GPU加速的氟盐冷却球床堆堆芯瞬态分析方法研究

发布时间：2018-06-23 16:06

本文选题：GPU加速 + 氟盐冷却球床堆　；参考：《中国科学院研究生院(上海应用物理研究所)》2017年博士论文

【摘要】：氟盐冷却球床堆(Pebble-bed fluoride-salt cooled high temperature reactor,PB-FHR)是结合了多种反应堆技术优势的新型第四代反应堆,采用液态氟化盐冷却剂、内嵌包裹燃料颗粒的球形燃料元件,具有良好的经济和安全特性。PB-FHR堆芯瞬态分析是反应堆设计和安全分析的重要环节,然而目前缺少针对PB-FHR开发的堆芯瞬态分析程序。堆芯瞬态分析是中子时空动力学和热工水力学的时空多维耦合计算问题,计算耗时非常长,GPU作为一种新型并行计算工具,能有效提升数值模拟程序的计算速度。本文针对PB-FHR堆芯的中子物理及热工水力特性,为PB-FHR建立合理的堆芯瞬态分析模型,结合GPU加速技术研发PB-FHR三维堆芯瞬态分析程序,并对PB-FHR进行三维堆芯瞬态分析。本文基于时空多群扩散理论建立堆芯三维中子动力学模型,在三维圆柱坐标的非均匀结构化网格系统下,采用细网有限体积法及隐式时间积分方法对模型进行离散和求解。本文基于多孔介质模型,在宏观尺度上建立堆芯热工水力模型,采用多孔介质湍流模型模拟冷却剂的热弥散效应,基于多孔介质非热平衡模型模拟冷却剂和燃料球的换热现象,建立燃料球的双重非均匀结构传热模型、热工水力辅助封闭模型及堆芯球床与侧反射层的耦合传热模型,采用同位网格的SIMPLEC算法(Semi-Implicit Method for Pressure Linked Equations-Consistent)求解堆芯热工水力模型。本文基于伪材料法和Lagrange插值法建立了连续温度点的宏观群常数计算方法,建立了瞬态物理热工半隐式耦合计算方法。本文研究了大型七对角稀疏线性方程组的GPU并行迭代求解算法,在GPU上实现了2种迭代求解算法(共轭梯度算法(CG)和稳定双共轭梯度法(BICGSTAB))及4种方程预处理算法(Neumann多项式预处理算法(POLYN)、不完全Cholesky分解预处理算法(IC0)、不完全LU分解预处理算法(ILU0)和几何代数多重网格预处理算法(GAMG),设计并研发了GPU加速的三维堆芯瞬态分析程序。本文对GPU堆芯瞬态分析程序的中子物理和热工水力求解器展开了校核工作,采用圆柱堆芯中子动力学基准题对中子动力学求解器的稳态和瞬态计算功能进行校核,采用与商用计算流体力学软件FLUENT进行结果对比的方式对热工水力求解器的稳态和瞬态计算功能进行校核,校核结果证明了所采用的数理模型、数值算法的合理性和正确性。本文对中子物理和热工水力求解器的GPU加速性能进行了详细的分析,证明了GPU加速的有效性,并发掘出最优的求解器组合。对中子物理求解器的分析结果表明,GPU并行POLYN-CG求解器具有最高的加速比率(21.65倍),GPU并行GAMG-CG求解器具有较快的收敛速度,但较低的加速比率(13.8倍)和较大的单次迭代计算量;当网格数量小于2万时,GPU求解器加速效果不明显,当网格数量适中时(2万到3百万),GPU并行POLYN-CG算法的计算耗时最少,当网格数量达到3百万以上时,GPU并行GAMG-CG求解器的计算耗时最少。对热工水力求解器的分析结果表明,使用GAMG预处理算法求解压力修正方程及冷却剂温度方程、POLYN预处理算法求解其他物理方程可使整体求解速度最快,加速比率最高达8.39倍。本文参考中科院上海应用物理研究所的PB-FHR实验堆设计方案,建立了PB-FHR堆芯模型,并对该模型进行了物理热工耦合稳态模拟和瞬态模拟,初步分析了PB-FHR堆芯的稳态及瞬态运行特性,同时证明了所研发的程序计算结果的合理性。堆芯物理热工耦合稳态模拟的结果表明,控制棒插入深度对堆芯中子通量密度和功率密度的分布形状具有显著影响;多孔介质孔隙率和阻力对堆芯压降及流速具有显著影响;氟盐冷却剂、燃料球表面、燃料球石墨中心及TRISO颗粒之间具有明显的温度梯度,燃料球表面温度受氟盐温度的影响较大,燃料球石墨、TRISO颗粒温度受堆芯功率密度的影响较大。本文进行了单根控制棒移动、堆芯入口温度变化、堆芯入口流量变化的瞬态工况模拟,分析了上述工况下堆芯功率、温度在时间和空间上的变化特性及物理热工的耦合效应。结果表明,单根控制棒移动会产生很大的局部扰动,使堆芯功率、温度的幅度和分布形状均产生较大变化;温度反馈效应能有效控制堆芯功率的变化,但有明显的滞后性;缓发中子会影响堆芯功率的变化速率;入口氟盐温度变化对堆芯氟盐和燃料球表面的温度影响显著,并引入温度反应性,使堆芯功率和温度发生变化,但不会影响堆芯功率及温度分布的形状;当侧反射层内表面附近的氟盐和燃料球表面温度变化时,侧反射层内表面温度也受到显著影响;入口氟盐流量变化引起堆芯氟盐、燃料球表面温度的变化,并引入温度反应性。
[Abstract]:The fluorine salt cooling ball bed reactor (Pebble-bed fluoride-salt cooled high temperature reactor, PB-FHR) is a new type of fourth generation reactor combined with a variety of reactor technical advantages, using liquid fluorinated salt coolant and spherical fuel element embedded with fuel particles, which has good economic and safety characteristics, and.PB-FHR core transient analysis is inverse. There is an important link in the design and safety analysis of the reactor. However, there is a lack of the core transient analysis program developed for PB-FHR. The transient analysis of the core is a time-space multi-dimensional coupling calculation problem of neutron space-time dynamics and thermal hydraulics. The time-consuming calculation is very long. As a new type of calculation tool, GPU can effectively improve the numerical simulation program. In this paper, based on the neutron physics and the thermal hydraulic characteristics of the PB-FHR core, this paper establishes a reasonable core transient analysis model for PB-FHR, and combines the GPU acceleration technology to develop the PB-FHR three-dimensional core transient analysis program, and carries out the three-dimensional core transient analysis of PB-FHR. In this paper, the model is discrete and solved by the finite volume method of fine mesh and the implicit time integration method. Based on the porous medium model, the thermal hydraulic model of the core is established on the macro scale, and the thermal dispersion effect of the coolant is simulated by the multi pore mesoporous turbulence model. In the non thermal equilibrium model of porous media, the heat transfer phenomenon of the coolant and fuel ball is simulated, the dual non-uniform heat transfer model of the fuel ball is established, the thermal hydraulic auxiliary sealing model and the coupled heat transfer model of the core ball bed and the side reflection layer are coupled with the SIMPLEC algorithm of the same bit grid (Semi-Implicit Method for Pressure Linked Equations-Con). Sistent) to solve the core thermal hydraulic model. Based on the pseudo material method and the Lagrange interpolation method, the method of calculating the macroscopic group constants of the continuous temperature points is established, and the transient physical thermal semi implicit coupling calculation method is established. This paper studies the GPU parallel iterative solution algorithm of the large seven diagonal sparse linear equations, and implements 2 kinds of algorithms on GPU. The iterative solution