Calibration method and system for nuclear work software model input parameters and electronic equipment

Abstract

The invention discloses a nuclear work software model input parameter calibration method and system and electronic equipment, and the method comprises the steps: generating a training data set and a verification data set according to an experiment data set related to a to-be-calibrated model; constructing a replacement large model corresponding to the to-be-calibrated model; training the replacement large model through the training data set to obtain a trained replacement large model; determining target posterior probability distribution of input parameters of the trained replacement large model according to the predicted value set and the data set for verification of the trained replacement large model; preset prior probability distribution of the to-be-calibrated model is replaced with the target posterior probability distribution, and the preset prior probability distribution is preset probability distribution corresponding to input parameters of the to-be-calibrated model. According to the method, the accuracy of the input parameters of the software model can be improved, and the uncertainty can be statistically quantified.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, system and electronic equipment for calibrating input parameters of a nuclear engineering software model. Background Art

[0002] With the advancement of computing technology, computer numerical simulation has been widely used in industrial practice. Large-scale industrial software with complex mathematical and physical models has been successfully applied in the design and evaluation work of various scientific research and industrial fields.

[0003] For nuclear engineering design and safety analysis software, especially best estimate software, safety review agencies in various countries have put forward the requirement of uncertainty quantification and sensitivity analysis when such software is applied to best estimate evaluation models or combined evaluation models, as one of the important components of software verification and validation.

[0004] In related technologies, in order to obtain uncertainty quantification results, forward uncertainty quantification analysis is usually performed using software. Forward uncertainty quantification requires sufficient knowledge of input uncertainty. However, in practical applications, the determination of many input parameters and their uncertainties comes from expert judgment or user experience, etc., lacking reasonable physical or mathematical basis support. In addition, due to the different cognition of experts and users, huge differences in calculation results may occur, that is, forward uncertainty quantification will lead to inaccurate input parameters of the software model. Summary of the invention

[0005] The present application aims to propose a method, system and electronic equipment for calibrating the input parameters of a nuclear engineering software model, which can improve the accuracy of the input parameters of the software model.

[0006] In a first aspect, an embodiment of the present application provides a method for calibrating input parameters of a nuclear engineering software model, the method comprising:

[0007] Generate a training data set and a validation data set based on an experimental data set related to the model to be calibrated;

[0008] Constructing a surrogate large model corresponding to the model to be calibrated;

[0009] Train the substitute large model by using the training data set to obtain a trained substitute large model;

[0010] Determine the target posterior probability distribution of the input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set;

[0011] The preset prior probability distribution of the model to be calibrated is replaced by the target posterior probability distribution, wherein the preset prior probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

[0012] In some embodiments, constructing a surrogate macromodel corresponding to the model to be calibrated includes:

[0013] Get the number of data in the experimental data set, N is a positive integer;

[0014] When N is less than a preset value, a linear deep Gaussian process structure is selected to construct an alternative large model.

[0015] In some embodiments, after obtaining the number N of experimental data, the method further comprises:

[0016] When N is greater than or equal to a preset value, the relationship between the input parameters and the output parameters of the model to be calibrated is analyzed to obtain a linear analysis result, wherein the linear analysis result is a linear correlation or a nonlinear correlation;

[0017] If the linear analysis results in nonlinear correlation, a nonlinear deep Gaussian process structure is selected to construct an alternative large model;

[0018] If the linear analysis result is linearly correlated, a linear deep Gaussian process structure is selected to construct an alternative large model.

[0019] In some embodiments, after training the surrogate large model with the training data set to obtain the trained surrogate large model, before determining the target posterior probability distribution of the input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set, the method further includes:

[0020] Calculate the mean square error of the trained surrogate model;

[0021] According to the predicted value set of the trained surrogate model and the validation data set, the target posterior probability distribution of the input parameters of the trained surrogate model is determined, including:

[0022] If the mean square error value meets the preset threshold range, the input data in the verification data set is input into the trained alternative large model to obtain the predicted value set of the trained alternative large model;

[0023] The target posterior probability distribution of the input parameters of the trained surrogate large model is determined according to the predicted values ​​in the predicted value set of the trained surrogate large model and the output true values ​​corresponding to the input data in the verification data set.

[0024] In some embodiments, determining a target posterior probability distribution of input parameters of the trained surrogate macromodel based on the predicted values ​​in the predicted value set of the trained surrogate macromodel and the true output values ​​corresponding to the input data in the validation data set comprises:

[0025] Obtaining the prior probability distribution of the model to be calibrated;

[0026] Compare the predicted values ​​in the predicted value set of the trained alternative large model with the output true values ​​in the validation data set to obtain a comparison result;

[0027] Based on the comparison results and the prior probability distribution, the maximum likelihood estimation method is used to calculate the posterior probability distribution of the input parameters of the trained surrogate large model;

[0028] If the maximum likelihood estimation method meets the preset acceptance criteria, the posterior probability distribution is determined as the target posterior probability distribution;

[0029] If the maximum likelihood estimation method does not meet the preset acceptance criteria, the posterior probability distribution is used to replace the prior probability distribution, and the execution is returned to calculate the posterior probability distribution of the input parameters of the trained replacement large model using the maximum likelihood estimation method based on the comparison result and the prior probability distribution, until the maximum likelihood estimation method meets the preset acceptance criteria, and the posterior probability distribution calculated after the maximum likelihood estimation method meets the preset acceptance criteria is determined as the target posterior probability distribution.

[0030] In some embodiments, generating a training dataset and a validation dataset based on an experimental dataset related to a model to be calibrated includes:

[0031] Obtain the number of experimental data M used for training the model to be calibrated, where M is a positive integer;

[0032] Select M data from the experimental data set;

[0033] Generate a training data set including M data.

[0034] In some embodiments, generating a training dataset and a validation dataset based on an experimental dataset related to a model to be calibrated includes:

[0035] Remove the training data set from the experimental data set to obtain the remaining experimental data set;

[0036] generating a plurality of subsets using at least a portion of the data in the remaining experimental data set;

[0037] Selecting a preset number of first data from each subset;

[0038] The first data selected from each subset is combined into a verification data set.

