Nuclear accident source item inversion method and device based on CatBoost ensemble learning algorithm

Abstract

The invention relates to a nuclear accident source item inversion method and device based on a CatBoost ensemble learning algorithm. Obtaining a target signal and an input characteristic variable of nuclear accident source item inversion; performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set of nuclear accident source item inversion; according to the training set and the test set, determining a CatBoost integrated learning model structure parameter and a parameter range of nuclear accident source item inversion; according to the training set, the structure parameters and the parameter range of the CatBoost integrated learning model, the Harlins eagle algorithm and the CatBoost integrated learning model, obtaining a hyper-parameter optimized CatBoost integrated learning model for nuclear accident source item inversion; and performing model performance evaluation and interpretability analysis of nuclear accident source item inversion on the CatBoost integrated learning model after hyper-parameter optimization by using the test set. According to the method, the automatic adjustment of the structure parameters of the CatBoost integrated learning model is realized, and the complexity of parameter adjustment and the uncertainty of manual parameter optimization are reduced; the method can meet the accuracy of source item inversion, and has higher interpretability.

Description

Technical Field

[0001] The present invention relates to the field of nuclear application technology, and more specifically, to a method and device for nuclear accident source term inversion based on CatBoost ensemble learning algorithm. Background Art

[0002] Nuclear accident source terms are key data for accident rating and nuclear accident consequence assessment. Nuclear accidents and accidental releases of ruthenium have proven that the traditional forward modeling method that relies on nuclear facility condition monitoring data is ineffective. The uncertainty of accident source terms will hinder the implementation of nuclear emergency response and bring challenges to nuclear safety.

[0003] There are usually various environmental monitoring equipment around nuclear facilities. These equipment are less affected by accidents when they occur, and various radioactive monitoring methods can be implemented during accidents to collect a large amount of monitoring data, which provides data support for the method of using environmental monitoring data to invert the accident source term. At present, the main methods for source term inversion include optimal interpolation, filtering algorithm, variational assimilation, iterative optimization methods such as genetic algorithms, and machine learning algorithms such as neural networks. The process of iterative optimization inversion of source terms requires continuous invocation of atmospheric diffusion models to improve source terms, narrow the gap between simulation and actual monitoring, and thus find a more ideal release source term. This type of method relies heavily on prior information and the accuracy of the forward diffusion model, and the inversion reasoning time cost is high due to the need for continuous forward calculations during iterations, which brings challenges to the rapid and accurate calculation of source terms required for nuclear emergency response. Machine learning algorithms such as neural networks have powerful processing capabilities for complex nonlinear problems such as nuclear accident source term inversion. By learning a large number of data samples, using hidden layers to extract and analyze features, and adaptively adjusting the model according to the gap between network output and expected output, the relationship between input features and output targets is established. This method requires adjusting the model hyperparameters for different tasks to obtain optimal performance, but there is currently no unified way to adjust hyperparameters. It will take a lot of time and it is difficult to find the optimal parameter combination through empirical selection and trial and error. At the same time, due to the fact that machine learning itself is a black box model, the established prediction model has not been well explained, so it is impossible to accurately determine which factors have the greatest impact on the inversion of nuclear accident source terms, resulting in the inability to achieve high-precision and explainable inversion prediction of accident source terms. Summary of the invention

[0004] The embodiments of the present application provide a method and device for inverting the source term of a nuclear accident based on a CatBoost ensemble learning algorithm. In order to have a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary is not a general review, nor is it intended to identify key / important components or describe the scope of protection of these embodiments. Its sole purpose is to present some concepts in a simple form as a preface to the detailed description that follows.

[0005] In a first aspect, the present application provides a method for inverting nuclear accident source terms based on a CatBoost ensemble learning algorithm, the method comprising:

[0006] Obtain target signals and input characteristic variables for nuclear accident source term inversion;

[0007] Performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion;

[0008] Determining the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set;

[0009] According to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model, a hyperparameter-optimized CatBoost ensemble learning model for the nuclear accident source term inversion is obtained;

[0010] The test set is used to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the CatBoost ensemble learning model after hyperparameter optimization.

