Nuclear reactor physical numerical simulation method and system based on CPU / GPU heterogeneous platform

Abstract

The invention belongs to the technical field of nuclear reaction, and discloses a nuclear reactor physical numerical simulation method and system based on a CPU / GPU heterogeneous platform, and the method comprises the steps: assuming that each resonance nuclide adopts a subgroup fixed source equation needing to be solved for N times on average; according to subgroup transport calculation, circulation is carried out on each subgroup of each resonance nuclide in a calculation system, for M resonance nuclides, the number of calling times of a characteristic line method transport solver by a subgroup fixed source equation is MN, and characteristic line scanning calculation needs to be carried out for each calling; the MN subgroups are combined together to form a'multi-group 'structure; the multi-group structure is applied to subgroup fixed source transport calculation, feature line scanning is carried out in the whole multi-group structure at the same time, and fixed source flux of all subgroups is solved at the same time; a multi-group grouping calculation method is adopted, it is assumed that the number of energy groups adopted by final multi-group transport calculation is G, and sub-group fixed source calculation of a multi-group structure is conducted through existing multi-group transport variables.

Description

Technical Field

[0001] The present invention belongs to the technical field of nuclear reaction, and in particular relates to a method and system for nuclear reactor physics numerical simulation based on a CPU / GPU heterogeneous platform. Background Art

[0002] With the continuous improvement of nuclear energy technology, the international requirements for the safety and economy of nuclear reactors are becoming higher and higher. Accurate and efficient reactor simulation calculations are an important way to ensure the safety of reactor design and operation, and reactor physics analysis calculations require accurate analysis and calculation of neutron behavior. The most critical link is to obtain the flux density distribution of neutrons in the reactor. The neutron flux density needs to be obtained by solving the neutron transport equation, and the resonance calculation provides the effective resonance cross section for the transport equation, which is the basis for the calculation of the neutron transport equation. How to accurately and efficiently calculate the effective resonance cross section has become one of the research focuses in the field of reactor simulation calculations.

[0003] At present, the subgroup method is one of the most widely used resonance solution methods in the world. The main process of subgroup resonance calculation includes the calculation of subgroup parameters and the solution of subgroup transport equations. Finally, the subgroup flux and subgroup parameters are weighted and merged to obtain the effective resonance cross section. The subgroup transport equation is consistent with the multi-group transport equation in format, so it can be coupled with any transport solution method. It has great advantages in geometric processing and has been widely used. At present, the most mainstream high-fidelity transport solution method in the world is the characteristic line method, and the use of the characteristic line method to solve the subgroup transport equation has become a research hotspot. However, the solution of the subgroup transport equation requires multiple calls to the characteristic line method solver, which is computationally huge and time-consuming, and has a serious impact on the efficiency of numerical simulation of reactor physics. In recent years, with the increasing application of fine energy group calculations, the process of merging fine group cross sections of hundreds or thousands of groups into wide group cross sections of dozens or several groups has been completed. The grouping process requires neutron slowing calculations for a large number of energy groups, which also has a profound impact on the computational efficiency. In certain cases, the computational time of the grouping process may even exceed the solution time of the subgroup fixed source equation. Therefore, how to improve the efficiency of the subgroup resonance calculation method while ensuring accuracy is of great significance to the numerical calculation of nuclear reactor physics.

[0004] In recent years, the characteristic line method based on the graphics processing unit (GPU) has developed rapidly. The subgroup resonance calculation method based on the fine energy group structure requires a large number of transport solution modules to solve the subgroup fixed source equation and the fine group slowing equation. Therefore, it is very potential to accelerate the resonance cross section calculation process based on the CPU / GPU heterogeneous computing platform, but the efficient coupling method of this process is still unclear and lacks corresponding technical research.

[0005] The main technical problems existing in the industrial application of existing technologies can be summarized as follows:

[0006] 1) Low computational efficiency: The subgroup transport equation in the subgroup resonance method and the neutron slowing equation in the fine energy group both require multiple calls to the transport module of the characteristic line method. This group-by-group solution method will lead to repeated calls to the transport module during the calculation process, increasing the repeated operations of characteristic line scanning and communication, thus significantly reducing the computational efficiency.

[0007] 2) High computational complexity and long time consumption: Since the subgroup resonance calculation needs to process a large number of subgroup fixed source equations and fine group slowdown equations, and each subgroup or energy group needs to be transported and solved independently, the entire calculation process is huge and time-consuming. This has a serious impact on the computational efficiency in reactor physics numerical simulation, especially when high-fidelity simulation is required, and the calculation time becomes the main bottleneck.

[0008] 3) Lack of efficient coupling calculation methods: Although GPU acceleration technology provides potential for accelerated calculation of characteristic line methods, the efficient coupling method of existing subgroup resonance calculation methods on CPU / GPU heterogeneous computing platforms is still unclear. The lack of efficient coupling technology for the characteristic line method and the subgroup transport equation and slowing-down equation makes it impossible to fully utilize the computing power of GPUs, limiting the effect of calculation acceleration.

