and sphere fuel surface temperature,inducing reactivity to the core.The inner interface temperature of side reflector changed significantly when fluoride salt and sphere fuel surface temperature changed.The inlet mass flow changing caused change of fluoride salt and sphere fuel surface temperature and,inducing reactivity to the core.