[0011] According to a preferred embodiment, the step of obtaining the target signal and input characteristic variables for the nuclear accident source term inversion includes:

[0012] Acquire the release rate of I-131 nuclides after the nuclear accident occurs in the inversion of the nuclear accident source term, and use the release rate of I-131 nuclides after the nuclear accident occurs as the target signal of the inversion of the nuclear accident source term;

[0013] According to the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables of the inversion of the nuclear accident source term.

[0014] According to a preferred embodiment, determining the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion based on the training set and the test set includes:

[0015] Determine the maximum number of trees, maximum depth, and learning rate in the decision tree parameters for the nuclear accident source term inversion according to the training set and the test set, and determine the search range of each parameter by a five-fold cross validation method;

[0016] The maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the nuclear accident source term inversion;

[0017] Each parameter search range is used as the parameter range for inversion of the nuclear accident source term.

[0018] According to a preferred embodiment, the hyperparameter-optimized CatBoost ensemble learning model for nuclear accident source term inversion is obtained based on the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model, including:

[0019] After the hyperparameter optimization, the CatBoost ensemble learning model has optimal hyperparameters.

[0020] According to a preferred embodiment, the use of the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model includes:

[0021] The CatBoost ensemble learning model after the hyperparameter optimization is tested and analyzed using the test set, and the CatBoost ensemble learning model after the hyperparameter optimization is compared with the ensemble tree model based on the ideas of Bagging and Boosting to evaluate the model performance of the nuclear accident source term inversion;

[0022] The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

[0023] In a second aspect, the present application provides a nuclear accident source term inversion device based on a CatBoost ensemble learning algorithm, the device comprising:

[0024] An acquisition module is used to obtain target signals and input characteristic variables for nuclear accident source term inversion;

[0025] A preprocessing module, used for performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion;

[0026] A parameter determination module, used to determine the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set;

[0027] A model optimization module, for obtaining a hyperparameter-optimized CatBoost ensemble learning model for nuclear accident source term inversion according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model;

[0028] A model evaluation module is used to use the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the CatBoost ensemble learning model after hyperparameter optimization.

[0029] According to a preferred embodiment, the acquisition module is specifically used for:

[0030] Acquire the release rate of I-131 nuclides after the nuclear accident occurs in the inversion of the nuclear accident source term, and use the release rate of I-131 nuclides after the nuclear accident occurs as the target signal of the inversion of the nuclear accident source term;

[0031] According to the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables of the inversion of the nuclear accident source term.

[0032] According to a preferred embodiment, the parameter determination module is specifically used to:

[0033] Determine the maximum number of trees, maximum depth, and learning rate in the decision tree parameters for the nuclear accident source term inversion according to the training set and the test set, and determine the search range of each parameter by a five-fold cross validation method;

[0034] The maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the nuclear accident source term inversion;

[0035] Each parameter search range is used as the parameter range for inversion of the nuclear accident source term.

[0036] According to a preferred embodiment, the model optimization module includes:

[0037] After the hyperparameter optimization, the CatBoost ensemble learning model has optimal hyperparameters.

[0038] According to a preferred embodiment, the model evaluation module is specifically used for:

[0039] The CatBoost ensemble learning model after the hyperparameter optimization is tested and analyzed using the test set, and the CatBoost ensemble learning model after the hyperparameter optimization is compared with the ensemble tree model based on the ideas of Bagging and Boosting to evaluate the model performance of the nuclear accident source term inversion;

[0040] The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

[0041] In a third aspect, the present application provides a computer storage medium, which stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

[0042] In a fourth aspect, the present application provides a terminal, which may include: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0043] The technical solution provided by this application may have the following beneficial effects:

[0044] In the present application, the nuclear accident source term inversion method based on the CatBoost ensemble learning algorithm obtains the target signal and input characteristic variables of the nuclear accident source term inversion; performs data preprocessing on the target signal and the input characteristic variables to obtain the training set and the test set of the nuclear accident source term inversion; determines the structural parameters and parameter ranges of the CatBoost ensemble learning model of the nuclear accident source term inversion according to the training set and the test set; obtains the hyperparameter-optimized CatBoost ensemble learning model of the nuclear accident source term inversion according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model; uses the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model. The present application realizes the automatic adjustment of the structural parameters of the CatBoost ensemble learning model through the joint application of HHO-CatBoost-SHAP. Compared with the prior art, the hyperparameters that the CatBoost ensemble learning model needs to optimize are simple, which can reduce the complexity of parameter adjustment. Combining with the Harris Eagle algorithm can reduce the uncertainty and time-consuming cost of manual parameter optimization. The model's own feature importance calculation, the CatBoost ensemble learning model after hyperparameter optimization and the SHAP calculation method are combined, which has higher interpretability while meeting the accuracy of source term inversion, and is of great significance for promoting the application of machine learning methods in the field of nuclear accident source term inversion and realizing the intelligent inversion of accident source terms.

[0045] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0047] Figure 1 It is a flow chart of a method for inverting a nuclear accident source term based on a CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0048] Figure 2 It is a schematic diagram of the overall process of a nuclear accident source term inversion method based on a CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0049] Figure 3 It is a schematic diagram of a hyperparameter tuning process of a nuclear accident source term inversion method based on a CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0050] Figure 4 It is a schematic diagram of absolute percentage error comparison of a source term inversion test of a CatBoost ensemble learning model after hyperparameter optimization of a nuclear accident source term inversion method based on a CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0051] Figure 5 It is a schematic diagram of the calculation results of the relative importance of features of a CatBoost ensemble learning model after hyperparameter optimization of a nuclear accident source term inversion method based on a CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0052] Figure 6 It is a schematic diagram of the SHAP feature importance calculation results of a nuclear accident source term inversion method based on the CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0053] Figure 7 It is a structural schematic diagram of a nuclear accident source term inversion device based on CatBoost ensemble learning algorithm provided in an embodiment of the present application;

[0054] Figure 8 It is a structural diagram of a terminal provided in an embodiment of the present application.

[0055] Reference numerals:

[0056] 10000, acquisition module, 20000, preprocessing module, 30000, parameter determination module, 40000, model optimization module, 50000, model evaluation module;

[0057] 1001. Processor, 1002. Communication bus, 1003. User interface, 1004. Network interface, 1005. Memory. DETAILED DESCRIPTION

[0058] The following description and the drawings sufficiently illustrate specific embodiments of the invention to enable those skilled in the art to practice them.

[0059] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are only examples of systems and methods consistent with some aspects of the present invention as detailed in the attached claims.

[0061] In the description of the present invention, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In addition, in the description of the present invention, unless otherwise specified, "plurality" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previously associated objects are in an "or" relationship.

[0062] The following will be combined with the attached Figure 1 -Attached Figure 6 , a nuclear accident source term inversion method based on CatBoost ensemble learning algorithm provided in an embodiment of the present application is introduced in detail.

[0063] See also Figure 1-6 The method of the embodiment of the present application may include the following steps:

[0064] As a type of machine learning algorithm, ensemble learning also has a strong ability to handle nonlinear problems. Compared with neural network algorithms, this method requires fewer parameters to be adjusted to establish an ensemble learning model, and can automatically process category features. At the same time, the ensemble learning model can rank the importance of each feature in general. With the development of theoretical research and computing equipment, this method has been applied in many fields, such as air quality prediction and equipment fault diagnosis. The SHAP interpretability analysis method is a post-model interpretation method based on cooperative game theory. By calculating the Shapley value, the impact of the model input on the output can be quantified. Combining the CatBoost ensemble learning algorithm with the SHAP calculation method can improve the model interpretability while meeting the speed and accuracy of the nuclear accident source term inversion. The embodiment of the present application proposes a method for inverting nuclear accident source terms based on the CatBoost ensemble learning algorithm, studies the inversion of nuclear accident source term information through radioactive air concentration monitoring values, and proposes a specific method and CatBoost ensemble learning model structure parameters suitable for inverting the I-131 nuclide release rate under the scenario of known release point location; the embodiment of the present application uses a Gaussian plume atmospheric diffusion model to simulate and obtain a data set for training and verification, uses the Harris Eagle algorithm to automatically adjust the hyperparameters of the CatBoost ensemble learning model, and trains and verifies the ensemble tree model, adds the SHAP calculation method to perform interpretability analysis of the CatBoost ensemble learning model after the hyperparameter optimization, and realizes the function of inverting nuclear accident source terms through radioactive air concentration.