[0009] 4) Low efficiency of the group merging process: In the fine energy group calculation, the slowing down of the calculation process takes a long time to merge the fine group cross sections of hundreds or thousands of groups into the wide group cross sections of dozens or several groups. In some cases, the group merging calculation time may even exceed the solution time of the subgroup fixed source equation, further dragging down the overall calculation efficiency.

[0010] These technical problems pose a major challenge to improving the efficiency of nuclear reactor physics numerical simulation in actual industrial applications. Therefore, while maintaining calculation accuracy, improving the efficiency of subgroup resonance calculation methods has become an urgent problem to be solved in the industry. Summary of the invention

[0011] In view of the problems existing in the prior art, the present invention provides a method and system for nuclear reactor physics numerical simulation based on a CPU / GPU heterogeneous platform.

[0012] The present invention is implemented as follows: a nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform, by which the neutron flux distribution and reaction cross-section changes under different fuel rod arrangements, cladding materials and coolant conditions are quickly simulated;

[0013] - Allows designers to evaluate multiple fuel assembly designs in a short period of time;

[0014] -Optimize fuel assembly design to improve fuel utilization and reduce production costs;

[0015] - Further used to analyze the performance of fuel assemblies under different burnup conditions and provide data support for the formulation of fuel management strategies;

[0016] Among them, the method significantly improves the computing efficiency and simulation accuracy in the nuclear fuel assembly design and optimization process by utilizing the efficient computing power of the CPU / GPU heterogeneous platform.

[0017] Further, specifically including:

[0018] S1, in the resonance calculation process, it is assumed that on average each resonant nuclide needs to solve the subgroup fixed source equation N times; the subgroup transport calculation will loop each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation;

[0019] S2, multi-group transport calculation based on characteristic line method: MN sub-groups are merged together to form a "multi-group" structure; the multi-group structure is applied to the sub-group fixed source transport calculation, and the characteristic line scanning will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all sub-groups;

[0020] S3, adopts the method of multi-group grouping calculation, assuming that the final multi-group transport calculation uses the energy group number G, and uses the existing multi-group transport variables to perform sub-group fixed source calculations of the multi-group structure.

[0021] Furthermore, in S2, the multi-group calculation transfers the loop on the subgroups to the innermost part of the entire calculation loop, and a global modular characteristic line scan can perform flux calculation on all subgroups; the entire multi-group structure performs flux convergence at the same time, so the number of global iterations is the same as the subgroup with the largest number of group-by-group iterations; since the subgroup fixed source equation is a pure absorption process, the source terms between subgroups are the same and no source iteration is required, the number of iterations of different subgroups when solving group by group is basically the same, so the number of iterations of the subgroup fixed source equation of the multi-group structure is also basically the same as the group-by-group calculation.

[0022] Furthermore, in S3, for MN subgroups (MN>G), each G subgroups can be divided into a group. When there are less than G groups, they can be divided into one group according to the original structure. When the multi-group fixed source equations are solved directly without grouping, only one global characteristic line scan is required. The number of characteristic line scans during group solution is the same as the number of groups. Although the amount of calculation is increased compared to not grouping, it can still be greatly reduced compared to solving group by group.

[0023] Furthermore, the slowing-down equation needs to establish a recursive relationship for solving each group. The calculation of the source term of the current energy group first needs to obtain the scalar flux of the upstream energy group. When solving the slowing-down equation of the "multi-group" structure composed of multiple energy groups at the same time, since the upstream energy group flux is still an unknown quantity in the first iteration, the source term of the downstream energy group is first calculated using the initial value of the flux, and the scattering source term is continuously updated during each iteration of the characteristic line method until the flux of all energy groups converges.

[0024] Furthermore, referring to the grouping method of the multi-group fixed source equation, assuming that the number of fine resonant energy groups is X, the X resonant fine groups are divided into Y groups from top to bottom according to energy, and each group has Z groups; the multi-group grouping solution is recursively calculated in groups, and the multi-group slowing-down equation is solved in each group, and the scattering source term is updated in real time according to the current iteration step flux; after the slowing-down flux of all energy groups in the current group converges, it is provided to the downstream group to calculate the scattering source term scattered from the current group to the downstream group. The calculation is recursively performed group by group until the slowing-down flux of all resonant fine groups converges.

[0025] Furthermore, each thread corresponds to only one energy group in a characteristic line segment, forming a two-level parallel computing mode of characteristic line-single energy group; in the calculation of subgroup fixed source equation and slowing-down equation, the cross-section preparation and merging, source term initialization and updating and other calculations are executed on the CPU side, and the transport scanning calculation is performed on the GPU side.