[0039] In some embodiments, multiple subsets are generated using at least a portion of the remaining experimental data set, including:

[0040] Determine the input variables of the model to be calibrated;

[0041] Acquire a target range, where the target range includes at least one of a range of an input variable and a range of an input parameter;

[0042] If the target range is within the range corresponding to the remaining experimental data sets, multiple subsets are generated using the data in the remaining experimental data sets that are within the target range;

[0043] If the target range is larger than the range corresponding to the remaining experimental data sets, all the data of the remaining experimental data sets are used to generate multiple subsets.

[0044] In a second aspect, the embodiment of the present application further provides a calibration system for input parameters of a nuclear engineering software model, the system comprising:

[0045] A data set generation module, used to generate a training data set and a verification data set according to an experimental data set related to the model to be calibrated;

[0046] A large model building module, used to build a substitute large model corresponding to the model to be calibrated;

[0047] A large model training module is used to train the substitute large model through a training data set to obtain a trained substitute large model;

[0048] A probability distribution determination module, used to determine a target posterior probability distribution of input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set;

[0049] The input parameter calibration module is used to replace the preset prior probability distribution of the model to be calibrated by the target posterior probability distribution, wherein the preset prior probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

[0050] In a third aspect, an embodiment of the present application also provides an electronic device comprising at least one control processor and a memory for communicating with the at least one control processor; the memory stores instructions executable by the at least one control processor, and the instructions are executed by the at least one control processor so that the at least one control processor can execute a method for calibrating input parameters of a nuclear engineering software model as described in the first aspect.

[0051] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute a method for calibrating input parameters of a nuclear engineering software model as described in the first aspect.

[0052] In an embodiment of the present application, a training data set and a verification data set are generated according to an experimental data set related to the model to be calibrated; a replacement large model corresponding to the model to be calibrated is constructed; the replacement large model is trained with the training data set to obtain a trained replacement large model; the target posterior probability distribution of the input parameters of the trained replacement large model is determined according to the predicted value set and the verification data set of the trained replacement large model; the preset prior probability distribution of the model to be calibrated is replaced by the target posterior probability distribution, wherein the preset prior probability distribution is a probability distribution corresponding to the input parameters of the model to be calibrated. In this way, the replacement large model is trained with the existing experimental data, and then the preset prior probability distribution of the model to be calibrated is replaced by the target posterior probability distribution of the input parameters of the trained replacement large model, so that the uncertainty can be statistically quantified, which can solve the problem of the difference in the quantification results of the positive uncertainty caused by the human bias of the input uncertainty, thereby improving the accuracy of the input parameters of the software model. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0054] Figure 1 It is a flow chart of an embodiment of a method for calibrating input parameters of a nuclear engineering software model provided by the present application;

[0055] Figure 2 It is a flow chart of the best embodiment of the calibration method of the nuclear engineering software model input parameters provided by the present application;

[0056] Figure 3 It is a structural schematic diagram of an embodiment of a calibration system for nuclear engineering software model input parameters provided by the present application;

[0057] Figure 4 It is a schematic diagram of the structure of an embodiment of the electronic device provided by the present application. DETAILED DESCRIPTION

[0058] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0059] In the description of this application, if there is a description of first, second, etc., it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0060] In the description of the present application, it should be understood that the descriptions involving orientation, such as the orientation or positional relationship indicated as up, down, etc., are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0061] In the description of this application, it should be noted that, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meaning of the above terms in this application based on the specific content of the technical solution.

[0062] With the advancement of computing technology, computer numerical simulation has been widely used in industrial practice. Large-scale industrial software with complex mathematical and physical models has been successfully applied in the design and evaluation work of various scientific research and industrial fields.

[0063] In the field of nuclear engineering, software simulation is widely used in design and safety analysis. For nuclear engineering design and safety analysis software, especially best estimate software, safety review units in various countries have proposed the need for forward uncertainty quantification and sensitivity analysis when such software is applied to best estimate evaluation models or combined evaluation models, as one of the important components of software verification and validation. In related technologies, in order to obtain uncertainty quantification results, it is usually necessary to use software to perform forward uncertainty quantification analysis. And forward uncertainty quantification requires sufficient knowledge of input uncertainty. However, in practical applications, the determination of many input parameters and their uncertainties comes from expert judgment or user experience, etc., lacking reasonable physical or mathematical foundations, and due to the different cognition of experts and users, there may be huge differences in calculation results, that is, forward uncertainty quantification will lead to inaccurate input parameters of the software model.

[0064] In order to solve the problem that the calculation results may be greatly different due to the different cognitions of experts and users, the present application proposes a calibration method, system and electronic equipment for the input parameters of a nuclear engineering software model.

[0065] Reference Figure 1 , a flow chart of a method for calibrating input parameters of a nuclear engineering software model provided in an embodiment of the present application. The method for calibrating input parameters of a nuclear engineering software model is applied to an electronic device, which may be a server or a mobile terminal, etc. Figure 1 As shown, the calibration method of the nuclear engineering software model input parameters may include the following steps:

[0066] Step 110: Generate a training data set and a validation data set according to an experimental data set related to the model to be calibrated;

[0067] Step 120, constructing a replacement large model corresponding to the model to be calibrated;

[0068] Step 130: train the substitute large model using the training data set to obtain a trained substitute large model;

[0069] Step 140, determining a target posterior probability distribution of input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set;

[0070] Step 150: Replace a preset priori probability distribution of the model to be calibrated with a target posterior probability distribution, wherein the preset priori probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

[0071] In this embodiment, a training data set and a verification data set are generated according to an experimental data set related to the model to be calibrated; a substitute large model corresponding to the model to be calibrated is constructed; the substitute large model is trained with the training data set to obtain a trained substitute large model; the target posterior probability distribution of the input parameters of the trained substitute large model is determined according to the predicted value set and the verification data set of the trained substitute large model; the preset prior probability distribution of the model to be calibrated is replaced by the target posterior probability distribution, wherein the preset prior probability distribution is a probability distribution corresponding to the input parameters of the model to be calibrated. In this way, the substitute large model is trained with the existing experimental data, and then the preset prior probability distribution of the model to be calibrated is replaced by the target posterior probability distribution of the input parameters of the trained substitute large model, so that the uncertainty can be statistically quantified, and the problem of the difference in the quantification results of the positive uncertainty caused by the human bias of the input uncertainty can be solved, thereby improving the accuracy of the input parameters of the software model.

[0072] In the above step 110, the electronic device generates a training data set and a verification data set according to an experimental data set related to the model to be calibrated.

[0073] The above-mentioned model to be calibrated may be a nuclear engineering software model to be calibrated in nuclear engineering design and safety analysis software.