algorithm (conjugate gradient algorithm (CG) and stable dual conjugate gradient method (BICGSTAB)) and 4 equation preprocessing algorithms (Neumann polynomial pre processing algorithm (POLYN), incomplete Cholesky decomposition preprocessing algorithm (IC0), incomplete LU decomposition preprocessing algorithm (ILU0) and geometric algebraic multigrid preprocessing algorithm (GAMG)) are designed and developed for GPU. In this paper, the neutron physics and thermal hydraulic solver of the GPU core transient analysis program is checked. The steady-state and transient calculation functions of the neutron dynamics solver are checked with the neutron dynamic benchmark of the column core, and the software FLUENT is used with the commercial computational fluid dynamics software. The calculation of the steady-state and transient functions of the thermal hydraulic solver is checked in the way of the comparison of the results. The results of the calculation prove the rationality and correctness of the mathematical model and the numerical algorithm used. This paper analyses the GPU acceleration performance of the neutron physics and thermal hydraulic solver, and proves the validity of the acceleration of the GPU. The analysis results of the neutron solver show that the GPU parallel POLYN-CG solver has the highest rate of acceleration (21.65 times), and the GPU parallel GAMG-CG solver has a faster convergence rate, but a lower acceleration ratio (13.8 times) and a larger single iteration calculation; when the number of grids is less than 20 thousand, GPU is solved. The acceleration effect is not obvious. When the number of grids is moderate (20 thousand to 3 million), the GPU parallel POLYN-CG algorithm takes the least time. When the number of grids is over 3 million, the calculation of GPU parallel GAMG-CG solver takes the least time. The analysis results of the thermal hydraulic solver show that the GAMG preprocessing algorithm is used to solve the pressure correction equation and cold. However, the solution of other physical equations by the POLYN pretreatment algorithm can make the overall solution speed the fastest and the acceleration ratio up to 8.39 times. In this paper, the PB-FHR core model is established by the PB-FHR experimental reactor design scheme of the Shanghai Institute of Applied Physics of the Academy of Sciences, and the model of physical thermal coupling steady state and transient simulation is carried out. The steady-state and transient operating characteristics of the PB-FHR core are preliminarily analyzed, and the results are proved to be reasonable. The results of the reactor core physical thermal coupling steady-state simulation show that the insertion depth of the control rod has a significant influence on the neutron flux density and the distribution of the power density of the core, and the porosity and resistance of the porous media. The pressure drop and flow velocity have a significant influence on the core pressure drop and flow velocity, and there is a clear temperature gradient between the fluorine coolant, the fuel ball surface, the fuel ball graphite center and the TRISO particles. The surface temperature of the fuel ball is greatly influenced by the temperature of the fluorine salt. The fuel ball graphite and the TRISO particle temperature are greatly influenced by the power density of the core. The change of core power and temperature in time and space and the coupling effect of physical heat are analyzed. The results show that the single control rod movement will produce a large local disturbance, which makes the core power, the temperature range and the distribution shape all produce. The temperature feedback effect can effectively control the change of the core power, but it has obvious hysteresis; the retarded neutrons will affect the change rate of the core power; the temperature change of the inlet fluorine salt has a significant influence on the temperature of the core fluorine salt and the surface of the fuel ball, and the temperature reactivity is introduced to make the core power and temperature change, but not the shadow. The shape of the core power and temperature distribution; when the surface temperature of the fluorine salt and the fuel ball near the inner surface of the side reflecting layer changes, the inner surface temperature of the side reflector is also significantly affected; the change of the flow rate of the inlet fluorine salt leads to the change of the core fluorine, the surface temperature of the fuel ball and the temperature reactivity.
【学位授予单位】：中国科学院研究生院(上海应用物理研究所)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TL351.1

【相似文献】