[0026] Another object of the present invention is to provide a subgroup resonance computing system based on a CPU / GPU heterogeneous platform to which the nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform is applied, comprising:

[0027] The resonance calculation module is used to assume that, during the resonance calculation process, each resonant nuclide adopts the subgroup fixed source equation that needs to be solved N times on average; the subgroup transport calculation will loop each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation;

[0028] Multi-group transport calculation module, used for multi-group transport calculation based on characteristic line method: MN sub-groups are merged together to form a "multi-group" structure; the multi-group structure is applied to the sub-group fixed source transport calculation, and the characteristic line scanning will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all sub-groups;

[0029] The subgroup fixed source calculation module is used to adopt the multi-group grouping calculation method, assuming that the final multi-group transport calculation adopts the energy group number G, and uses the existing multi-group transport variables to perform subgroup fixed source calculations of the multi-group structure.

[0030] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform.

[0031] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform.

[0032] Another object of the present invention is to provide an information data processing terminal, which is used to implement the subgroup resonance computing system based on the CPU / GPU heterogeneous platform.

[0033] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0034] First, in view of the technical problems existing in the above-mentioned prior art, some creative technical effects are brought about after solving the problems. The specific description is as follows:

[0035] 1) Fast calculation speed: A technique for accelerating the solution of subgroup fixed source equations and fine group slowing equations using CPU / GPU heterogeneous platforms is proposed to improve parallel efficiency. By reasonably parallelizing computing resources on the CPU and GPU, the speed of subgroup resonance calculation is improved while ensuring accuracy.

[0036] 2) Low memory consumption: A multi-group solution idea for subgroup fixed source equations and slowing equations is proposed. Compared with the traditional group-by-group solution, it can significantly reduce the redundant calculation of geometric processing and feature line scanning in the transport module. The multi-group group calculation method is adopted to reduce the memory growth caused by multi-group calculation.

[0037] 3) Wide range of applications: This technology can not only be used for pressurized water reactor subgroup resonance calculations, but can also be applied to the effective cross-section calculations of heavy water reactors, liquid metal cooled fast reactors and other reactor types. It can also efficiently handle various complex conditional geometries and has a wide range of applicability.

[0038] second,

[0039] The present invention aims to solve the problem of low efficiency of existing subgroup resonance calculation methods. By introducing CPU / GPU heterogeneous platform and multi-group grouping calculation technology, the following technical effects and advantages are achieved:

[0040] 1) Fast calculation speed:

[0041] First, CPU / GPU heterogeneous parallelism is adopted: the feature line scanning calculation is assigned to the GPU for execution, and the powerful parallel computing capability of the GPU is used to significantly improve the calculation speed.

[0042] Secondly, multi-group grouping calculation is proposed: multiple sub-groups are merged into a multi-group structure for calculation, which reduces the number of feature line scans, reduces the amount of calculation, and improves parallel efficiency.

[0043] 2) Low memory consumption:

[0044] A multi-group grouping calculation is proposed. Compared with the traditional group-by-group solution, the multi-group grouping calculation reduces the redundant calls to the transport module and reduces the memory consumption.

[0045] 3) Wide application scope:

[0046] Firstly, the present invention is not only applicable to the resonance calculation of pressurized water reactor subgroups, but also can be applied to heavy water reactors, liquid metal cooled fast reactors and other reactor types, and has wide applicability.

[0047] Secondly, this technology can efficiently handle various complex geometric conditions and adapt to different reactor design requirements.

[0048] 4)Technological innovation:

[0049] Multi-group grouping calculation: A multi-group grouping calculation method is proposed for the sub-group fixed source equation and the slowing-down equation to reduce the redundant calculation amount of the transport module and improve the calculation efficiency.

[0050] CPU / GPU heterogeneous parallelism: feature line scanning calculations are assigned to the GPU for execution, and the powerful parallel computing capabilities of the GPU are utilized to achieve efficient computing.

[0051] Third, as auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

[0052] (1) The expected benefits and commercial value of the technical solution of the present invention after transformation are:

[0053] 1) Improve reactor design efficiency:

[0054] Shorten the design cycle: Subgroup resonance calculation is a key link in reactor design and safety analysis. The present invention can significantly improve the calculation speed, shorten the reactor design cycle, and accelerate the development and construction of new reactor types.

[0055] Reduce design costs: Improved computing efficiency means less input in computing resources and lower reactor design costs.

[0056] 2) Enhanced reactor safety:

[0057] More accurate calculation results: The present invention can improve calculation accuracy, provide a more reliable basis for reactor design and safety analysis, and enhance reactor safety.

[0058] More comprehensive analysis capabilities: Efficient computing capabilities allow for more detailed analysis, such as sensitivity analysis, uncertainty analysis, etc., providing more comprehensive information for reactor design and operation. (2) The technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0059] 1) Principle breakthrough: Existing subgroup resonance calculation methods are mainly based on the CPU platform, and fail to fully utilize the powerful parallel computing capabilities of GPU. This invention applies the CPU / GPU heterogeneous platform to subgroup resonance calculation for the first time, achieving a significant improvement in computing efficiency.

[0060] 2) Technical implementation: The existing subgroup resonance calculation method mainly adopts group-by-group solution, which has redundant calculation and low efficiency. The present invention proposes a multi-group grouping calculation method for subgroup fixed source equation and slowing-down equation, which reduces the number of characteristic line scans, reduces the amount of calculation, and improves the calculation efficiency.