[0065] S100, obtaining target signals and input characteristic variables for nuclear accident source term inversion, including:

[0066] The I-131 nuclide release rate after the nuclear accident occurs for the inversion of the nuclear accident source term is obtained, and the I-131 nuclide release rate after the nuclear accident occurs is used as the target signal for the inversion of the nuclear accident source term; based on the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables for the inversion of the nuclear accident source term.

[0067] In the embodiment of the present application, the release amount of I-131 nuclide can be used as one of the basis for grading severe nuclear accidents, and the release rate of I-131 nuclide after the nuclear accident is used as the target signal output by the following CatBoost ensemble learning model.

[0068] The nuclear accident monitoring data is the monitoring data around the nuclear accident facilities. The nuclear accident monitoring data includes the radioactive air concentration of the following five wind direction monitoring points, environmental influencing factors such as meteorological parameters monitored outside the nuclear power plant, and source term auxiliary parameters. Specifically, the five wind direction monitoring points are at distances of 0.5km, 1km, 2km, 5km, and 10km in the wind direction. Referring to the range of I-131 nuclide release rates in nuclear accidents in Table 1, the wind speed, atmospheric stability, mixing layer height, precipitation, and release height are randomly matched and combined by inputting the RADC Gaussian plume atmospheric diffusion model, and the radioactive air concentrations of the five wind direction monitoring points are calculated in a manner that includes as many scenarios as possible.

[0069] Table 1 Some examples of I-131 nuclide source term inversion

[0070]

[0071] Subsequently, the hyperparameter-optimized CatBoost ensemble learning model can be constructed according to the target signal and the input feature variables, and the target signal is used as the output layer of the CatBoost ensemble learning model, and the input feature variables are used as the input layer of the CatBoost ensemble learning model. The CatBoost ensemble learning model is also called a machine learning model or a network model.

[0072] S200, performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion.

[0073] In an embodiment of the present application, data normalization is performed on a data set containing a target signal and input feature variables as required. The data set can be normalized to between 0 and 1 using a standardized processing method using maximum and minimum values ​​to increase the speed of model training convergence; the data set is randomly divided into a training set and a test set in a ratio of 80%:20%.

[0074] S300, determining the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set, including:

[0075] According to the training set and the test set, the maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters for the inversion of the nuclear accident source term are determined, and the search range of each parameter is determined by a five-fold cross-validation method; the maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the inversion of the nuclear accident source term, and the structural parameters of the CatBoost ensemble learning model are also called the key hyperparameters of the CatBoost ensemble learning model. The search range of each parameter is used as the parameter range of the inversion of the nuclear accident source term, and the maximum number of trees search range is [1000, 2500], the maximum depth search range is [0.001, 0.1], and the learning rate search range is [5, 30].

[0076] S400, obtaining a hyperparameter-optimized CatBoost ensemble learning model for nuclear accident source term inversion according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model, including:

[0077] After the hyperparameter optimization, the CatBoost ensemble learning model has optimal hyperparameters.

[0078] In the embodiment of the present application, the Harris Hawk algorithm, referred to as HHO, is used to perform hyperparameter tuning on the CatBoost ensemble learning model, and the training set is divided into a training group and a test group using 5-fold cross validation to obtain the CatBoost ensemble learning model after hyperparameter optimization. The test group is also called the validation group.