[0074] The above-mentioned model to be calibrated can be judged by the user based on the phenomenon identification ranking table (PIRT table), analyzing different accident conditions in the PIRT table, determining the key phenomenon related to the accident condition, and then making expert judgment to determine the model to be calibrated related to the key phenomenon.

[0075] The experimental data set related to the model to be calibrated may include a plurality of experimental data generated by separation effect experiments and overall effect experiments related to the model to be calibrated.

[0076] The above-mentioned separation effect experiment can be a separation effect experiment (SET) bench established based on the similar proportion law in the process of conducting safety assessment of nuclear reactors and nuclear power plants to provide data support for safety performance verification and evaluation.

[0077] The above-mentioned overall effect experiment can be an overall effect test (IET) bench established based on the similarity proportionality law in the process of safety assessment of nuclear reactors and nuclear power plants, providing data support for safety performance verification and evaluation. Overall effect experiment refers to a full-system simulation experiment of a thermal process containing multiple effects, which is used to observe the mutual influence and feedback relationship between various effects, and to verify, improve and perfect the computer program describing the process.

[0078] An example of an overall effect experiment: natural circulation experiment of the reactor primary loop, small and large rupture accident experiments of the reactor coolant system, and overall effect experiment of the entrainment and de-entrainment of droplets during core reflooding.

[0079] The above-mentioned generation of the training data set and the verification data set based on the experimental data set related to the model to be calibrated may be that a certain amount of data is selected from the experimental data set related to the model to be calibrated as the data in the training data set and the verification data set, and the amount of data in the training data set and the verification data set may be unequal.

[0080] Before generating the training data set and the validation data set, the separation effect experiment and the overall effect experiment corresponding to the experimental data set are first checked to see whether they cover the key phenomena related to the model to be calibrated. If they do, the experimental data set is available. If they do not, the experimental data set is unavailable. When the experimental data set is unavailable, it is necessary to reselect the experimental data set.

[0081] In the above step 120, a replacement large model corresponding to the model to be calibrated is constructed.

[0082] The embodiments of the present application do not limit the type or structure of the alternative large model. In some embodiments, the alternative large model corresponding to the model to be calibrated can be an alternative large model constructed using a deep Gaussian process structure. In other embodiments, an alternative large model constructed using an existing deep machine learning algorithm, such as a deep neural network structure, can also be used.

[0083] In the above step 130, the substitute large model is trained by using the training data set to obtain a trained substitute large model.

[0084] The embodiments of the present application do not limit the training method of the replacement large model, and any method that can be used to train the replacement large model can fall within the scope of protection of the present application. In some embodiments, the above-mentioned training of the replacement large model with the training data set to obtain the trained replacement large model can be to use the maximum likelihood estimation method to train the replacement large model with the training data set to obtain the trained replacement large model.

[0085] The above-mentioned maximum likelihood estimation method can be an important and common method for finding an estimator. The maximum likelihood estimation method explicitly uses a probability model, and its goal is to find a phylogenetic tree that can generate the observed data with a higher probability.

[0086] In the above step 140, the target posterior probability distribution of the input parameters of the trained surrogate large model is determined based on the prediction value set of the trained surrogate large model and the validation data set.

[0087] The predicted value set of the trained replacement large model includes at least one predicted value, and each predicted value is obtained by inputting an input data in the validation data set into the trained replacement large model for prediction.

[0088] The above-mentioned determination of the target posterior probability distribution of the input parameters of the trained alternative large model can be determined by using the maximum likelihood estimation method based on the prediction value set and the verification data set of the trained alternative large model.

[0089] The above-mentioned determination of the target posterior probability distribution of the input parameters of the trained alternative large model can also be carried out by using the maximum a posteriori estimation method to determine the target posterior probability distribution of the input parameters of the trained alternative large model based on the prediction value set and the verification data set of the trained alternative large model.

[0090] In the above step 150, the preset prior probability distribution of the model to be calibrated is replaced by the target posterior probability distribution, wherein the preset prior probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

[0091] The input parameters of the above-mentioned model to be calibrated can be determined by the user based on the PIRT table, analyzing different accident conditions, determining the key phenomena related to the accident conditions, and then making expert judgments to determine the input parameters of the model to be calibrated related to the key phenomena.

[0092] The above-mentioned replacement of the preset prior probability distribution of the model to be calibrated with the target posterior probability distribution can realize the calibration of the input parameters of the model to be calibrated.

[0093] Since the preset prior probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated, and the target posterior probability distribution is a determined and trained target posterior probability distribution that replaces the input parameters of the large model, the calibration of the input parameters of the model to be calibrated can be achieved by replacing the preset prior probability distribution with the target posterior probability distribution.

[0094] In some embodiments, constructing a replacement large model corresponding to the model to be calibrated may include:

[0095] Get the number of data in the experimental data set, N is a positive integer;

[0096] When N is less than a preset value, a linear deep Gaussian process structure is selected to construct an alternative large model.

[0097] In this embodiment, when the number of data N in the experimental data set is less than the preset value, a linear deep Gaussian process structure is selected to construct a large alternative model, which can effectively reduce the requirements for the required training data, and can be trained with limited data to form an effective large alternative model. It is more accurate to use the target posterior probability distribution of the input parameters of the trained large alternative model to replace the preset prior probability distribution of the model to be calibrated for input parameter calibration in the later stage.

[0098] The above experimental data set may be an integration of multiple experimental data related to the model to be calibrated.

[0099] In the above case where N is less than a preset value, a linear deep Gaussian process structure is selected to construct an alternative large model. This may be a case where the overall data volume in the experimental data set is relatively small, in which a linear deep Gaussian process structure is selected to construct an alternative large model.

[0100] The above-mentioned linear deep Gaussian process structure can be a linear activation function in which the hidden layer in the deep Gaussian process adopts a single hidden layer or multiple hidden layers.

[0101] In some embodiments, after obtaining the number N of experimental data, the method may further include:

[0102] When N is greater than or equal to a preset value, the relationship between the input parameters and the output parameters of the model to be calibrated is analyzed to obtain a linear analysis result, wherein the linear analysis result is a linear correlation or a nonlinear correlation;

[0103] If the linear analysis results in nonlinear correlation, a nonlinear deep Gaussian process structure is selected to construct an alternative large model;

[0104] If the linear analysis result is linearly correlated, a linear deep Gaussian process structure is selected to construct an alternative large model.