[0061] 3) Practical application: The present invention proposes a two-level parallel computing mode of characteristic line and single energy group, which allocates the characteristic line scanning calculation to the GPU for execution, and executes the calculations such as section preparation and merging, source term initialization and update on the CPU side, giving full play to the computing advantages of the CPU and GPU and achieving efficient computing.

[0062] Fourth, the technical effects of the subgroup resonance calculation method based on CPU / GPU heterogeneous platform in industrial applications are mainly reflected in the following aspects:

[0063] 1) Significantly improve computational efficiency: By merging multiple subgroups into a multigroup structure, the independent solution process of each group and each subgroup is avoided, thereby reducing the repeated calls to the characteristic line method transport solver. This greatly reduces the number of characteristic line scans and communications, and improves computational efficiency. In the numerical simulation of nuclear reactor physics, this means faster computing speed and higher simulation efficiency, especially in complex scenarios where a large number of resonant nuclides need to be processed, which can effectively shorten the simulation time.

[0064] 2) Make full use of GPU parallel computing capabilities: This method is implemented on a CPU / GPU heterogeneous platform, distributing computing tasks to the GPU for parallel processing, which accelerates the calculation of the characteristic line method. The high parallel processing capability of the GPU can process a large number of characteristic line scanning tasks in a short time, thereby further improving the speed of subgroup resonance calculation. This has important practical significance for scenarios in the industry that require frequent high-fidelity simulations, such as reactor design optimization and safety analysis.

[0065] 3) Reduce computing resource consumption: Since the repeated calls of the characteristic line method and the communication overhead in the calculation process are reduced, this method can reduce the demand for computing resources while ensuring the calculation accuracy. This not only reduces the hardware investment required for simulation, but also reduces energy consumption and operating costs, and improves the economy and sustainability of the entire calculation process.

[0066] 4) Enhance the adaptability of industrial simulation systems: By adopting a multi-group grouping calculation method, the technology can flexibly adapt to different energy group structures and meet the needs of various complex working conditions in nuclear reactor physics simulation. This enables simulation tasks under different reactor types, different fuel types, and different operating conditions to be completed with higher accuracy and efficiency, enhancing the adaptability and reliability of industrial simulation systems.

[0067] 5) Supporting the widespread application of high-fidelity simulation: This method significantly improves the computational efficiency in numerical simulation of nuclear reactor physics, making high-fidelity simulation more feasible in actual industrial applications. For example, in the fields of nuclear reactor design, fuel optimization, accident analysis, operation and maintenance, high-fidelity simulation can provide more accurate data support, help optimize design decisions and improve safety performance, thereby promoting industrial technological progress. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0069] Figure 1 It is a flow chart of a nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform provided by an embodiment of the present invention;

[0070] Figure 2 It is a structural diagram of a subgroup resonance computing system based on a CPU / GPU heterogeneous platform provided by an embodiment of the present invention;

[0071] Figure 3 is a flow chart of subgroup fixed source equation calculation provided by an embodiment of the present invention; wherein (a) group-by-group calculation; (b) multi-group calculation;

[0072] Figure 4 is a flow chart of the calculation of the slowing-down equation provided by an embodiment of the present invention; wherein (a) group-by-group calculation; (b) multi-group calculation;

[0073] Figure 5 It is a flow chart of transport solution of subgroup fixed source equation or slowing-down equation based on GPU provided by an embodiment of the present invention;

[0074] Figure 6 is a diagram of calculation results of the subgroup fixed source equation provided by an embodiment of the present invention, wherein (a) calculation time; (b) number of iterations;

[0075] Figure 7 1 is a diagram of the calculation results of the slowing-down equation provided by an embodiment of the present invention, wherein (a) is the calculation time; (b) is the number of iterations;

[0076] Figure 8 is the computational acceleration ratio of the GPU provided by the embodiment of the present invention relative to the CPU: (a) subgroup fixed source equation; (b) slowed-down equation. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0078] In view of the problems existing in the prior art, the present invention provides a method and system for numerical simulation of nuclear reactor physics based on a CPU / GPU heterogeneous platform. The present invention is described in detail below in conjunction with the accompanying drawings.

[0079] The following are two specific examples of industrial applications based on the above technology:

[0080] 1) Nuclear power plant fuel assembly design and optimization

[0081] Background: The design of nuclear power plant fuel assemblies requires accurate neutron transport calculations to ensure safety and economy under different operating conditions. Traditional subgroup resonance calculation methods cannot quickly provide high-precision simulation results due to their low computational efficiency.

[0082] Example: By adopting a subgroup resonance calculation method based on a CPU / GPU heterogeneous platform, the design process of fuel assemblies can be greatly accelerated. When designing new fuel assemblies, the neutron flux distribution and reaction cross-section changes under different fuel rod arrangements, cladding materials, and coolant conditions can be quickly simulated. This efficient calculation method enables designers to evaluate multiple design schemes in a short period of time and optimize the design of fuel assemblies, thereby improving fuel utilization and reducing production costs. This method can also be used to analyze the performance of fuel assemblies under different burnup conditions and provide data support for the formulation of fuel management strategies.

