Construction method, device and equipment of reactor safety analysis model and storage medium

Abstract

The invention relates to the technical field of nuclear reactor safety analysis, and provides a reactor safety analysis model construction method, device and equipment and a storage medium, and the method comprises the steps: taking a target parameter related to a transient working condition of a reactor as a target output, and recognizing a plurality of input parameters corresponding to the target parameter by constructing a phenomenon recognition sorting table, a plurality of important parameters are screened out from a plurality of input parameters by using a moment independent global sensitivity analysis method, target input corresponding to the target parameters is constructed after the plurality of important parameters are randomly sampled for multiple times, and then a neural network model is trained by using a sample set constructed according to the target input and the target output; and obtaining a reactor safety analysis model. According to the method provided by the invention, the high-precision construction of the nuclear reactor multi-field coupling system transient safety analysis substitution model can be realized, and the constructed reactor safety analysis model can obtain relatively high precision with a relatively small sample, so that the method is used for nuclear reactor fault diagnosis and accident prediction.

Description

Technical Field

[0001] The present application relates to the technical field of reactor safety analysis, and in particular to a method, device, equipment and storage medium for constructing a reactor safety analysis model. Background Art

[0002] In the transient safety analysis of nuclear reactors, computer programs are often relied upon for modeling and simulation. Since a nuclear reactor is a highly nonlinear system, there is a very complex and difficult-to-quantify functional relationship between its input and output, so it is difficult to use conventional methods to predict the accident conditions of a nuclear reactor. Therefore, for complex nuclear reactor systems and thermal-hydraulic phenomena existing in the transient process of nuclear reactor accidents, it is necessary to develop a multi-physics field coupling system and use a high-precision program model to provide a basis for the transient safety analysis of nuclear reactors.

[0003] However, when multiple physics are coupled, the number of models included in the computer program is very large, which makes the computer program very time-consuming to run. For a single load case, it may take several days to perform calculations using a multiphysics coupled system.

[0004] In order to overcome the time-consuming and inaccurate nature of safety analysis in high-precision simulated multi-physics field coupling systems, and to meet the needs of accident prediction and fault diagnosis, it is necessary to study alternative models for high-precision nuclear reactor transient safety analysis. Summary of the invention

[0005] To solve the above problems, the present application provides a method, device, equipment and storage medium for constructing a reactor safety analysis model, taking the target parameters related to the transient operating conditions of the reactor as the target output, constructing a phenomenon recognition sorting table to first identify multiple input parameters, and using moment independent sensitivity measurement to screen important parameters that have an important impact on the target output from the input parameters, and constructing the target input based on the important parameters, so as to train the neural network model based on the target input and target output, and obtain a reactor safety analysis model that can accurately predict the transient operating conditions of nuclear reactor coupling system accidents.

[0006] This application adopts the following technical solutions:

[0007] In a first aspect, the present application provides a method for constructing a reactor safety analysis model, the method comprising:

[0008] Acquire the transient operating condition of the reactor, determine the target parameters related to the transient operating condition, and use the target parameters as target output;

[0009] Constructing a phenomenon recognition ranking table related to the target parameter, and identifying multiple input parameters corresponding to the target parameter according to the phenomenon recognition ranking table;

[0010] The moment-independent global sensitivity analysis method is used to rank the importance of multiple input parameters, and multiple important parameters are selected from the multiple input parameters according to the importance ranking;

[0011] Perform multiple random sampling on multiple important parameters, and construct the target input corresponding to the target parameters based on the random sampling results;

[0012] A sample set is constructed according to the target input and target output, and the neural network model is trained using the sample set to obtain a reactor safety analysis model.

[0013] In a second aspect, the present application provides a device for constructing a reactor safety analysis model, the device comprising:

[0014] A target output determination unit, used to obtain the transient operating conditions of the reactor, determine target parameters related to the transient operating conditions, and use the target parameters as target output;

[0015] An input parameter determination unit, used to construct a phenomenon recognition ranking table related to the target parameter, and identify multiple input parameters corresponding to the target parameter according to the phenomenon recognition ranking table;

[0016] An important parameter determination unit is used to sort the importance of multiple input parameters by using a moment-independent global sensitivity analysis method, and select multiple important parameters from the multiple input parameters according to the importance sorting;

[0017] A target input determination unit, used for performing multiple random samplings on multiple important parameters, and constructing a target input corresponding to the target parameters based on the results of the multiple random samplings;

[0018] The model training unit is used to construct a sample set according to the target input and target output, and use the sample set to train the neural network model to obtain a reactor safety analysis model.

[0019] In a third aspect, the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for constructing a reactor safety analysis model when executing the computer program.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, which implements the steps of the method for constructing the reactor safety analysis model when instructed by a processor.

