Nuclear power plant safety analysis method using RELAP5 and MCDET coupled parallel computation and DET branch truncation

Abstract

The application discloses a nuclear power plant safety analysis method adopting RELAP5 and MCDET coupling parallel computing and DET branch truncation, the method comprises the following steps: configuring a corresponding RELAP5 thermal hydraulic file according to a set number of running threads; generating an input i file according to MC sampling and DET branch rules; in the DET branch process, calling RELAP5 according to the branch state of each node, calculating the probability of each DET branch, and truncating the process with the minimum branch probability. The application realizes root node selection by adopting a Monte Carlo (MC) random sampling method, improves the calculation efficiency by coupling MCDET and RELAP5 programs through parallel computing, and stops the calculation consumption of the DET branch with extremely low occurrence probability by truncating the DET branch with extremely low occurrence probability, so that the calculation time is reduced under the condition of ensuring the calculation accuracy, and the method is an important improvement and improvement on the nuclear power plant safety analysis method.

Description

Technical Field

[0001] The technical solution of this application belongs to the field of nuclear science and engineering technology. In particular, when conducting safety analysis on the operation of nuclear power facilities (nuclear power plants), it utilizes the automatic coupling of MCDET (Monte Carlo Dynamic Event Tree) and RELAP5 to implement parallel computing, and adopts the low-probability branch truncation method of DET (Dynamic Event Tree) for efficient and high-accuracy nuclear power plant safety analysis. That is, a nuclear power plant safety analysis method that uses parallel computing coupled with RELAP5 and MCDET and DET branch truncation. Background Technology

[0002] Nuclear power plant safety is the most critical aspect of its operation. Safety analysis and assessment must be conducted during the design and construction of nuclear power facilities and systems. Common analysis and assessment methods fall into two categories: deterministic safety analysis and probabilistic safety evaluation. Deterministic safety analysis commonly uses RELAP5 software, while probabilistic safety evaluation is based on traditional event tree / fault tree methods. However, either a single deterministic or probabilistic method has significant limitations in analyzing the dynamic characteristics of nuclear power plant accidents. To overcome the shortcomings of deterministic and probabilistic safety analyses, random sampling is typically used for both types of parameters to construct an uncertainty parameter model of the nuclear power plant's input space. This model aims to simulate the actual operating conditions of the nuclear power plant as closely as possible, yielding more detailed and accurate safety analysis results. Of the two types of parameters mentioned above, one type is the physical process parameter of the initial state of the nuclear power plant, such as power, temperature, and pressure, which specifies the initial state of the nuclear power plant. It is generally simulated using Monte Carlo (MC) sampling. The other type is the random state transition parameter, such as the operating time of equipment or system, which reflects the changes in the state of the nuclear power plant during operation. In order to achieve explicit modeling of the possible complex interaction models between hardware / software / processes / human behavior in the evolution process of complex systems, the Discrete Dynamic Event Tree (DDET) method was developed. Moreover, the main difference between the Discrete Dynamic Event Tree method and traditional analysis methods is that the physical process of a nuclear power plant accident is driven by the dynamic simulation evolution of the nuclear power plant system, which reduces the dependence on the experience of analysts and reduces unnecessary conservative engineering judgments.

[0003] The MCDET method combines MC sampling and DET methods. Through comprehensive uncertainty analysis of the initial state physical process parameters and stochastic state transition parameters of a nuclear power plant, it aims to approximate the actual response and operation process after a nuclear power plant accident as closely as possible, thus obtaining more detailed safety analysis results. Figure 1The MCDET (Multi-Channel Difference-Effective) simulation process is presented. It is assumed that n MC (Multi-Channel Difference) samplings are performed on the initial physical process parameters of a nuclear power plant to obtain n initial state models. These models are used as the root nodes of the DET (Difference-Effective) branches to establish DET branch models to simulate the nuclear power plant response process. When simulating the state evolution of a nuclear power plant, the MCDET method starts from the root node. After a certain time, branches are generated based on the state changes of the nuclear power plant equipment or systems, creating intermediate nodes. After traversing all branches, the simulation ends at the final node. The state of the final node determines whether the nuclear power plant is safe or has failed.