[0083] 2) Analysis of nuclear power plant accident conditions

[0084] Background: During the operation of nuclear power plants, various accident condition analyses are required, such as loss of coolant accident (LOCA). These analyses require high-fidelity numerical simulations to accurately predict the reactor behavior after an accident, and then formulate effective emergency measures.

[0085] Embodiment: In LOCA accident simulation, the subgroup resonance calculation method based on the CPU / GPU heterogeneous platform can significantly improve the simulation speed. By combining the efficient characteristic line method with the subgroup resonance calculation, the neutron energy spectrum changes and the dynamic response of the reactor core after the accident can be quickly simulated. This technology enables the simulation system to provide accurate reactor state predictions in real time when an accident occurs, helping operators to make decisions quickly, thereby improving the efficiency of accident response and reducing potential safety risks. This technology can also be used to verify and optimize accident emergency plans to improve the overall safety of nuclear power plants.

[0086] like Figure 1 As shown, the nuclear reactor physics numerical simulation method based on the CPU / GPU heterogeneous platform provided by the embodiment of the present invention includes:

[0087] S1, in the resonance calculation process, it is assumed that on average each resonant nuclide needs to solve the subgroup fixed source equation N times; the subgroup transport calculation will loop each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation;

[0088] S2, multi-group transport calculation based on characteristic line method: MN sub-groups are merged together to form a "multi-group" structure; the multi-group structure is applied to the sub-group fixed source transport calculation, and the characteristic line scanning will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all sub-groups;

[0089] S3, adopts the method of multi-group grouping calculation, assuming that the final multi-group transport calculation uses the energy group number G, and uses the existing multi-group transport variables to perform sub-group fixed source calculations of the multi-group structure.

[0090] like Figure 2 As shown, the subgroup resonance computing system based on the CPU / GPU heterogeneous platform provided by the embodiment of the present invention includes:

[0091] The resonance calculation module is used to assume that, during the resonance calculation process, each resonant nuclide adopts the subgroup fixed source equation that needs to be solved N times on average; the subgroup transport calculation will loop each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation;

[0092] Multi-group transport calculation module, used for multi-group transport calculation based on characteristic line method: MN sub-groups are merged together to form a "multi-group" structure; the multi-group structure is applied to the sub-group fixed source transport calculation, and the characteristic line scanning will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all sub-groups;

[0093] The subgroup fixed source calculation module is used to adopt the multi-group grouping calculation method, assuming that the final multi-group transport calculation adopts the energy group number G, and uses the existing multi-group transport variables to perform subgroup fixed source calculations of the multi-group structure.

[0094] In view of the ability of the characteristic line method to handle multi-group transport, the present invention proposes a method to calculate the subgroup fixed source equation and the slowing-down equation in a "multi-group" manner, thereby reducing the redundant calculation amount of the transport module. The specific process is as follows:

[0095] During the resonance calculation, it is assumed that on average each resonant nuclide needs to solve the subgroup fixed source equation N times. Since the subgroup fixed source equation does not update the source term, the subgroups do not affect each other during the calculation. In the actual calculation process, the subgroup transport calculation will loop for each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation.

[0096] Based on the characteristic line method multi-group transport calculation, the present invention merges MN subgroups together to form a "multi-group" structure. The multi-group structure is applied to the subgroup fixed source transport calculation, and the characteristic line scan will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all subgroups. Multi-group calculation transfers the loop of the subgroup to the innermost side of the entire calculation loop, and a global modular characteristic line scan can perform flux calculation on all subgroups. The entire multi-group structure performs flux convergence at the same time, so the global iteration number is the same as the subgroup with the most group-by-group iterations. Since the subgroup fixed source equation is a pure absorption process, the source terms between each subgroup are the same and no source iteration is required. The number of iterations of different subgroups when solving group by group is basically the same, so the number of iterations of the subgroup fixed source equation of the multi-group structure is also basically the same as the group-by-group calculation. Compared with the frequent characteristic line scanning in group-by-group calculation, multi-group calculation can significantly reduce the amount of transport calculation.

[0097] However, the single group solution only needs to store the cross-section and flux information of a single energy group each time, while the memory required for the subgroup fixed source calculation of the multi-group structure is positively correlated with the total number of subgroups. For examples with more resonant nuclides, the memory usage of the multi-group fixed source calculation will increase significantly. In order to alleviate the computational memory pressure, the present invention adopts a multi-group grouping calculation method. Assuming that the final multi-group transport calculation uses an energy group number of G, the existing multi-group transport variables can be used to perform subgroup fixed source calculations of the multi-group structure. For MN subgroups (MN>G), each G subgroup can be divided into a group, and when there are less than G groups, they are divided into a group according to the original structure. Figure 1 The flowcharts of solving the subgroup fixed source equation group by group and solving the multi-group group by group are given. When solving the multi-group fixed source equation directly without grouping, only one global characteristic line scan is required. The number of characteristic line scans in grouped solution is the same as the number of groups. Although the amount of calculation is increased compared with the non-grouping solution, it can still be greatly reduced compared with solving group by group.