[0021] At least one of the above technical solutions adopted in this application can achieve the following beneficial effects:

[0022] The method for constructing a reactor safety analysis model provided by the present application first obtains the transient operating conditions of the reactor, determines the target parameters related to the transient operating conditions, and uses the target parameters as the target output; then constructs a phenomenon recognition ranking table related to the target parameters, and identifies multiple input parameters corresponding to the target parameters according to the phenomenon recognition ranking table; then uses a moment independent global sensitivity analysis method to sort the importance of multiple input parameters, and selects multiple important parameters from multiple input parameters according to the importance sorting; then performs multiple random sampling on multiple important parameters, and constructs the target input corresponding to the target parameter based on the results of multiple random sampling; finally, constructs a sample set based on the target input and the target output, and uses the sample set to train the neural network model to obtain the reactor safety analysis model. The method for constructing a reactor safety analysis model provided by the present application determines the transient operating conditions and target parameters of the reactor, establishes PIRT to determine the input parameters, quantifies the uncertainty distribution of the input parameters and performs moment independent global sensitivity analysis, and selects the previous input parameters with a higher importance as the target input, thereby constructing a sample set training neural network model based on the target input and the target output. The method for constructing a reactor safety analysis model provided in the present application can realize the construction of a high-precision transient safety analysis alternative model for a nuclear reactor multi-field coupling system. The constructed reactor safety analysis model can obtain higher accuracy with a smaller sample, so as to be used for nuclear reactor fault diagnosis and accident prediction. The reactor safety analysis model constructed by the present application can effectively guide the operator to control the process of the reactor accident, and at the same time provide predictive protective measures for the safe operation of the reactor. Due to its rapid calculation, the neural network can predict the process of nuclear reactor accidents in real time, and buy more reaction time for the operator after the accident occurs. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0024] Figure 1 A schematic diagram showing a flow chart of a method for constructing a reactor safety analysis model according to an embodiment of the present application;

[0025] Figure 2 A schematic diagram showing an unconditional PDF and a conditional PDF output according to an embodiment of the present application;

[0026] Figure 3 A schematic diagram showing an unconditional CDF and a conditional CDF of an output according to an embodiment of the present application;

[0027] Figure 4A schematic flow chart showing a method for constructing a reactor safety analysis model according to another embodiment of the present application;

[0028] Figure 5 A schematic structural diagram showing a device for constructing a reactor safety analysis model according to an embodiment of the present application;

[0029] Figure 6 A schematic diagram of the structure of a computer device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0031] The idea of ​​this application is that neural networks can accurately predict high-dimensional nonlinear models and have strong feasibility, so they can be used as alternative models for multi-physics field coupling systems. However, there are a large number of inputs in complex nuclear reactor coupling systems, and there are high-order interactions between the inputs. The interactions of these inputs pose a huge challenge to building high-precision neural networks.

[0032] In order to achieve the goal of high-precision simulation output, if it is possible to identify in advance the important inputs that have a greater impact on the output, and then build an alternative model relationship between the input and output based on the important inputs, it will be of great benefit to improving the accuracy of the neural network.

[0033] Based on this, the present application proposes a method for constructing a reactor safety analysis model, taking the target parameters related to the transient operating conditions of the reactor as the target output, constructing a phenomenon recognition sorting table to first identify multiple input parameters, and using moment independent sensitivity measurement to screen important parameters that have an important impact on the target output from the input parameters, and constructing a target input based on the important parameters, so as to train a neural network model based on the target input and target output, and obtain a reactor safety analysis model that can accurately predict the transient operating conditions of nuclear reactor coupling system accidents, thereby improving the prediction accuracy of the reactor safety analysis model.

[0034] The present application is described in detail below through specific embodiments.

[0035] Figure 1 The flowchart of the method for constructing a reactor safety analysis model proposed in one embodiment of the present application is shown. Figure 1As shown, the method of this embodiment includes steps S110 to S150:

[0036] Step S110, obtaining the transient operating condition of the reactor, determining the target parameters related to the transient operating condition, and outputting the target parameters as the target.

[0037] Obtain the transient operating conditions of the nuclear reactor. The operating conditions are not specifically limited here, and theoretically all transient operating conditions are included. Each operating condition has its specific corresponding target parameters. The target parameters can be used to indicate the objects of concern and safety acceptance criteria in the transient operating conditions.

[0038] The safety of nuclear reactors depends most on ensuring the integrity of fuel elements. Therefore, the objects of concern may include, but are not limited to, fuel pellets, fuel cladding, etc. The damage of nuclear reactor fuel elements is mainly attributed to temperature changes. Therefore, the safety acceptance criteria may include, but are not limited to, temperature criteria, such as the material temperature limit of fuel pellets or fuel cladding, etc.

[0039] Different target parameters can be determined for different transient conditions. The target parameters are used as target outputs. The target outputs are used as part of the samples used to train the neural network model.

[0040] Step S120, constructing a phenomenon recognition ranking table related to the target parameter, and identifying multiple input parameters corresponding to the target parameter according to the phenomenon recognition ranking table.