[0004] When using the MCDET method for accident analysis, to ensure the accuracy of the calculation results, extensive sampling is typically performed, and a detailed DET model is established to comprehensively simulate the state changes of nuclear power plants under accident conditions. This generates numerous branches. However, the RELAP5 program itself is a single-threaded software, and only one input file can be specified for a single simulation during the RELAP5 calculation process. If multiple accidents need to be simulated, RELAP5 needs to be called multiple times. In the coupled calculation process of RELAP5 and the MCDET method, if only single-threaded calculation can be performed each time, it will incur an unacceptable time cost. Therefore, a multi-threaded parallel calculation method is required. Running multiple RELAP5 instances simultaneously can also effectively reduce calculation time and improve calculation efficiency. At the same time, branch truncation is performed for DET branches with extremely low probability of occurrence to eliminate the calculation of branches with little impact and improve calculation efficiency. This application is an improvement on the coupled calculation method of RELAP5 and MCDET, proposing a nuclear power plant safety analysis method that uses coupled parallel calculation of RELAP5 and MCDET and DET branch truncation. Summary of the Invention

[0005] This application proposes an algorithm improvement to address the shortcomings of combining the MCDET method with the RELAP5 program, which leads to reduced computational efficiency due to single-threaded computation and excessive computational consumption for the DET branch, which has an extremely low probability of occurrence. Specifically, it proposes a nuclear power plant safety analysis method that couples RELAP5 and MCDET for parallel computation and uses DET branch truncation. The specific technical steps of this method include:

[0006] (1) Generate n root nodes by MC (Monte Carlo) sampling, and select one of the root nodes to start DET calculation;

[0007] (2) Set up the DET branch calculation thread and copy the thermal hydraulic files according to the number of threads and name them sequentially;

[0008] (3) Start the first calculation from the i-th file of the root node. Name the root node folder root.i in the root folder. Call RELAP5 to run the calculation and get the corresponding output root.o and root.r files. The root node is the initial parent node.

[0009] (4) Generate child node branches starting from the parent node. Starting from the branch time corresponding to the branch node, generate the corresponding restart input *.i file according to the DET branch logic. The newly generated node serves as the parent node of the next level child node.

[0010] (5) Copy the r file from the parent node folder to each child node folder, and modify the name of the copied r file so that the name of the copied r file is the same as the name of the i file in this child node folder;

[0011] (6) Calculate the branch probability of the child node and determine whether the branch probability is lower than the set probability threshold: if it is lower than the set probability threshold, terminate the branch of the node, start the calculation of the next node, and continue to the next step; if it is not lower than the set probability threshold, directly run the next step.

[0012] (7) Determine the relationship between the number of branches and the number of threads: If the number of branches is less than the number of threads, start multi-threaded parallel computation with the number of branches; otherwise, start multi-threaded parallel computation with the number of threads.

[0013] (8) Call the i file and r file to restart the calculation, obtain the output o file, and update the r file in this node to obtain the r file after running in the DET branch state, which serves as the basis for restarting the calculation of the next level branch.

[0014] (9) Determine whether the calculation of this node branch is complete: If it is complete, continue to determine whether the DET branch has been traversed, and then execute the next step; if it is not complete, continue to start several threads to calculate this node branch;

[0015] (10) Determine whether the DET branch has been traversed: If the traversal calculation is completed, determine whether the DET branch model corresponding to all root nodes generated by MC sampling has been calculated, i.e., the next step; otherwise, return to step (3) to start the calculation of the next child node branch.

[0016] (11) Determine whether the DET branch model corresponding to all root nodes generated by MC sampling has been calculated. If the DET branch model corresponding to all root nodes has been calculated, the calculation ends; otherwise, return to step (1) to continue execution.

[0017] When calling RELAP5 to run the calculation in step (3) above, the r file of the previous node is automatically deleted and the r file of the current node is regenerated. Therefore, the r file needs to be copied, that is, the *.r file is copied from the parent node folder to each child node folder, and the name of the copied r file is modified so that the name of the copied r file is the same as the i file in the current child node folder.

[0018] The branch probabilities in step (6) above are divided into demand-type branch probabilities and operation-type branch probabilities. The probability of a successful state transition and the probability of a failed state transition are (1-P1) and P1, respectively. The operation-type branch probabilities follow a cumulative probability density distribution cdf(t), and the branch probabilities corresponding to the branch time t are P(t1), P(t2)-P(t1), P(t3)-P(t2)...1-P(t3 ... m The branch probability of each child node is the product of the probabilities of all branches on its branch path; here, P(t1)....P(t) m ) represents different times t1…t m The probability of a failed branch during a state transition.

[0019] The calculation of the branch probability of each child node requires first determining the branch type of the node and calculating the branch probability of the node, and then multiplying the branch probability of the node with the branch probability of the parent node to obtain the final probability of the node.

[0020] In step (9) above, the calculation of the number of all branches on the node is completed before the calculation of the next child node is started.