[0079] like Figure 3 As shown in the figure, the hyperparameter tuning of the CatBoost ensemble learning model is performed using the Harris Eagle algorithm, including the following steps:

[0080] Preset the basic parameters of Harris Hawk Algorithm, including population size, maximum number of iterations, and the range and dimension of the hawk group and prey positions;

[0081] Initialize the hawk group position and prey position in the Harris Hawk algorithm respectively, and select the root mean square error RMSE of the CatBoost ensemble learning model as the fitness function; then perform iterative calculations, and select different position update strategies in each iteration to update the position vectors of the hawk group position and the prey position according to whether the prey is found, that is, whether the prey escapes successfully, the energy state of the prey, and the size of the fitness function; finally, when the number of iterations of the entire process reaches the preset maximum number of iterations, it is determined that the termination condition is met, and the tuning is completed. The position vector after the iteration is the optimal hyperparameter after optimization. At this time, the output result is the output of the target position vector parameters calculated for each iteration and the optimal hyperparameter after optimization. The target position vector parameters are the parameters of a position vector of size 1*dim represented by the following hawk group position and prey position.

[0082] Among them, the iterative calculation process first performs fitness calculation to update the prey position and the optimal fitness value, and determines whether to enter the exploration phase or the development phase according to the absolute value of the energy factor E of the prey's escape energy. When |E|>1, the algorithm enters the exploration phase. In the exploration phase, the Harris Eagle algorithm selects a search method to search for prey according to the size of the random number q(0,1) to update the prey position. When E|￡1, the algorithm enters the development phase. In the development phase, the Harris Eagle algorithm determines whether the prey has successfully escaped according to the energy factor E of the prey's escape energy, and executes four capture strategies to hunt the prey according to the energy factor E of the prey's escape energy and the size of the random number q, and updates the prey position.

[0083] The positions of the hawk flock and the prey are both represented as a position vector of size 1*dim, where dim represents the dimension, and the size of the dimension represents the number of hyperparameters that need to be optimized. The position vector is updated in each iteration process, and the position vector is formed by the hyperparameters. The position vector is represented as:

[0084] X(t)=(l[0],l[1],l[2])

[0085] Where X(t) represents the position vector at the tth iteration, t represents the tth iteration, l[0] represents the maximum number of trees, l[1] represents the learning rate, and l[2] represents the maximum depth.

[0086] The energy factor E, which represents the prey's escape energy, is used to adjust the development and exploration phases of the Harris Hawk algorithm, and its expression is as follows:

[0087]

[0088] Among them, t represents the current iteration number, T max Represents the maximum number of iterations, and E0 is a random number between (-1, 1).

[0089] The calculation expressions of the two search methods in the search phase are as follows:

[0090]

[0091] Where t+1 represents the next iteration number of the current iteration number t, X(t+1) represents the position vector at the t+1th iteration, and r1, r2, r3 and r4 are random numbers in the range of (0,1). rand (t), X m (t), X rabbit (t) are the position of a randomly selected eagle, the average position of all eagles in the current eagle group, and the position of the target prey. LB and UB represent the lower and upper limits of each optimization parameter, respectively.

[0092] The calculation expressions of the four round-up strategies in the development stage are as follows:

[0093] The first one: soft siege strategy, at this time, q ≥ 0.5 & |E| ≥ 0.5:

[0094] X(t+1)=X rabbit (t)-X(t)-E|JX rabbit (t)-X(t)|

[0095] J=2(1-r5)

[0096] The second type: hard siege strategy, at this time, q ≥ 0.5 & |E| < 0.5:

[0097] X(t+1)=X rabbit (t)-E|X rabbit (t)-X(t)|

[0098] The third type: progressive rapid dive soft encirclement strategy, at this time, q<0.5&|E|≥0.5:

[0099] V=X rabbit (t)-E|JX rabbit (t)-X(t)|

[0100] Z=V+S·LN(D)

[0101]

[0102] The fourth type: progressive rapid dive hard siege strategy, at this time, q<0.5&|E|<0.5:

[0103] P=X rabbit (t)-E|JX rabbit (t)-X m (t)|

[0104] Q=P+S·LF(D)

[0105]

[0106] Among them, r5 is a random number between 0 and 1, J is a disturbance factor between 0 and 2, V and Z are two ways to calculate the target position vector parameters under the progressive rapid dive soft encirclement strategy, P and Q represent two ways to calculate the target position vector parameters under the progressive rapid dive hard encirclement strategy, D is the dimension of the data, S is a random number with 1 row and D columns, LF is the Lévy flight function, N is the fitness function, N = N(Z), N(V), N(Q) or N(P) = RMSE.