[0105] In this embodiment, when N is greater than or equal to a preset value, the relationship between the input parameters and the output parameters of the model to be calibrated is analyzed to obtain a linear analysis result. If the linear analysis result is nonlinearly correlated, a nonlinear deep Gaussian process structure is selected to construct an alternative large model. If the linear analysis result is linearly correlated, a linear deep Gaussian process structure is selected to construct an alternative large model. In this way, when the amount of data is relatively large, by analyzing the relationship between the input parameters and the output parameters, a suitable deep Gaussian process structure is selected to construct an alternative large model, and in the later stage, the target posterior probability distribution of the input parameters of the trained alternative large model is used to replace the preset prior probability distribution of the model to be calibrated for input parameter calibration, which is more accurate.

[0106] The above output parameters may correspond to the input parameters of the model to be calibrated, that is, the input parameters of the model to be calibrated may correspond to the output parameters.

[0107] When N is greater than or equal to a preset value, analyzing the relationship between the input parameters and output parameters of the model to be calibrated may be analyzing the relationship between the input parameters and output parameters of the model to be calibrated when the overall data volume in the experimental data set is relatively large.

[0108] The above analysis of the relationship between the input parameters and output parameters of the model to be calibrated can be determined by expert judgment, or by calculating the linear correlation coefficient between the input parameters and output parameters of the model to be calibrated through sensitivity analysis, for example, Pearson correlation coefficient (Pearson coefficient), Kendall correlation coefficient (Kendall coefficient) and Spearman correlation coefficient (Spearman coefficient) for linear judgment.

[0109] The above input parameters and output parameters may be caused by multiple key phenomena, and one key phenomenon may correspond to multiple input parameters and output parameters.

[0110] The above-mentioned deep Gaussian process can be regarded as a multi-layer perceptron (MLP) with a single hidden layer and infinite hidden layer nodes. The deep Gaussian process (DGP) is the Gaussian process obtained when the MLP is extended from a single hidden layer to multiple hidden layers, that is, an MLP with multiple hidden layers and infinite width.

[0111] The above-mentioned linear deep Gaussian process structure can be a linear activation function in which the hidden layer in the deep Gaussian process adopts a single hidden layer or multiple hidden layers.

[0112] The nonlinear deep Gaussian process structure may be a deep Gaussian process in which the hidden layer adopts a nonlinear activation function of a single hidden layer or multiple hidden layers.

[0113] In some embodiments, after training the surrogate large model with the training data set to obtain the trained surrogate large model, before determining the target posterior probability distribution of the input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set, the method may further include:

[0114] Calculate the mean square error of the trained surrogate model;

[0115] According to the predicted value set of the trained surrogate model and the validation data set, the target posterior probability distribution of the input parameters of the trained surrogate model is determined, including:

[0116] If the mean square error value meets the preset threshold range, the input data in the verification data set is input into the trained alternative large model to obtain the predicted value set of the trained alternative large model;

[0117] The target posterior probability distribution of the input parameters of the trained surrogate large model is determined according to the predicted values ​​in the predicted value set of the trained surrogate large model and the output true values ​​corresponding to the input data in the verification data set.

[0118] In this embodiment, by comparing the mean square error value of the trained alternative large model with the preset threshold range, a prediction value set of the trained alternative large model that meets the preset threshold range is obtained, and then the target posterior probability distribution of the input parameters of the trained alternative large model is determined according to the prediction value in the prediction value set of the trained alternative large model and the output true value corresponding to the input data in the verification data set, which can improve the accuracy of the calculation, thereby improving the accuracy of the input parameters of the posterior expectation calibration model. The alternative large model is constructed through a deep Gaussian process structure. After the model is established and tested, the waste of computing resources caused by repeated running of the computing software can be improved, and the computing efficiency can be improved.

[0119] The above mean square error can be a measure of the degree of difference between the estimator and the estimated quantity.

[0120] The above-mentioned preset threshold range can be set according to the actual needs of the user and can be changed according to the actual situation. This embodiment does not make any specific limitations.

[0121] If the mean square error value does not meet the preset threshold range, the replacement large model is retrained until the mean square error value of the trained replacement large model meets the preset threshold range.

[0122] The target posterior probability distribution of the input parameters of the trained alternative large model is determined according to the predicted values ​​in the predicted value set of the trained alternative large model and the true output values ​​corresponding to the input data in the verification data set. The target posterior probability distribution of the input parameters of the trained alternative large model can be determined by using the maximum likelihood estimation method according to the predicted values ​​in the predicted value set of the trained alternative large model and the true output values ​​corresponding to the input data in the verification data set.

[0123] In some embodiments, determining the target posterior probability distribution of the input parameters of the trained surrogate macromodel according to the predicted values ​​in the predicted value set of the trained surrogate macromodel and the output true values ​​corresponding to the input data in the validation data set may include:

[0124] Obtaining the prior probability distribution of the model to be calibrated;

[0125] Compare the predicted values ​​in the predicted value set of the trained alternative large model with the output true values ​​in the validation data set to obtain a comparison result;

[0126] Based on the comparison results and the prior probability distribution, the maximum likelihood estimation method is used to calculate the posterior probability distribution of the input parameters of the trained surrogate large model;

[0127] If the maximum likelihood estimation method meets the preset acceptance criteria, the posterior probability distribution is determined as the target posterior probability distribution;

[0128] If the maximum likelihood estimation method does not meet the preset acceptance criteria, the posterior probability distribution is used to replace the prior probability distribution, and the execution is returned to calculate the posterior probability distribution of the input parameters of the trained replacement large model using the maximum likelihood estimation method based on the comparison result and the prior probability distribution, until the maximum likelihood estimation method meets the preset acceptance criteria, and the posterior probability distribution calculated after the maximum likelihood estimation method meets the preset acceptance criteria is determined as the target posterior probability distribution.

[0129] In this embodiment, the predicted values ​​in the predicted value set of the trained alternative large model are compared with the output true values ​​in the verification data set to obtain a comparison result, and then the maximum likelihood estimation method is used to calculate the posterior probability distribution of the input parameters of the trained alternative large model based on the comparison result and the prior probability distribution, until the maximum likelihood estimation method reaches the preset acceptance criteria and obtains the target posterior probability distribution. A relatively accurate posterior probability distribution can be obtained, laying the foundation for improving the accuracy of the input parameters of the model to be calibrated in the later stage.