[0021] This application employs the Monte Carlo (MC) random sampling method to select the root node. By coupling MCDET with the RELAP5 program, computational efficiency is improved through parallel computing. At the same time, in order to reduce the amount of computation, DET branches with extremely low occurrence probability are truncated and their computation is stopped. This shortens the computation time while ensuring computational accuracy. This method is an important improvement and enhancement to the safety analysis method for nuclear power plants. Attached Figure Description

[0022] Figure 1 This is a diagram illustrating the MCDET operation process;

[0023] Figure 2 A diagram illustrating the branch probabilities corresponding to different branch times;

[0024] Figure 3 A schematic diagram illustrating the branch probability calculation process for demand-based and operation-based branches;

[0025] Figure 4 Illustration of MCDET branch calculation thread setup and file allocation;

[0026] Figure 5 This is a flowchart of the nuclear power plant safety analysis method using RELAP5 coupled with MCDET parallel computing and DET branch truncation, as described in this application. Detailed Implementation

[0027] To more clearly illustrate the security analysis method of this application, the technical solution of this application is described below with reference to the accompanying drawings. First, a brief explanation of file calls in RELAP5 calculations is given: When calculating each thread, RELAP5 requires specifying the paths to its input *.i files, output *.o files, *.r files, and the thermal hydraulic file tpfh2o. Here, "*" represents any filename that meets the requirements of the RELAP5 program, and ".i", ".o", and ".r" represent file types.

[0028] The DET branch states are explained as follows: There are two types of branches: one is the demand-based branch, which is divided into a successful state transition branch and a failed state transition branch. If we assume the probability of a failed state transition branch is P1, then the probability of a successful state transition branch is (1-P1); the other is the running branch, whose generation is related to the running time. Assuming the running branch follows a cumulative probability density distribution cdf(t), such as... Figure 2 As shown, the branch probabilities corresponding to the branch time t are P(t1), P(t2)-P(t1), P(t3)-P(t2)...1-P(t3) respectively. m The branch probability calculation process for demand-based and operation-based branches is as follows: Figure 3 As shown, it represents that the branch probability of each endpoint is the product of the branch probabilities of all branches on its branch path.

[0029] During the parallel computation of DET branches, the RELAP5 restart computation function is required. Restarting computation means continuing the computation from the previous result; therefore, restarting computation requires specifying the *.r file output from the previous computation, and the next computation segment is based on the results of the *.r file from the previous computation. The number of DET branches gradually increases as the computation progresses. For each branch, the parallel computation process copies the *.r file from the previous branch node to the current branch folder and starts the computation on that branch. The specific computation flow is as follows: Figure 4 As shown.

[0030] Based on a clear understanding of file allocation, DET branch states and probabilities, and the parallel computation process of DET branches in the RELAP5 calculation, the specific steps of the nuclear power plant safety analysis method using RELAP5 coupled with MCDET parallel computation and DET branch truncation in this application are as follows:

[0031] S1. The MC method is used to sample and generate n root nodes, and one of the root nodes is selected to start the DET calculation.

[0032] S2. Set the number of calculation threads for the DET branch to x, then copy x thermal hydraulic files tpfh2o and name them sequentially from tpfh2o_1 to tpfh2o_x. In subsequent multi-threaded calculations, provide each thread with an independent thermal hydraulic file to prevent thread conflicts.

[0033] S3. Start the first calculation from the root node i file. Name the root node folder root. Place the root node root.i file in the root folder. Call RELAP5 to run the calculation and obtain the corresponding output root.o and root.r files. The root node is the initial parent node.

[0034] S4. Generate child node branches starting from the parent node. Starting from the branch time corresponding to the branch node, generate the corresponding restart input i file according to the DET branch logic. The newly generated node becomes the parent node of the next level child node.

[0035] S5. Copy the r file from the parent node folder to each child node folder, and modify the name of the copied r file so that the name of the copied r file is the same as the i file in this child node folder;

[0036] S6. Calculate the branch probability of the child node and determine whether the branch probability is lower than the set probability threshold. If it is lower than the set probability threshold, terminate the branch of the node, start the calculation of the next node, and continue to step S6; if it is not lower than the set probability threshold, directly continue to step S7.

[0037] S7. Determine if the number of branches is less than the number of threads. If the number of branches is less than the number of threads, start multi-threaded parallel computation using the number of branches; otherwise, start multi-threaded parallel computation using the number of threads.

[0038] S8. Call the i file and r file to restart the calculation, obtain the output o file, and update the r file in this node to obtain the r file after running in the DET branch state, which serves as the basis for restarting the calculation in the next level branch.