[0041] Analyze and construct a phenomenon identification ranking table related to the target parameters. Phenomenon identification can refer to a comprehensive evaluation from physical phenomena to macroscopic systems, from the inside of the fuel heat source to the outside of the thermal system, etc., based on the design characteristics and operating condition mitigation characteristics of the nuclear reactor, so as to determine the phenomena that have an impact on the target parameters related to the transient operating conditions of the nuclear reactor.

[0042] Since different phenomena may have different importance at different stages, in some scenarios, the target parameters can also be refined according to the time stage. For example, in a nuclear reactor coolant loss accident, the time period can be divided into: the initial rapid coolant loss stage; the high-temperature heating stage of the subsystem before the use of emergency coolant; the initial stage of introducing emergency coolant into the high-temperature subsystem; the stage when the system is filled with emergency coolant, etc. In other scenarios, it can also be refined spatially to better identify the phenomenon.

[0043] For each target parameter, construct a phenomenon identification ranking table (PIRT) for each of the identified phenomena. The phenomena can be listed in any order in the table and can be grouped into general phenomena or process categories. For example, in a heat transfer problem, the phenomena can be listed as conduction, convection, radiation, etc.

[0044] The importance level of phenomena can be analyzed from different dimensions, and the determined phenomena can be classified and then labeled according to the classification. For example, the phenomenon classification can be carried out based on the existing research results on phenomena related to transient conditions of nuclear reactors; each phenomenon can be quantitatively analyzed by program or calculation, so as to evaluate the contribution of each phenomenon to the target parameter in the form of proportion; the importance level of each phenomenon can be determined by experiment, so as to classify the phenomenon; and the phenomena can be classified based on the scores of professionals.

[0045] Generally speaking, phenomena can be divided into three levels, marked as very important, generally important, and not important. Alternatively, phenomena can be divided into five levels, marked as extremely important, relatively important, generally important, not important, and extremely not important. Based on the classification of phenomena, phenomena marked as very important and generally important (or marked as extremely important, relatively important, and generally important) can be screened out as input parameters.

[0046] Step S130: sorting the importance of the multiple input parameters using a moment-independent global sensitivity analysis method, and selecting multiple important parameters from the multiple input parameters according to the importance sorting.

[0047] The moment-independent global sensitivity analysis method aims to evaluate the impact of input on the output probability density function. By calculating the moment-independent global sensitivity index of each input parameter, the input parameters are sorted according to their importance, and the leading input parameters with higher importance are selected as important parameters.

[0048] For example, based on the obtained importance ranking of the input parameters, the top input parameters with an importance accounting for 95% are determined, that is, the impact of the top input parameters on the output accounts for 95% of the total impact. The selected input parameters are taken as important parameters.

[0049] Step S140, performing multiple random sampling on multiple important parameters, and constructing a target input corresponding to each target parameter based on the results of the multiple random samplings.

[0050] Based on the determined important parameters, random sampling can be performed from the confidence intervals of these important parameters to form target inputs for training the neural network model. The target inputs serve as another part of the samples for training the neural network model.

[0051] Step S150, constructing a sample set according to the target input and the target output, and using the sample set to train the neural network model to obtain a reactor safety analysis model.

[0052] The target input is coupled with the corresponding target output to form a sample set for training the neural network model. The neural network model is trained using the sample set to obtain a reactor safety analysis model. The trained reactor safety analysis model characterizes the substitution function relationship between the target input and the target output, which can improve the accuracy of the prediction of the transient operating conditions of nuclear reactor accidents.

[0053] from Figure 1 It can be seen from the method shown that the construction method of the reactor safety analysis model provided by the present application first obtains the transient operating conditions of the reactor, determines the target parameters related to the transient operating conditions, and uses the target parameters as the target output; then constructs a phenomenon recognition ranking table related to the target parameters, and identifies multiple input parameters corresponding to the target parameters according to the phenomenon recognition ranking table; then uses the moment independent global sensitivity analysis method to sort the importance of multiple input parameters, and selects multiple important parameters from multiple input parameters according to the importance sorting; then performs multiple random sampling on multiple important parameters, and constructs the target input corresponding to the target parameter based on the results of multiple random sampling; finally, constructs a sample set based on the target input and the target output, and uses the sample set to train the neural network model to obtain the reactor safety analysis model. The construction method of the reactor safety analysis model provided by the present application determines the transient operating conditions and target parameters of the reactor, establishes PIRT to determine the input parameters, quantifies the uncertainty distribution of the input parameters and performs moment independent global sensitivity analysis, and selects the previous input parameters with a higher importance as the target input, thereby constructing a sample set training neural network model based on the target input and the target output. The method for constructing a reactor safety analysis model provided in the present application can realize the construction of a high-precision transient safety analysis alternative model for a nuclear reactor multi-field coupling system. The constructed reactor safety analysis model can obtain higher accuracy with a smaller sample, so as to be used for nuclear reactor fault diagnosis and accident prediction. The reactor safety analysis model constructed by the present application can effectively guide the operator to control the process of the reactor accident, and at the same time provide predictive protective measures for the safe operation of the reactor. Due to its rapid calculation, the neural network can predict the process of nuclear reactor accidents in real time, and buy more reaction time for the operator after the accident occurs.