[0130] The above-mentioned preset acceptance criteria may be acceptance criteria set by the user.

[0131] The above-mentioned obtaining of the prior probability distribution of the model to be calibrated may be obtaining an initial prior probability distribution, where the initial prior probability distribution is a preset probability distribution. However, the prior probability distribution will be updated with the iteration of the maximum likelihood estimation method.

[0132] The above method of calculating the posterior probability distribution of the input parameters of the trained alternative large model using the maximum likelihood estimation method based on the comparison results and the prior probability distribution can be performed by inputting the comparison results and the prior probability distribution into the maximum likelihood estimation method to calculate the posterior probability distribution of the input parameters of the trained alternative large model.

[0133] In some embodiments, generating a training dataset and a validation dataset based on an experimental dataset related to the model to be calibrated may include:

[0134] Obtain the number of experimental data M used for training the model to be calibrated, where M is a positive integer;

[0135] Select M data from the experimental data set;

[0136] Generate a training data set including M data.

[0137] In this embodiment, the training data set is first screened out to ensure the independence of large model training and provide higher reliability.

[0138] The number M of experimental data obtained for training the model to be calibrated may be determined based on specific user requirements. The number M of experimental data may be changed based on actual conditions and is not specifically limited in this embodiment.

[0139] Before selecting M data from the experimental data set, the separation effect experiment and the overall effect experiment corresponding to the experimental data set are first checked to see whether they cover the key phenomena related to the model to be calibrated. If they do, the experimental data set is available. If they do not, the experimental data set is unavailable. When the experimental data set is unavailable, it is necessary to reselect the experimental data set.

[0140] In some embodiments, generating a training dataset and a validation dataset based on an experimental dataset related to the model to be calibrated may include:

[0141] Remove the training data set from the experimental data set to obtain the remaining experimental data set;

[0142] generating a plurality of subsets using at least a portion of the data in the remaining experimental data set;

[0143] Selecting a preset number of first data from each subset;

[0144] The first data selected from each subset is combined into a verification data set.

[0145] In this embodiment, since the verification data set is screened out from the remaining experimental data sets, and the remaining experimental data set is the data set after the training data set is removed from the experimental data set, the verification data set is screened out by adopting a cross-validation method, which ensures the independence of the large model usability evaluation and can provide higher reliability.

[0146] The above-mentioned generation of multiple subsets by using at least part of the data in the remaining experimental data set may be based on part of the data in the remaining experimental data set, or may be based on all of the data in the remaining experimental data set.

[0147] The above preset number can be set according to user needs and can be changed according to actual conditions, and this embodiment does not impose any specific restrictions.

[0148] The combining of the first data selected from each subset into a verification data set may be combining a preset number of first data selected from each subset to obtain a verification data set.

[0149] For example, each subset randomly selects N_val number of experimental data (ie, a preset number of first data), and then combines the N_val number of experimental data selected by each subset into a verification data set A_val.

[0150] In some embodiments, generating multiple subsets using at least part of the data in the remaining experimental data set may include:

[0151] Determine the input variables of the model to be calibrated;

[0152] Acquire a target range, where the target range includes at least one of a range of an input variable and a range of an input parameter;

[0153] If the target range is within the range corresponding to the remaining experimental data sets, multiple subsets are generated using the data in the remaining experimental data sets that are within the target range;

[0154] If the target range is larger than the range corresponding to the remaining experimental data sets, all the data of the remaining experimental data sets are used to generate multiple subsets.

[0155] In this embodiment, by generating multiple appropriate subsets according to the range of input variables and the range of input parameters, a good data foundation is laid for later screening out more appropriate verification data sets.

[0156] The above-mentioned determination of the input variables of the model to be calibrated can be determined by the user based on the PIRT table, determining the key phenomena related to the accident conditions for different accident conditions to be analyzed, and then making expert judgment to determine the input variables of the model to be calibrated related to the key phenomena.

[0157] The above input variables can be boundary conditions or initial conditions.

[0158] The range of the above-mentioned input variables and the range of the input parameters can be determined according to the range of the operating parameters related to the key phenomena that need to be analyzed.

[0159] The range of the operating condition parameters related to the key phenomena mentioned above may be the operating condition range obtained according to the analysis range of the PIRT table.

[0160] The above operating condition parameters may be input parameters or input variables.

[0161] If the target range is within the range corresponding to the remaining experimental data sets, multiple subsets are generated using the data within the target range in the remaining experimental data sets. This may be when the range of the operating parameters related to the key phenomenon corresponding to the remaining experimental data sets can cover the range of the input variables and the range of the input parameters, at least one of the range of the input variables and the range of the input parameters is evenly divided into a preset number of groups to obtain the grouping range of each group, and then experimental data corresponding to the grouping range of each group is selected from the remaining experimental data sets to obtain multiple subsets, the number of data in each subset is the same, and the number of multiple subsets is equal to the preset number of groups.

[0162] For example, the range of an input variable is [0, N), and it needs to be divided into n subsets. The grouping range of each subset is [0, N / n), [N / n, 2N / n), ..., [(n-1)N / n, N), and then experimental data are selected from the remaining experimental data sets according to the grouping range of each subset to obtain the experimental data contained in each subset.

[0163] The above-mentioned preset number of groups can be grouped according to the actual needs of the user and can be changed according to the actual situation. This embodiment does not make any specific restrictions.

[0164] If the target range is larger than the range corresponding to the remaining experimental data sets, then all the data of the remaining experimental data sets are used to generate multiple subsets. When the range of the operating parameters related to the key phenomena corresponding to the remaining experimental data sets cannot cover the range of the input variables and the range of the input parameters, the range corresponding to the remaining experimental data sets is evenly divided into a preset number of groups to obtain the grouping range of each group, and then the experimental data corresponding to the grouping range of each group is selected from the remaining experimental data sets to obtain multiple subsets. The amount of data in each subset is the same, and the number of multiple subsets is equal to the preset number of groups.

[0165] The target range includes at least one of a range of an input variable and a range of an input parameter, and the target range may include a range of an input variable and a range of an input parameter.

[0166] When the target range includes two ranges, namely the range of the input variable and the range of the input parameter, if one of the ranges is within the range corresponding to the remaining experimental data sets and the other range is larger than the range corresponding to the remaining experimental data sets, the range within the range corresponding to the remaining experimental data sets is evenly divided into a preset number of groups, and if the other range is larger than the range corresponding to the remaining experimental data sets, the range corresponding to the remaining experimental data sets is evenly divided into the preset number of groups.