[0039] S9. Determine whether the calculation of this node branch is complete. If it is complete, continue to determine whether the DET branch has been traversed. If it is not complete, continue to start several threads to calculate this node branch.

[0040] S10. Determine whether the DET branch has been traversed. If the traversal calculation is complete, determine whether the DET branch model corresponding to all root nodes generated by MC sampling has been calculated. Otherwise, return to step S3 to start the calculation of the next child node branch.

[0041] S11. Determine whether the DET branch models corresponding to all root nodes generated by MC sampling have been calculated. If the DET branch models corresponding to all root nodes have been calculated, then end the calculation.

[0042] The specific embodiments described in this invention are merely illustrative of the invention and are not intended to limit it. Those skilled in the art can make modifications to these embodiments without contributing any inventive step after reading this specification, but such modifications are protected by patent law as long as they fall within the scope of the claims of this invention.

Claims

1. A nuclear power plant safety analysis method employing RELAP5 and MCDET coupled parallel computation and DET branch truncation, characterized in that: The method includes the following steps: (1) MC sampling generates n root nodes, and selects one of the root nodes to start DET calculation; (2) Set up the DET branch calculation thread and copy the thermal hydraulic files according to the number of threads and name them sequentially; (3) Start the first calculation from the i-th file of the root node. Name the root node folder root.i in the root folder. Call RELAP5 to run the calculation and get the corresponding output root.o and root.r files. The root node is the initial parent node. (4) Generate child node branches starting from the parent node. Starting from the branch time corresponding to the branch node, generate the corresponding restart input *.i file according to the DET branch logic. The newly generated node serves as the parent node of the next level child node. (5) Copy the r file from the parent node folder to each child node folder, and modify the name of the copied r file so that the name of the copied r file is the same as the name of the i file in this child node folder; (6) Calculate the branch probability of the child node and determine whether the branch probability is lower than the set probability threshold: if it is lower than the set probability threshold, terminate the branch of the node, start the calculation of the next node, and continue to run the next step; if it is not lower than the set probability threshold, run the next step directly. (7) Determine the relationship between the number of branches and the number of threads: If the number of branches is less than the number of threads, start multi-threaded parallel computation with the number of branches; otherwise, start multi-threaded parallel computation with the number of threads. (8) Call the i file and r file to restart the calculation, obtain the output o file, and update the r file in this node to obtain the r file after running in the DET branch state, which serves as the basis for restarting the calculation of the next level branch. (9) Determine whether the calculation of this node branch is complete: If it is complete, continue to determine whether the DET branch has been traversed, and then execute the next step; if it is not complete, continue to start several threads to calculate this node branch. (10) Determine whether the DET branch has been traversed: If the traversal calculation is completed, determine whether the DET branch model corresponding to all root nodes generated by MC sampling has been calculated, i.e., the next step; otherwise, return to step (3) to start the calculation of the next child node branch. (11) Determine whether the DET branch model corresponding to all root nodes generated by MC sampling has been calculated. If the DET branch model corresponding to all root nodes has been calculated, the calculation ends; otherwise, return to step (1) to continue execution.

2. The nuclear power plant safety analysis method using RELAP5 and MCDET coupled parallel computation and DET branch truncation as described in claim 1, is characterized in that, When calling RELAP5 to run the calculation in step (3), the new r file generated by this calculation will be automatically deleted and generated first.

3. The nuclear power plant safety analysis method using RELAP5 and MCDET coupled parallel computation and DET branch truncation as described in claim 1, is characterized in that, The branch probabilities in step (6) are divided into demand-type branch probabilities and operation-type branch probabilities; the probability of a successful state transition and a failed state transition for demand-type branch probabilities are P1 and (1-P1), respectively; the operation-type branch probabilities follow a cumulative probability density distribution cdf(t), and the branch probabilities corresponding to branch time t are P(t1), P(t2)-P(t1), P(t3)-P(t2)...1-P(t1), respectively. m The branch probability of each child node is the product of the probabilities of all branches on its branch path; here, P(t1)....P(t) m ) represents different times t1…t m The probability of a failed branch during a state transition.

4. The nuclear power plant safety analysis method using RELAP5 and MCDET coupled parallel computation and DET branch truncation as described in claim 3, is characterized in that, The probability of each child node branch is calculated by first determining the branch type of the current node and calculating the branch probability of the current node, and then multiplying the branch probability of the current node by the branch probability of the parent node to obtain the final probability of the current node.

5. The nuclear power plant safety analysis method using RELAP5 and MCDET coupled parallel computation and DET branch truncation as described in claim 1, characterized in that, In step (9), the calculation of the number of all branches on the node is completed before the calculation of the next child node is started.

Images