[0098] Compared with the characteristics of the subgroup fixed source equation that each group is solved independently, the slowing-down equation needs to establish a recursive relationship for solving each group. The calculation of the source term of the current energy group first needs to obtain the scalar flux of the upstream energy group. When solving the slowing-down equation of the "multi-group" structure composed of multiple energy groups at the same time, since the upstream energy group flux is still an unknown quantity in the first iteration, the source term of the downstream energy group is first calculated using the initial value of the flux, and the scattering source term is continuously updated during each iteration of the characteristic line method until the flux of all energy groups converges.

[0099] The idea of ​​multi-group grouping can also be applied to the solution of the slowdown equation. Combined with the characteristics of the top-down recursive solution of the slowdown equation, the group-by-group calculation can be adjusted to group-by-group calculation. Referring to the grouping method of the multi-group fixed source equation, assuming that the number of fine resonance energy groups is X, the X resonance fine groups are divided into Y groups from top to bottom according to energy, and each group has Z groups. Figure 4 The calculation flow chart of group-by-group solution and multi-group group solution of the slowdown equation is given. The multi-group group solution is recursively calculated in groups, and each group solves the multi-group slowdown equation, and updates the scattering source term in real time according to the current iteration step flux. After the slowdown flux of all energy groups in the current group converges, it is provided to the downstream group to calculate the scattering source term scattered from the current group to the downstream group. The calculation is recursively calculated group by group until the slowdown flux of all resonant fine groups converges.

[0100] Characteristic line scanning is the most computationally intensive part of the transport module, requiring cyclic calculations of all characteristic line segments and energy groups at each azimuth. Since the computational tasks of each part are heavy and relatively independent, combined with the parallel characteristics of the CUDA core, the two-layer cyclic calculations can be assigned to each thread for parallel solution. Each thread corresponds to only one energy group in a characteristic line segment, forming a two-level parallel calculation mode of characteristic line-single energy group. In combination with the above characteristics, in the calculation of subgroup fixed source equations and slowing-down equations, calculations such as section preparation and merging, source term initialization and updating can be performed on the CPU side, while transport scanning calculations are performed on the GPU side. Therefore, Figure 3 and Figure 4 The transport solution in Figure 5 As shown. The front-end calculation includes input definition, geometry initialization, database reading, etc., all of which are calculated on the CPU side. After the cross section and source terms are initialized, the CPU calls the GPU to perform characteristic line scanning calculations. After the converged fixed source flux or slowing flux is obtained, the effective cross section is merged and calculated on the CPU side. Subsequent calculations also include the final multi-group transport calculations, which call the GPU side again to solve the characteristic lines and obtain the final multi-group flux and effective proliferation factor.

[0101] The present invention is implemented by taking a series of problems from single gate element to component as examples. The case series components range from 3×3 to 17×17, and the computing platforms are all single CPU or single GPU core. The hardware used for the calculation is a Xeon E5-2690 v4 model CPU, whose parameters are 2.60GHz, 14 cores, and 28 threads; the graphics card is a K210 model GPU, each K210 GPU contains 13 stream processors, each stream processor contains 192 CUDA cores, and a total of 2496 CUDA cores can be allocated to multiple GPU threads, and each thread is independent and parallel to solve.

[0102] 1) Subgroup fixed source calculation

[0103] UO2 fuel contains 235 U and 238 U two resonant nuclides, a total of 16 subgroup fixed source equations need to be solved. In the multi-group solution mode, 16 subgroups form a multi-group structure, and the multi-group fixed source equations are solved by CPU and GPU transport solvers respectively, and the calculation time is compared with the group-by-group solution. The results are as follows Figure 6 As shown. The horizontal axis in the figure is the number of grid cells on each edge of the component. Figure 6 (a) The left ordinate is the total time for solving the subgroup fixed source equation, and the right ordinate is the ratio of the group-by-group calculation time to the multi-group calculation time, that is, the multi-group calculation speedup ratio. Figure 6 (b) The total number of MOC iterations for each calculation scheme is given.

[0104] Depend on Figure 6(a) It can be seen that compared with traditional group-by-group computing, multi-group computing has a better acceleration effect on the subgroup fixed source equation on the CPU and GPU side. Figure 6 (b) It can be seen that on the same computing platform, the total number of MOC iterations for the fixed source equations of the group-by-group and multi-group subgroups is almost unchanged, with only a slight difference in the minimum scale components. For CPU solving, the speedup ratio increases with the increase in the calculation scale, and the speedup ratio for 17×17 components is 1.46. For GPU solving, multi-group computing increases the parallelism between subgroups on the basis of reducing the repetitive feature line scanning of group-by-group computing, so the acceleration effect is more obvious, with a speedup ratio of 5 to 10 times for different components. In summary, compared with group-by-group computing, multi-group computing has a significant acceleration effect on the subgroup fixed source equations on both the CPU and GPU sides, and the speedup ratio on the GPU side is even greater.