[0054] In some embodiments of the present application, in the above method, step S110, obtaining the transient operating condition of the reactor, determining the target parameters related to the transient operating condition, and outputting the target parameters as the target, includes: obtaining the transient operating condition according to the reactor accident condition classification method; determining the objects of concern and the safety acceptance criteria related to the transient operating condition; and outputting the objects of concern and the safety acceptance criteria as the target.

[0055] In this embodiment, a full range of accident lists can be constructed based on the design features of the nuclear reactor, combined with the operation mode of the nuclear reactor, and with reference to the classification method of nuclear reactor accident conditions, so as to obtain transient conditions.

[0056] Determine the objects of concern and safety acceptance criteria related to transient conditions in nuclear reactors. For example, the operating condition of a nuclear reactor may be a nuclear reactor loss of coolant accident. In a nuclear reactor loss of coolant accident, the temperature of the fuel cladding will rise rapidly, and there is a risk of core damage. Therefore, the object of concern related to a nuclear reactor loss of coolant accident is the cladding peak temperature, and the safety acceptance criterion related to a nuclear reactor loss of coolant accident is not exceeding a specific limit (the specific limit may be 1204°C). In other words, obtain the transient condition of "nuclear reactor loss of coolant accident", and determine that the target parameter related to the transient condition is "the cladding peak temperature does not exceed a specific limit".

[0057] The objects of interest as well as the safety acceptance criteria are output as targets as part of the samples used to train the neural network model.

[0058] In some embodiments of the present application, in the above method, step S120, constructs a phenomenon identification ranking table related to the target parameters, and identifies multiple input parameters corresponding to the target parameters according to the phenomenon identification ranking table, including: determining multiple phenomena related to the target parameters according to reactor design characteristics and operating condition mitigation characteristics; grading multiple phenomena to determine the importance level of each phenomenon, and forming a phenomenon identification ranking table; and determining multiple input parameters that have a significant impact on the target parameters according to the phenomenon levels.

[0059] Phenomenon identification is the premise and basis for forming the phenomenon identification ranking table. Phenomena that have an impact on target parameters related to transient conditions of nuclear reactors can be determined based on the design characteristics of nuclear reactors and the characteristics of operating condition mitigation. When determining phenomena, they can be identified and determined based on experiments, expert opinions, inverse uncertainty analysis, engineering judgment, etc. The identified phenomena should be as comprehensive as possible, that is, discovery rather than evaluation is more important in the phenomenon identification stage.

[0060] For all reasonable phenomena that have an impact on the target parameters, a phenomenon identification ranking table is established for each transient operating condition, and the level of each phenomenon is marked in the phenomenon identification ranking table. The process of marking the level can be carried out through various methods such as experimental research, proportional analysis, and empirical judgment. The degree of influence of the phenomenon on the target parameters is analyzed from different dimensions, thereby providing strong data support for the phenomenon classification.

[0061] After the phenomenon identification ranking table is formed, multiple input parameters that have a significant impact on the target parameters are screened out based on the phenomenon level. Generally speaking, the phenomena marked as unimportant in the phenomenon identification ranking table can be eliminated, and the retained phenomena can be used as input parameters. For example, for the cladding peak temperature of a nuclear reactor loss of coolant accident not exceeding a specific limit, the determined input parameters may include critical flow, initial pressure, fuel specific heat, etc.

[0062] In some embodiments of the present application, in the above method, step S130, uses a moment independent global sensitivity analysis method to rank the importance of multiple input parameters, and screens out multiple important parameters from the multiple input parameters according to the importance ranking, including: fusing the results of program simulation calculation and the results of separation effect test measurement to determine the uncertainty distribution type and confidence interval of each input parameter; performing a moment independent global sensitivity analysis on each input parameter according to the uncertainty distribution type and the confidence interval, and determining the importance ranking of each input parameter according to the results of the moment independent global sensitivity analysis; and screening out multiple important parameters from the multiple input parameters according to the importance ranking.

[0063] First, the uncertainty distribution of the input parameters can be quantified by integrating the results of the program simulation calculation with the results of the separation effect experimental measurement, thereby determining the uncertainty distribution type and confidence interval of the input parameters. The uncertainty distribution type can include but is not limited to normal distribution, uniform distribution, etc., and the confidence interval can include but is not limited to selecting a 95% confidence level or a 90% confidence level, etc.