[0167] It should be noted that the various implementation methods provided in the embodiments of the present application can be implemented independently or in combination without conflicting with each other. The specific implementation method can be determined based on actual needs, and the embodiments of the present application are not limited to this.

[0168] To facilitate understanding by those skilled in the art, a set of best embodiments is provided below:

[0169] This embodiment will establish an alternative large model through deep learning algorithms to achieve input parameter calibration and reverse uncertainty quantification of the software model, so as to replace the input uncertainty determination methods such as expert judgment and user experience that may cause human bias. Figure 2 , the specific process is as follows:

[0170] Step (1) determining the software calculation model to be analyzed, the software calculation model to be analyzed is nuclear engineering design and safety analysis software, which includes multiple models, for example, a heat transfer calculation model and a flow calculation model, etc., and the model to be calibrated is a model in the software calculation model to be analyzed. Determine the input variables and input model parameters (i.e., input parameters) related to the model to be calibrated.

[0171] Step (2) Linear analysis: Perform a preliminary analysis of the linear relationship between the output parameters of the calibration model and the input model parameters to determine whether a linear assumption can be made for the calibration model.

[0172] Linear analysis can be determined through expert judgment or by calculating linear correlation coefficients through sensitivity analysis, such as Pearson, Kendall, Spearman coefficients, etc. to make linear judgments and establish linear assumptions.

[0173] Step (3) Select the training experimental data set (i.e., training data set) and the verification experimental data set (i.e., verification data set). The selection of relevant training experimental data sets and verification experimental data sets is carried out for the problems that need to be quantified. It is necessary to ensure the adequacy of the training experimental data set and the coverage of the verification experimental data set for the verification range. The coverage refers to the degree to which the separation effect experiment and the overall effect experiment in the experimental data set cover all key phenomena. The experimental data should cover all key phenomena and have redundancy, so as to determine whether the training experimental data set and the verification experimental data set are available or not. The specific implementation steps are as follows:

[0174] 1) Collect experimental data A generated by separation effect experiments and overall effect experiments related to the model to be calibrated, determine the number of available experimental data N, and number the experimental data;

[0175] 2) Select the initial experimental data set for training: Determine the number of experimental data N_test for training according to the specific needs of the user, and randomly select N_test experimental data from the collected experimental data A as the training data set A_test;

[0176] 3) Select the validation data set: In the (A-A_test) data set (i.e., the remaining experimental data set), group the (A-A_test) data set according to the range of the input variables and the range of the input model parameters to form n subsets, and randomly extract N_val of experimental data from each subset, and then combine the experimental data in each subset as the validation experimental data set A_val.

[0177] Step (4) Constructing a large alternative model: Deep Gaussian Process (DGP) is used to implement deep learning training. This method can effectively reduce the requirements for the required training data. For the problem of lack of relevant experimental data that is common in the nuclear engineering field, limited data can be used for training to form an effective large alternative model. The large alternative model is constructed according to the characteristics of the software model to be analyzed and the number of experimental data sets, specifically:

[0178] 1) The relationship between the output parameters and input parameters of the software calculation model to be analyzed is nonlinear: a nonlinear DGP structure is selected, and the hidden layer in the DGP structure uses a single hidden layer or a multi-hidden layer nonlinear activation function to fit the nonlinear relationship existing in the nuclear engineering software;

[0179] 2) The relationship between the output parameters and input parameters of the software calculation model to be analyzed is linear or the number of experimental data sets is small: Select the linear DGP structure, and the hidden layer in the DGP structure uses a single hidden layer or multiple hidden layers of linear activation functions to fit the linear relationship existing in the nuclear engineering software.

[0180] Step (5) training the alternative large model: based on the training experimental data set, the constructed alternative large model is trained using the maximum likelihood estimation method.

[0181] Step (6) Evaluate model availability: Calculate the mean square error (MSE) value of the trained alternative large model to determine whether the DGP performance meets the requirements. By setting an acceptance threshold (i.e., a preset threshold range), when the MSE value meets the acceptance threshold, it can be determined that the DGP performance meets the requirements. When the MSE value does not meet the acceptance threshold, the alternative large model can be retrained.

[0182] Step (7) Model calibration or reverse uncertainty quantification: Use the established DGP to replace the large model to replace the predicted output value of the original software calculation model to be analyzed.

[0183] Specifically, the verification experimental data set includes input data (i.e., experimental conditions) and output true values ​​corresponding to the input data (i.e., known experimental results); the input data in the verification experimental data set is input into the alternative large model selected in step (6) for prediction to obtain a prediction value set (i.e., since the verification experimental data set includes multiple input data, each input data will have a corresponding prediction value); the prediction values ​​in the prediction value set are compared with the output true values ​​in the verification experimental data set to obtain a comparison result; finally, based on the comparison result and the prior probability distribution (the prior probability distribution when the maximum likelihood estimation method is calculated for the first time is the initial prior probability distribution pre-set by the user), the maximum likelihood estimation method is used to estimate the posterior probability distribution of the input parameters of the alternative large model, and the posterior probability distribution obtained after the maximum likelihood estimation method reaches the preset acceptance standard is used to replace the initial prior probability distribution (i.e., the preset prior probability distribution), thereby realizing the calibration of the input parameters of the model to be calibrated.

[0184] In this embodiment, the input parameters of the nuclear engineering software model are calibrated according to the existing experimental data, and the uncertainty is statistically quantified. The overall process can minimize the human bias caused by the introduction of expert judgment or user effect, and can solve the current problem of differences in positive uncertainty quantification results caused by human bias in input uncertainty.

[0185] The replacement large model is constructed through the deep machine learning algorithm. After the replacement large model is established and tested, the replacement large model is used to replace the original calculation model of the software to be analyzed, rather than the model to be calibrated. This is because some models are not within the scope of calibration. If the entire calculation model of the software to be analyzed is run, the amount of calculation will increase. Therefore, the replacement large model is used to replace the original calculation model of the software to be analyzed, which can improve the waste of computing resources caused by repeated running of the calculation software, improve computing efficiency, and calibrate the parameters of the model to be calibrated. The deep Gaussian process method can effectively reduce the requirements for the required training data. For the problem of lack of relevant experimental data that is common in the nuclear engineering field, limited data can be used to form an effective replacement model. The cross-validation method is used to screen the experimental data set for training and the experimental data set for verification, respectively, to ensure the independence of the construction, training and usability evaluation of the replacement large model, and to provide higher reliability.