[0105] 2) Calculation of resonance fine group slowing-down equation

[0106] Figure 7 The computation time of the slowed-down equations solved by group-by-group and multi-group on the CPU and GPU sides, as well as the total number of MOC iterations of the slowed-down equations for each example are given. On the CPU side, the amount of characteristic line scanning calculations reduced by multi-group calculations cannot offset the increase in computational time caused by the increase in the total number of MOC iterations. As the scale increases, the multi-group speedup ratio drops to less than 0.3, indicating that multi-group calculations have significantly reduced the computational efficiency. However, the performance of multi-group calculations on the GPU side is completely opposite to that of the CPU, and multi-group calculations still have a good acceleration effect on the slowed-down equations on the GPU. Although the number of iterations has increased, the parallel characteristics of multi-groups still have a higher efficiency improvement compared to group-by-group calculations.

[0107] Even under the same solution model, GPU still has a significant efficiency improvement compared to CPU. Figure 8 The calculation speedup ratio of GPU relative to CPU under the same calculation model is given. As the scale of components increases, the advantage of GPU parallel processing becomes more obvious, so the speedup ratio is larger. For solving the subgroup fixed source equation, the GPU speedup ratio of group-by-group calculation is 8.56 for 17×17 components. Under multi-group calculation conditions, due to the increase of subgroup parallelism, the acceleration effect of GPU is more obvious, and the speedup ratio for 17×17 components is 30.17. For multi-group slowed equations, the increase in the number of MOC iterations leads to a decrease in the calculation efficiency on the CPU side, while the parallel setting on the GPU side still has a high calculation efficiency. Therefore, the GPU speedup ratio of the multi-group slowed equation is further improved, and the speedup ratio for 17×17 components reaches 73.47.

[0108] In summary, for the subgroup fixed source equation, multi-group calculation has a good acceleration effect on both the CPU and GPU sides, and the GPU has a higher computational efficiency than the CPU. For the slowed-down equation, since the multi-group calculation needs to update the source term, the number of MOC iterations increases. The multi-group slowed-down calculation is less efficient on the CPU platform and has no acceleration effect. For the GPU platform, due to the parallel calculation of the characteristic line-energy group, the multi-group slowed-down has a higher computational efficiency than the group-by-group calculation.

[0109] An application embodiment of the present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of a nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform.

[0110] An application embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of a nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform.

[0111] An application embodiment of the present invention provides an information data processing terminal, which is used to implement a subgroup resonance computing system based on a CPU / GPU heterogeneous platform.

[0112] In order to prove the creativity and technical value of the technical solution of the present invention, this section provides application examples of the claimed technical solution on specific products or related technologies.

[0113] 1) Design of new reactors: The efficient calculation method of the present invention can provide fast and accurate calculation results for the design and optimization of new reactors, thus accelerating the research and development process of new reactors.

[0114] 2) Nuclear power plant safety analysis:

[0115] Accident analysis: The present invention can be used to analyze the causes and processes of nuclear power plant accidents and provide decision support for emergency response to accidents.

[0116] Sensitivity analysis: The present invention can be used to analyze the impact of reactor parameters on reactor performance and safety, and provide optimization solutions for reactor design and operation.

[0117] Uncertainty analysis: This invention can be used to assess the uncertainty in reactor design and operation and improve the safety of nuclear power plants.

[0118] 3) Nuclear fuel cycle:

[0119] The present invention can be used to analyze the performance of nuclear fuel and provide support for fuel design and optimization, and can also be used to analyze the performance of industrial nuclear reactors and provide support for industrial applications.

[0120] 4) Scientific research and teaching:

[0121] Nuclear reactor physics research: This invention can provide efficient computing tools for nuclear reactor physics research and accelerate the scientific research process.

[0122] Nuclear engineering education: The present invention can be used as an important tool for nuclear engineering education to help students better understand and master nuclear reactor physics knowledge.

[0123] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0124] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the field of the present invention within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A nuclear reactor physics numerical simulation method based on CPU / GPU heterogeneous platform, characterized in that: The following steps are involved: -This method can be used to quickly simulate the neutron flux distribution and reaction cross section changes under different fuel rod arrangements, cladding materials and coolant conditions; - Allows designers to evaluate multiple fuel assembly designs in a short period of time; -Optimize fuel assembly design to improve fuel utilization and reduce production costs; - Further used to analyze the performance of fuel assemblies under different burnup conditions and provide data support for the formulation of fuel management strategies; Among them, the method significantly improves the computing efficiency and simulation accuracy in the nuclear fuel assembly design and optimization process by utilizing the efficient computing power of the CPU / GPU heterogeneous platform.