[0064] For example, take the critical flow input parameters related to the peak temperature of the cladding in a nuclear reactor loss of coolant accident not exceeding a specific limit as an example. The result of the separation effect experiment measurement is the result obtained by measuring the flow rate sprayed at the rupture in the separation effect experiment related to the critical flow. The result of the program simulation calculation is the result of the simulation calculation using the module in the program for calculating the critical flow under the same conditions as the separation effect experiment. The program simulation calculation can use the simulation calculation method of the prior art, which is not specifically limited here. By integrating the results of the simulation calculation with the results of the experimental measurement, the uncertainty distribution of the quantified critical flow obeys the standard normal distribution N(0,1). If the confidence interval is determined according to the 95% confidence level, its confidence interval is [-1.96,1.96].

[0065] Then, a moment-independent global sensitivity analysis is performed on each input parameter according to the uncertainty distribution type and confidence interval, and the importance ranking of each input parameter is determined according to the results of the moment-independent global sensitivity analysis.

[0066] Specifically, multiple random samplings are performed on the input parameters within the confidence interval; the calculation results of the target parameters are determined based on the multiple random samplings; the moment independent global sensitivity index of the input parameters is determined according to the calculation results of the target parameters; and the input parameters are sorted according to importance from large to small based on the moment independent global sensitivity index.

[0067] Based on the uncertainty distribution type and confidence interval of the input parameters, random sampling is performed within the confidence interval of the input parameters and calculations are performed. The number of random samplings can be as large as possible to ensure the accuracy of the calculation results; for example, assuming that there are 20 input parameters, there are 20 uncertainty distribution types and confidence intervals corresponding to the input parameters. Randomly select one from the confidence interval corresponding to each of the 20 input parameters, combine them, and perform calculations to determine the calculation results of the target parameters. Preferably, the above process is repeated no less than 1000 times.

[0068] According to the calculation results of the target parameters, the moment-independent sensitivity index of each input parameter is determined using the moment-independent global sensitivity analysis method.

[0069] The principle of the moment-independent global sensitivity analysis method is as follows:

[0070] Assume that the function Y = g(X) has k inputs, that is, X = (X1, X2…, X k ) T , each input obeys a certain probability distribution, then the uncertainty of the input can be propagated to the output through function calculation, and the distribution of the unconditional output is obtained. The unconditional probability density function (PDF) and unconditional cumulative distribution function CDF of Y are represented as f Y (y) and F Y (y), the i-th input X i The conditional PDF and conditional CDF of Y obtained when taking a certain fixed value are expressed as and See the schematic diagram of the output unconditional PDF and conditional PDF. Figure 2 As shown, the schematic diagram of the output unconditional CDF and conditional CDF is shown in Figure 3 shown.

[0071] Reference Figure 2 and Figure 3 As shown in Figure 1, fixing the value of an input will cause the output conditional PDF to change compared to the output unconditional PDF / the output conditional CDF to change compared to the output unconditional CDF. The goal of moment-independent global sensitivity analysis is to evaluate the offset between the output conditional PDF and the output unconditional PDF / the offset between the output conditional CDF and the output unconditional CDF. That is, the area between the conditional function and the unconditional function.

[0072] Define the offset of the output conditional PDF for a fixed input condition as s(X i ), the offset of the output conditional CDF is A(X i ),but

[0073]

[0074]

[0075] That is, s(X i )for Figure 2 The sum of the areas of s1, s2, and s3 in the figure is used to represent the offset of the conditional PDF output after fixing the i-th input, A(X i )for Figure 3 The sum of the areas of parts A1 and A2.

[0076] Since each input follows a certain distribution, we generally consider the input X i The expected output PDF and output CDF shifts when varying in their distribution.

[0077]

[0078] Among them, x i Represents input X i The value in the distribution.

[0079] Based on this, two moment-independent global sensitivity indices of the i-th parameter can be obtained.

[0080]

[0081] Where E(Y) represents the first-order origin moment of the output value. When E(Y) = 0, use (Formula 4) to calculate

[0082]

[0083] calculate and The difficulty lies in solving s(X i ) and A(X i ). For solving s(X i ) can be used to solve s1, s2 and s3 respectively. s1, s2 and s3 are solved according to the following relationship:

[0084]

[0085] Among them, a and b represent Figure 2 The intersection of the output unconditional PDF and the output conditional PDF.

[0086] It can be seen that the key point is to determine fY (y) and The intersection point. Y (y) and The intersection point can be expressed as:

[0087]

[0088] Furthermore, (Formula 10) can be expressed as:

[0089]

[0090] Since the output F Y (y) and can be calculated, so it can be determined by The extreme point calculation s(X i ).

[0091] Assume f Y (y) and There are m intersection points, denoted as a1, a2, ..., a m , then s(X i ) can be expressed as the sum of (m+1) sub-areas, that is:

[0092] s(X i )=s1+s2+…+s j …+s m +s m+1 , (j=1,2,…,m+1), (Formula

[0093] 12).