[0186] Reference Figure 3 , is a schematic diagram of the structure of a calibration system for nuclear engineering software model input parameters provided in an embodiment of the present application. The system may include:

[0187] A data set generation module 310 is used to generate a training data set and a validation data set according to an experimental data set related to the model to be calibrated;

[0188] A large model construction module 320 is used to construct a substitute large model corresponding to the model to be calibrated;

[0189] A large model training module 330 is used to train the substitute large model through a training data set to obtain a trained substitute large model;

[0190] The probability distribution determination module 340 is used to determine the target posterior probability distribution of the input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the verification data set;

[0191] The input parameter calibration module 350 is used to replace the preset prior probability distribution of the model to be calibrated with the target posterior probability distribution, wherein the preset prior probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

[0192] In some implementations, the large model building module 320 may be specifically used to:

[0193] Get the number of experimental data N, where N is a positive integer;

[0194] When N is less than a preset value, a linear deep Gaussian process structure is selected to construct an alternative large model.

[0195] In some implementations, the large model building module 320 may be specifically used to:

[0196] When N is greater than or equal to a preset value, the relationship between the input parameters and the output parameters of the model to be calibrated is analyzed to obtain a linear analysis result, wherein the linear analysis result is a linear correlation or a nonlinear correlation;

[0197] If the linear analysis results in nonlinear correlation, a nonlinear deep Gaussian process structure is selected to construct an alternative large model;

[0198] If the linear analysis result is linearly correlated, a linear deep Gaussian process structure is selected to construct an alternative large model.

[0199] In some implementations, the probability distribution determination module 340 may be specifically configured to:

[0200] Calculate the mean square error of the trained surrogate model;

[0201] According to the predicted value set of the trained surrogate model and the validation data set, the target posterior probability distribution of the input parameters of the trained surrogate model is determined, including:

[0202] If the mean square error value meets the preset threshold range, the input data in the verification data set is input into the trained alternative large model to obtain the predicted value set of the trained alternative large model;

[0203] The target posterior probability distribution of the input parameters of the trained surrogate large model is determined according to the predicted values ​​in the predicted value set of the trained surrogate large model and the output true values ​​corresponding to the input data in the verification data set.

[0204] In some implementations, the probability distribution determination module 340 may be specifically configured to:

[0205] Obtaining the prior probability distribution of the model to be calibrated;

[0206] Compare the predicted values ​​in the predicted value set of the trained alternative large model with the output true values ​​in the validation data set to obtain a comparison result;

[0207] Based on the comparison results and the prior probability distribution, the maximum likelihood estimation method is used to calculate the posterior probability distribution of the input parameters of the trained surrogate large model;

[0208] If the maximum likelihood estimation method meets the preset acceptance criteria, the posterior probability distribution is determined as the target posterior probability distribution;

[0209] If the maximum likelihood estimation method does not meet the preset acceptance criteria, the posterior probability distribution is used to replace the preset prior probability distribution, and the execution is returned to calculate the posterior probability distribution of the input parameters of the trained replacement large model using the maximum likelihood estimation method based on the comparison result and the prior probability distribution, until the maximum likelihood estimation method meets the preset acceptance criteria, and the posterior probability distribution calculated after the maximum likelihood estimation method meets the preset acceptance criteria is determined as the target posterior probability distribution.

[0210] In some implementations, the data set generation module 310 may be specifically configured to:

[0211] Obtain the amount of experimental data M used for training the model to be calibrated;

[0212] Select M data from the experimental data set;

[0213] Generate a training data set including M data.

[0214] In some implementations, the data set generation module 310 may be specifically configured to:

[0215] Determine the input variables of the model to be calibrated;

[0216] Remove the training data set from the experimental data set to obtain the remaining experimental data set;

[0217] generating a plurality of subsets using at least a portion of the data in the remaining experimental data set;

[0218] Selecting a preset number of first data sets from each subset;

[0219] The first data set selected from each subset is combined into a validation data set.

[0220] In some implementations, the data set generation module 310 may be specifically configured to:

[0221] Acquire a target range, where the target range includes at least one of a range of an input variable and a range of an input parameter;

[0222] If the target range is within the range corresponding to the remaining experimental data sets, multiple subsets are generated using the data in the remaining experimental data sets that are within the target range;

[0223] If the target range is larger than the range corresponding to the remaining experimental data sets, all the data of the remaining experimental data sets are used to generate multiple subsets.

[0224] It should be noted that since the calibration device for the input parameters of a nuclear engineering software model in this embodiment and the calibration method for the input parameters of a nuclear engineering software model mentioned above are based on the same inventive concept, the corresponding contents in the method embodiment are also applicable to the present system embodiment and will not be described in detail here.

[0225] Reference Figure 4 The embodiment of the present application further provides an electronic device, which may include:

[0226] at least one memory;

[0227] at least one processor;

[0228] at least one program;

[0229] The programs are stored in the memory, and the processor executes at least one program to implement the calibration method of the nuclear engineering software model input parameters described above in the present disclosure.

[0230] The electronic device may be any intelligent terminal including a mobile phone, a tablet computer, a personal digital assistant (PDA), a car computer, etc.

[0231] The electronic device according to the embodiment of the present application is described in detail below.

[0232] The processor 410 may be implemented by a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present disclosure;

[0233] The memory 420 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 420 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 420, and the processor 410 calls and executes the calibration method of the nuclear engineering software model input parameters of the embodiment of the present disclosure.

[0234] Input / output interface 430, used to implement information input and output;

[0235] Communication interface 440, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0236] bus 450 , which transmits information between the various components of the device (e.g., processor 410 , memory 420 , input / output interface 430 , and communication interface 440 );

[0237] The processor 410 , the memory 420 , the input / output interface 430 , and the communication interface 440 are connected to each other in communication within the device via the bus 450 .

[0238] The embodiment of the present disclosure also provides a storage medium, which is a computer-readable storage medium. The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the calibration method of the input parameters of the nuclear engineering software model.