[0107] The fitness function is expressed as follows:

[0108] Among them, z represents the zth sample, Y z，true represents the sample monitoring value of the zth sample, Y z，pred represents the model inversion prediction value of the zth sample, and m represents the number of samples. The sample monitoring value is the radioactive air concentration monitoring value.

[0109] When the optimal hyperparameters of the CatBoost ensemble learning model after the hyperparameter optimization are obtained through model training by the training group, the model training ends. At this time, the optimal hyperparameters of the CatBoost ensemble learning model after the hyperparameter optimization can be verified by the test group. The optimal hyperparameters can be: the maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are 1500, 0.05, and 10 respectively.

[0110] S500, using the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model, including:

[0111] The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the ensemble tree model based on Bagging and Boosting ideas is used to compare the CatBoost ensemble learning model after hyperparameter optimization to evaluate the model performance of the nuclear accident source term inversion; the test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

[0112] In the embodiments of the present application, the CatBoost ensemble learning model after hyperparameter optimization is also called the Catboost nuclear accident source term inversion model, the machine learning model for nuclear accident source term inversion based on the CatBoost ensemble learning algorithm, and the source term inversion model based on machine learning.

[0113] The hyperparameter-optimized CatBoost ensemble learning model was used to perform a nuclear accident source term inversion test, and the results were analyzed and compared with those of three typical ensemble tree models. Specifically, the hyperparameter-optimized CatBoost ensemble learning model was tested and analyzed using the test set, and the determination coefficient R was used to determine the optimal model. 2 The model performance is comprehensively evaluated by using the mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). Random forest RF, XGBoost, and LightGBM, three typical integrated tree models based on bagging and boosting ideas, are used to compare the model performance evaluation of the CatBoost integrated learning model after hyperparameter optimization. As shown in Table 2.

[0114] Table 2 Comparison results of model inversion performance

[0115]

[0116]

[0117] The comparison results of the inversion prediction error of the CatBoost ensemble learning model after hyperparameter optimization based on 4000 samples are as follows: Figure 4 As shown in Figure 2, the inversion prediction error of the CatBoost ensemble learning model after hyperparameter optimization is only 4.25%.

[0118] In the embodiment of the present application, the relative importance of features is calculated using the feature importance calculation method of the CatBoost ensemble learning model after hyperparameter optimization, and the SHAP calculation method is used to output the key hyperparameter feature importance, the main effects of the top six features ranked in importance, and the SHAP interaction values ​​between key meteorological parameters and radioactivity monitoring data input into the CatBoost ensemble learning model after hyperparameter optimization, so as to achieve a comparison of the two interpretability of model-related explanation and model-post-explanation. The calculation result of the relative importance of features is shown in Figure 5 shown.

[0119] The SHAP calculation method relies on the Shapley value to quantify the importance of the key hyperparameter features of the CatBoost ensemble learning model input after the hyperparameter optimization. The TreeShap interpreter based on the ensemble tree model is used to complete the interpretability calculation of the model. The calculation method of the Shapley value is as follows:

[0120]

[0121] Among them, i represents the i-th feature, Φ i is the Shapely value of the i-th feature, F represents the feature sorting set containing all features, |F|! represents the factorial of the number of features contained in F, S represents all feature subsets that do not contain feature i, |S|! represents the factorial of the number of features contained in S, and x S represents the input eigenvalue in S, f S∪{i} represents the model trained with feature i, x S∪{i} represents the input feature value in all feature subsets containing feature i, f S Represents a model without feature training.