[0094] Each sub-area s j The calculation is based on the following relationship:

[0095]

[0096] Based on (Formula 11) to (Formula 13), we can calculate s(X i ), thereby calculating the moment-independent global sensitivity index of each input

[0097] Based on (Formula 2), we can calculate A(X i ), thereby calculating the moment-independent global sensitivity index of each input

[0098] Based on the moment-independent global sensitivity index, the input parameters are sorted by importance from large to small. Multiple important parameters are screened out from multiple input parameters according to the importance sorting. The top input parameters with an importance of 95% can be determined, that is, the impact of the top input parameters on the output accounts for 95% of the total impact. The screened input parameters are taken as important parameters.

[0099] In some embodiments of the present application, in the above method, step S140, performs multiple random sampling on multiple important parameters, and constructs a target input corresponding to the target parameter based on the results of the multiple random samplings, including: performing multiple random sampling on the important parameters within the confidence interval; the result of each random sampling constitutes an input sample, and the results of multiple random samplings constitute the target input.

[0100] After determining the important parameters, the target input is constructed based on the important parameters. The target input is a set that includes multiple input samples.

[0101] When constructing the target input, randomly sample within the confidence interval of the important parameters. For example, assuming there are 10 important parameters, randomly select one from the confidence interval corresponding to each of the 10 important parameters, and combine them to obtain an input sample. Repeat the above process to obtain multiple input samples to form a set of target inputs.

[0102] In some embodiments of the present application, in the above method, step S150, constructing a sample set according to the target input and the target output, and using the sample set to train the neural network model to obtain a reactor safety analysis model, includes: constructing a neural network model; constructing a sample set according to the target input and the target output, and dividing the sample set into a training set, a test set and a verification set; training the neural network model with the training set, verifying the trained neural network model with the verification set, and testing the trained neural network model with the test set to obtain the trained neural network model as the reactor safety analysis model.

[0103] The sample set is divided into a training set, a test set, and a validation set, and a neural network model is constructed. The neural network model is trained using the training set; the performance indicators of the neural network model are evaluated based on the validation set to understand the performance of the model on the validation set, and the model parameters are adjusted or the model is fine-tuned according to the performance of the model; the performance and generalization ability of the neural network model are tested based on the test set. When the model performance meets the requirements, the trained neural network model is used as a reactor safety analysis model to predict accident transient conditions and fault diagnosis of nuclear reactor coupling systems.

[0104] Figure 4 A schematic diagram showing a process flow of a method for constructing a reactor safety analysis model proposed in another embodiment of the present application. Figure 4 As shown, the method of this embodiment includes steps S401 to S410:

[0105] Step S401, obtain transient conditions according to the reactor accident condition classification method, determine the objects of concern and safety acceptance criteria related to the transient conditions, and output the objects of concern and the corresponding safety acceptance criteria as targets; go to step S402 and step S407.

[0106] Step S402, multiple phenomena related to the target parameters are determined according to the reactor design characteristics and the operating condition mitigation characteristics, the multiple phenomena are graded to determine the importance level of each phenomenon to form a phenomenon identification ranking table, and multiple input parameters having a significant impact on the target parameters are determined according to the phenomenon level. Go to step S403.

[0107] Step S403, the results of program simulation calculation and the results of separation effect test measurement are integrated to determine the uncertainty distribution type and confidence interval of each input parameter. Then, the process goes to step S404.

[0108] Step S404, perform multiple random sampling on the input parameters within the confidence interval, determine the calculation results of the target parameters based on the multiple random sampling, determine the moment independent global sensitivity index of the input parameters according to the calculation results of the target parameters, and sort the input parameters according to importance from large to small based on the moment independent global sensitivity index. Go to step S405.

[0109] Step S405, filter out multiple important parameters from multiple output parameters according to the importance ranking. Go to step S406.

[0110] Step S406, perform multiple random sampling on important parameters within the confidence interval, the result of each random sampling constitutes an input sample, and the results of multiple random samplings constitute the target input. Go to step S407.

[0111] Step S407, construct a neural network, construct a sample set according to the target input and the target output and divide the sample set into a training set, a test set and a verification set, use the training set to train the neural network model, use the verification set to verify the trained neural network model, use the test set to test the trained neural network model, and obtain the trained neural network model as the reactor safety analysis model.

[0112] Figure 5 A schematic diagram of the structure of a device for constructing a reactor safety analysis model according to an embodiment of the present application is shown. Figure 5 As shown, the device 500 includes:

[0113] The target output determination unit 510 is used to obtain the transient operating conditions of the reactor, determine the target parameters related to the transient operating conditions, and use the target parameters as the target output;

[0114] An input parameter determination unit 520 is used to construct a phenomenon recognition ranking table related to the target parameter, and identify multiple input parameters corresponding to the target parameter according to the phenomenon recognition ranking table;

[0115] An important parameter determination unit 530 is used to sort the importance of multiple input parameters using a moment-independent global sensitivity analysis method, and select multiple important parameters from the multiple input parameters according to the importance sorting;

[0116] A target input determination unit 540 is used to perform multiple random samplings on multiple important parameters and construct a target input corresponding to the target parameters based on the results of the multiple random samplings;

[0117] The model training unit 550 is used to construct a sample set according to the target input and the target output, and use the sample set to train the neural network model to obtain a reactor safety analysis model.