[0239] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0240] The embodiments described in the embodiments of the present disclosure are intended to more clearly illustrate the technical solutions of the embodiments of the present disclosure and do not constitute a limitation on the technical solutions provided by the embodiments of the present disclosure. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present disclosure are also applicable to similar technical problems.

[0241] Those skilled in the art will appreciate that the technical solutions shown in the figures do not limit the embodiments of the present disclosure and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0242] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0243] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0244] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0245] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0246] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0247] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0248] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0249] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store programs. The above is a detailed description of the embodiments of the present application in conjunction with the accompanying drawings, but the present application is not limited to the above embodiments. Within the scope of knowledge possessed by ordinary technicians in the relevant technical field, various changes can be made without departing from the purpose of the present application.

Claims

1. A method for calibrating input parameters of a nuclear engineering software model, characterized in that: The method comprises: Generate a training data set and a validation data set based on an experimental data set related to the model to be calibrated; Constructing a substitute large model corresponding to the model to be calibrated; Training the replacement large model using the training data set to obtain a trained replacement large model; Determining a target posterior probability distribution of input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set; The preset priori probability distribution of the model to be calibrated is replaced by the target posterior probability distribution, wherein the preset priori probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

2. The method for calibrating input parameters of a nuclear engineering software model according to claim 1, characterized in that: The step of constructing a substitute large model corresponding to the model to be calibrated comprises: Obtain the number N of data in the experimental data set, where N is a positive integer; When N is less than a preset value, a linear deep Gaussian process structure is selected to construct an alternative large model.

3. The method for calibrating input parameters of a nuclear engineering software model according to claim 2, characterized in that: After obtaining the number N of experimental data, the method further includes: When N is greater than or equal to the preset value, analyzing the relationship between the input parameters and the output parameters of the model to be calibrated to obtain a linear analysis result, wherein the linear analysis result is a linear correlation or a nonlinear correlation; If the linear analysis result is a nonlinear correlation, a nonlinear deep Gaussian process structure is selected to construct an alternative large model; If the linear analysis result is a linear correlation, a linear deep Gaussian process structure is selected to construct an alternative large model.

4. The method for calibrating input parameters of a nuclear engineering software model according to claim 1, characterized in that: After training the surrogate large model with the training data set to obtain the trained surrogate large model, before determining the target posterior probability distribution of the input parameters of the trained surrogate large model based on the predicted value set of the trained surrogate large model and the validation data set, the method further includes: Calculate the mean square error value of the trained alternative large model; Determining the target posterior probability distribution of the input parameters of the trained alternative large model according to the predicted value set of the trained alternative large model and the verification data set includes: If the mean square error value satisfies a preset threshold range, inputting the input data in the verification data set into the trained replacement large model to obtain a prediction value set of the trained replacement large model; The target posterior probability distribution of the input parameters of the trained alternative large model is determined according to the predicted values ​​in the predicted value set of the trained alternative large model and the output true values ​​corresponding to the input data in the verification data set.

5. The method for calibrating input parameters of a nuclear engineering software model according to claim 4, characterized in that: The step of determining the target posterior probability distribution of the input parameters of the trained replacement large model according to the predicted values ​​in the predicted value set of the trained replacement large model and the output true values ​​corresponding to the input data in the verification data set comprises: Obtaining a priori probability distribution of the model to be calibrated; Compare the predicted values ​​in the predicted value set of the trained replacement large model with the output true values ​​in the verification data set to obtain a comparison result; According to the comparison result and the prior probability distribution, a posterior probability distribution of the input parameters of the trained replacement large model is calculated using a maximum likelihood estimation method; If the maximum likelihood estimation method meets the preset acceptance criteria, the posterior probability distribution is determined as the target posterior probability distribution; If the maximum likelihood estimation method does not meet the preset acceptance criteria, the posterior probability distribution is used to replace the prior probability distribution, and the process returns to executing the method of calculating the posterior probability distribution of the input parameters of the trained replacement large model using the maximum likelihood estimation method based on the comparison result and the prior probability distribution, until the maximum likelihood estimation method meets the preset acceptance criteria, and the posterior probability distribution calculated after the maximum likelihood estimation method meets the preset acceptance criteria is determined as the target posterior probability distribution.

6. The method for calibrating input parameters of a nuclear engineering software model according to claim 1, characterized in that: The step of generating a training data set and a verification data set according to an experimental data set related to the model to be calibrated includes: Obtaining the number M of experimental data used for training the model to be calibrated, where M is a positive integer; Select M data from the experimental data set; Generate a training data set including the M data.

7. The method for calibrating input parameters of a nuclear engineering software model according to claim 1, characterized in that: The step of generating a training data set and a verification data set according to an experimental data set related to the model to be calibrated includes: Removing the training data set from the experimental data set to obtain a remaining experimental data set; Using at least part of the data in the remaining experimental data set, generating a plurality of subsets; Selecting a preset number of first data from each subset; The first data selected from each subset are combined into a verification data set.

8. The method for calibrating input parameters of a nuclear engineering software model according to claim 7, characterized in that: The step of generating a plurality of subsets by using at least part of the data in the remaining experimental data set comprises: Determining input variables of the model to be calibrated; Acquire a target range, wherein the target range includes at least one of a range of the input variable and a range of the input parameter; If the target range is within the range corresponding to the remaining experimental data sets, generate multiple subsets using the data in the remaining experimental data sets that are within the target range; If the target range is larger than the range corresponding to the remaining experimental data sets, all data of the remaining experimental data sets are used to generate multiple subsets.

9. A calibration system for input parameters of a nuclear engineering software model, characterized in that: The system comprises: A data set generation module, used to generate a training data set and a verification data set according to an experimental data set related to the model to be calibrated; A large model construction module, used to construct a substitute large model corresponding to the model to be calibrated; A large model training module, used for training the replacement large model with the training data set to obtain a trained replacement large model; A probability distribution determination module, used to determine a target posterior probability distribution of input parameters of the trained replacement large model based on the predicted value set of the trained replacement large model and the verification data set; An input parameter calibration module is used to replace a preset priori probability distribution of the model to be calibrated by the target posterior probability distribution, wherein the preset priori probability distribution is a pre-set probability distribution corresponding to the input parameters of the model to be calibrated.

10. An electronic device, characterized in that: comprising at least one control processor and a memory for communicatively coupling with the at least one control processor; The memory stores instructions that can be executed by at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to perform the calibration method of nuclear engineering software model input parameters according to any one of claims 1 to 8.

Images