2. The method for numerical simulation of nuclear reactor physics based on a CPU / GPU heterogeneous platform according to claim 1, characterized in that: include: S1, in the resonance calculation process, on average each resonant nuclide adopts the fixed source equation that needs to be solved N times of subgroups; The subgroup transport calculation will loop for each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation; S2, multi-group transport calculation based on characteristic line method: MN sub-groups are merged together to form a "multi-group" structure; The multi-group structure is applied to the sub-group fixed source transport calculation, and the characteristic line scanning will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all sub-groups; S3, adopts the method of multi-group grouping calculation, and the final multi-group transport calculation uses the energy group number G, and uses the existing multi-group transport variables to perform sub-group fixed source calculations of the multi-group structure.

3. The method for numerical simulation of nuclear reactor physics based on CPU / GPU heterogeneous platform according to claim 1, characterized in that: In S2, the multi-group calculation transfers the loop of subgroups to the innermost part of the entire calculation loop, and a global modular characteristic line scan can perform flux calculation on all subgroups; the entire multi-group structure performs flux convergence at the same time, so the number of global iterations is the same as the subgroup with the largest number of group-by-group iterations; since the subgroup fixed source equation is a pure absorption process, the source terms between subgroups are the same and no source iteration is required, the number of iterations of different subgroups when solving group by group is basically the same, so the number of iterations of the subgroup fixed source equation of the multi-group structure is also basically the same as the group-by-group calculation.

4. The method for numerical simulation of nuclear reactor physics based on CPU / GPU heterogeneous platform according to claim 1, characterized in that: In S3, for MN subgroups, every G subgroups can be divided into a group. When there are less than G groups, they can be divided into one group according to the original structure. When the multi-group fixed source equations are solved directly without grouping, only one global characteristic line scan is required. The number of characteristic line scans during group solution is the same as the number of groups. Although the amount of calculation is increased compared to no grouping, it can still be greatly reduced compared to solving group by group.

5. The method for numerical simulation of nuclear reactor physics based on CPU / GPU heterogeneous platform according to claim 1, characterized in that: The slowing-down equation needs to establish a recursive relationship for solving each group. The calculation of the source term of the current energy group first needs to obtain the scalar flux of the upstream energy group. When solving the slowing-down equation of the "multi-group" structure composed of multiple energy groups at the same time, since the upstream energy group flux is still an unknown quantity in the first iteration, the source term of the downstream energy group is first calculated using the initial value of the flux, and the scattering source term is continuously updated during each iteration of the characteristic line method until the flux of all energy groups converges.

6. The method for numerical simulation of nuclear reactor physics based on CPU / GPU heterogeneous platform according to claim 1, characterized in that: Referring to the grouping method of the multi-group fixed source equation, assuming that the number of fine resonance energy groups is X, the X resonance fine groups are divided into Y groups from top to bottom according to energy, and each group has Z groups; The multi-group group solution is recursively calculated in groups. Each group solves the multi-group slowing-down equation and updates the scattering source term in real time according to the current iteration step flux. After all the energy group slowing fluxes in the current group converge, they are provided to the downstream groups to calculate the scattering source terms from the current group to the downstream groups. The calculation is repeated group by group until all the resonant fine group slowing fluxes converge.

7. The method for numerical simulation of nuclear reactor physics based on CPU / GPU heterogeneous platform according to claim 1, characterized in that: Each thread corresponds to only one energy group in a characteristic line segment, forming a two-level parallel computing mode of characteristic line-single energy group; in the calculation of subgroup fixed source equation and slowing-down equation, the calculations such as section preparation and merging, source term initialization and updating are executed on the CPU side, and the transport scanning calculation is performed on the GPU side.

8. A subgroup resonance computing system based on a CPU / GPU heterogeneous platform using the nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform as claimed in any one of claims 1 to 7, characterized in that: include: The resonance calculation module is used to assume that, during the resonance calculation process, each resonant nuclide adopts a fixed source equation that needs to be solved N times of subgroups; The subgroup transport calculation will loop for each subgroup of each resonant nuclide in the calculation system. For M resonant nuclides, the subgroup fixed source equation calls the characteristic line method transport solver MN times, and each call requires a characteristic line scanning calculation; Multi-group transport calculation module, used for multi-group transport calculation based on characteristic line method: MN sub-groups are merged together to form a "multi-group" structure; The multi-group structure is applied to the sub-group fixed source transport calculation, and the characteristic line scanning will be performed simultaneously in the entire multi-group structure to achieve the simultaneous solution of the fixed source flux of all sub-groups; The subgroup fixed source calculation module is used to adopt the multi-group grouping calculation method, assuming that the final multi-group transport calculation adopts the energy group number G, and uses the existing multi-group transport variables to perform subgroup fixed source calculations of the multi-group structure.

9. A computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the nuclear reactor physics numerical simulation method based on a CPU / GPU heterogeneous platform as described in any one of claims 1 to 6.

10. An information data processing terminal, used for implementing the subgroup resonance computing system based on a CPU / GPU heterogeneous platform as claimed in claim 7.

Images