[0118] In some embodiments of the present application, in the above-mentioned device 500, the target output determination unit 510 is specifically used to: obtain transient conditions according to the reactor accident condition classification method; determine the objects of concern and safety acceptance criteria related to the transient conditions; and output the objects of concern and the corresponding safety acceptance criteria as target outputs.

[0119] In some embodiments of the present application, in the above-mentioned device 500, the input parameter determination unit 520 is specifically used to: determine multiple phenomena related to the target parameters based on reactor design characteristics and operating condition mitigation characteristics; classify multiple phenomena to determine the importance level of each phenomenon, and form a phenomenon identification ranking table; determine multiple input parameters that have an important impact on the target parameters based on the phenomenon level.

[0120] In some embodiments of the present application, in the above-mentioned device 500, the important parameter determination unit 530 is specifically used to: determine the uncertainty distribution type and confidence interval of each input parameter by integrating the results of program simulation calculation and the results of separation effect experimental measurement; perform moment independent global sensitivity analysis on each input parameter according to the uncertainty distribution type and confidence interval, and determine the importance ranking of each input parameter according to the results of the moment independent global sensitivity analysis; and screen out multiple important parameters from multiple input parameters according to the importance ranking.

[0121] In some embodiments of the present application, in the above-mentioned device 500, the important parameter determination unit 530 is also specifically used to: perform multiple random sampling on the input parameters within the confidence interval; determine the calculation results of the target parameters based on the multiple random samplings; determine the moment independent global sensitivity index of the input parameters according to the calculation results of the target parameters; and sort the input parameters according to importance from large to small based on the moment independent global sensitivity index.

[0122] In some embodiments of the present application, in the above-mentioned device 500, the target input determination unit 540 is specifically used to: perform multiple random samplings on important parameters within the confidence interval; the result of each random sampling constitutes an input sample, and the results of multiple random samplings constitute the target input.

[0123] In some embodiments of the present application, in the above-mentioned device 500, the model training unit 550 is specifically used to: construct a neural network model; construct a sample set according to the target input and the target output and divide the sample set into a training set, a test set and a verification set; train the neural network model using the training set, verify the trained neural network model using the verification set, and test the trained neural network model using the test set to obtain the trained neural network model as a reactor safety analysis model.

[0124] It should be noted that the above-mentioned reactor safety analysis model construction device 500 can implement the above-mentioned reactor safety analysis model construction method one by one, which will not be described in detail.

[0125] Figure 6 A schematic diagram showing the structure of a computer device according to an embodiment of the present application is shown. Figure 6 As shown, the internal structure of the computer device may include a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, the functions or steps of the method for constructing a reactor safety analysis model are realized.

[0126] In one embodiment, the computer device provided by the present application includes a memory and a processor, the memory stores a database and a computer program that can be run on the processor, and the processor implements the following steps when executing the computer program:

[0127] Acquire the transient operating condition of the reactor, determine the target parameters related to the transient operating condition, and use the target parameters as target output;

[0128] Constructing a phenomenon recognition ranking table related to the target parameter, and identifying multiple input parameters corresponding to the target parameter according to the phenomenon recognition ranking table;

[0129] The moment-independent global sensitivity analysis method is used to rank the importance of multiple input parameters, and multiple important parameters are selected from the multiple input parameters according to the importance ranking;

[0130] Perform multiple random sampling on multiple important parameters, and construct target inputs corresponding to target parameters based on the results of multiple random samplings;

[0131] A sample set is constructed according to the target input and target output, and the neural network model is trained using the sample set to obtain a reactor safety analysis model.

[0132] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0133] Acquire the transient operating condition of the reactor, determine the target parameters related to the transient operating condition, and use the target parameters as target output;

[0134] Constructing a phenomenon recognition ranking table related to the target parameter, and identifying multiple input parameters corresponding to the target parameter according to the phenomenon recognition ranking table;

[0135] The moment-independent global sensitivity analysis method is used to rank the importance of multiple input parameters, and multiple important parameters are selected from the multiple input parameters according to the importance ranking;

[0136] Perform multiple random sampling on multiple important parameters, and construct target inputs corresponding to target parameters based on the results of multiple random samplings;

[0137] A sample set is constructed according to the target input and target output, and the neural network model is trained using the sample set to obtain a reactor safety analysis model.

[0138] It should be noted that the above functions or steps that can be implemented by the computer device or computer-readable storage medium can refer to the relevant description in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0139] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0140] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0141] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for constructing a reactor safety analysis model, characterized in that: The method comprises: Acquiring a transient operating condition of the reactor, determining a target parameter related to the transient operating condition, and using the target parameter as a target output; Constructing a phenomenon recognition ranking table related to the target parameter, and identifying a plurality of input parameters corresponding to the target parameter according to the phenomenon recognition ranking table; Using a moment-independent global sensitivity analysis method to sort the importance of the plurality of input parameters, and selecting a plurality of important parameters from the plurality of input parameters according to the importance sorting; Performing multiple random sampling on the multiple important parameters, and constructing a target input corresponding to the target parameter based on the results of the multiple random samplings; A sample set is constructed according to the target input and the target output, and a neural network model is trained using the sample set to obtain a reactor safety analysis model.

2. The method for constructing a reactor safety analysis model according to claim 1, characterized in that: The step of obtaining a transient operating condition of the reactor, determining a target parameter related to the transient operating condition, and outputting the target parameter as a target comprises: According to a reactor accident condition classification method, obtaining the transient condition; Determine the objects of concern and safety acceptance criteria associated with the transient conditions; The object of interest and the corresponding security acceptance criteria are output as the target.

3. The method for constructing a reactor safety analysis model according to claim 1, characterized in that: The step of constructing a phenomenon recognition ranking table related to the target parameter and identifying a plurality of input parameters corresponding to the target parameter according to the phenomenon recognition ranking table includes: determining a plurality of phenomena related to the target parameter based on reactor design characteristics and operating condition mitigation characteristics; Classifying the plurality of phenomena to determine the importance level of each phenomenon, and forming a phenomenon identification ranking table; According to the phenomenon level, a plurality of the input parameters having a significant influence on the target parameter are determined.

4. The method for constructing a reactor safety analysis model according to claim 1, characterized in that: The step of using the moment-independent global sensitivity analysis method to sort the plurality of input parameters by importance, and selecting a plurality of important parameters from the plurality of input parameters according to the importance sorting, comprises: The uncertainty distribution type and confidence interval of each input parameter are determined by integrating the results of program simulation calculation and the results of separation effect experimental measurement; Performing a moment-independent global sensitivity analysis on each of the input parameters according to the uncertainty distribution type and the confidence interval, and determining the importance ranking of each of the input parameters according to the result of the moment-independent global sensitivity analysis; A plurality of the important parameters are screened out from the plurality of the input parameters according to the importance ranking.

5. The method for constructing a reactor safety analysis model according to claim 4, characterized in that: The step of performing moment-independent global sensitivity analysis on each of the input parameters according to the uncertainty distribution type and the confidence interval, and determining the importance ranking of each of the input parameters according to the result of the moment-independent global sensitivity analysis, comprises: Performing multiple random sampling on the input parameter within the confidence interval; Determine the calculation result of the target parameter based on the multiple random samplings; Determining a moment-independent sensitivity index of the input parameter according to a calculation result of the target parameter; The input parameters are sorted according to importance from large to small based on the moment-independent sensitivity index.

6. The method for constructing a reactor safety analysis model according to claim 4, characterized in that: The step of performing multiple random sampling on the multiple important parameters and constructing a target input corresponding to the target parameter based on the results of the multiple random samplings includes: Performing multiple random sampling on the important parameters within the confidence interval; The result of each random sampling constitutes an input sample, and the results of multiple random samplings constitute the target input.

7. The method for constructing a reactor safety analysis model according to claim 1, characterized in that: The step of constructing a sample set according to the target input and the target output, and using the sample set to train a neural network model to obtain a reactor safety analysis model includes: Build a neural network model; Constructing a sample set according to the target input and the target output, and dividing the sample set into a training set, a test set, and a validation set; The neural network model is trained using the training set, the trained neural network model is verified using the verification set, and the trained neural network model is tested using the test set to obtain the trained neural network model as the reactor safety analysis model.

8. A device for constructing a reactor safety analysis model, characterized in that: The device comprises: A target output determination unit, used to obtain a transient operating condition of the reactor, determine a target parameter related to the transient operating condition, and use the target parameter as a target output; An input parameter determination unit, configured to construct a phenomenon recognition ranking table related to the target parameter, and to identify a plurality of input parameters corresponding to the target parameter according to the phenomenon recognition ranking table; An important parameter determination unit, used for ranking the plurality of input parameters by importance using a moment-independent global sensitivity analysis method, and selecting a plurality of important parameters from the plurality of input parameters according to the importance ranking; a target input determination unit, configured to perform multiple random sampling on the multiple important parameters, and construct a target input corresponding to the target parameters based on the results of the multiple random samplings; The model training unit is used to construct a sample set according to the target input and the target output, and use the sample set to train the neural network model to obtain a reactor safety analysis model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for constructing a reactor safety analysis model as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: The computer program, when instructed by a processor, implements the steps of the method for constructing a reactor safety analysis model as described in any one of claims 1 to 7.

Images