[0122] SHAP calculation is used to further explain the relationship between the input feature variables of the CatBoost ensemble learning model after hyperparameter optimization and the output target signal. The calculation results are as follows: Figure 6 As shown. The radioactive air concentration has the greatest impact on the output of the CatBoost ensemble learning model after the hyperparameter optimization; among the key meteorological parameter features, the wind speed has the greatest impact on the output of the CatBoost ensemble learning model after the hyperparameter optimization. The greater the wind speed, the more positive impact on the output of the CatBoost ensemble learning model after the hyperparameter optimization; the main effects of the top six features selected according to the calculation results of their own feature importance are the radioactive air concentration monitoring values, wind speed, release height, and precipitation at the wind direction monitoring points at 0.5km, 1km, and 5km distances in the wind direction. The radioactive monitoring data includes radioactive air concentration monitoring values.

[0123] Since atmospheric stability is a categorical feature, one-hot encoding is generally selected for processing, so it is not considered in the interpretability analysis.

[0124] In the present application, the nuclear accident source term inversion method based on the CatBoost ensemble learning algorithm obtains the target signal and input characteristic variables of the nuclear accident source term inversion; performs data preprocessing on the target signal and the input characteristic variables to obtain the training set and the test set of the nuclear accident source term inversion; determines the structural parameters and parameter ranges of the CatBoost ensemble learning model of the nuclear accident source term inversion according to the training set and the test set; obtains the hyperparameter-optimized CatBoost ensemble learning model of the nuclear accident source term inversion according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model; uses the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model. The present application realizes the automatic adjustment of the structural parameters of the CatBoost ensemble learning model through the joint application of HHO-CatBoost-SHAP. Compared with the prior art, the hyperparameters that the CatBoost ensemble learning model needs to optimize are simple, which can reduce the complexity of parameter adjustment. Combining with the Harris Eagle algorithm can reduce the uncertainty and time-consuming cost of manual parameter optimization. The model's own feature importance calculation, the CatBoost ensemble learning model after hyperparameter optimization and the SHAP calculation method are combined, which has higher interpretability while meeting the accuracy of source term inversion, and is of great significance for promoting the application of machine learning methods in the field of nuclear accident source term inversion and realizing the intelligent inversion of accident source terms.

[0125] The following are embodiments of the apparatus of the present invention, which can be used to implement the embodiments of the method of the present invention. For details not disclosed in the embodiments of the apparatus of the present invention, please refer to the embodiments of the method of the present invention.

[0126] See also Figure 7 , which shows a schematic diagram of the structure of a nuclear accident source term inversion device based on CatBoost ensemble learning algorithm provided by an exemplary embodiment of the present invention. The device includes: an acquisition module 10000, a preprocessing module 20000, a parameter determination module 30000, a model optimization module 40000 and a model evaluation module 50000.

[0127] An acquisition module 10000 is used to acquire target signals and input characteristic variables for nuclear accident source term inversion;

[0128] A preprocessing module 20000 is used to perform data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion;

[0129] A parameter determination module 30000 is used to determine the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set;

[0130] A model optimization module 40000 is used to obtain the hyperparameter-optimized CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model;

[0131] The model evaluation module 50000 is used to use the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the CatBoost ensemble learning model after hyperparameter optimization.

[0132] According to a preferred embodiment, the acquisition module 10000 is specifically used for:

[0133] Acquire the release rate of I-131 nuclides after the nuclear accident occurs in the inversion of the nuclear accident source term, and use the release rate of I-131 nuclides after the nuclear accident occurs as the target signal of the inversion of the nuclear accident source term;

[0134] According to the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables of the inversion of the nuclear accident source term.

[0135] According to a preferred implementation, the parameter determination module 30000 is specifically used for:

[0136] Determine the maximum number of trees, maximum depth, and learning rate in the decision tree parameters for the nuclear accident source term inversion according to the training set and the test set, and determine the search range of each parameter by a five-fold cross validation method;

[0137] The maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the nuclear accident source term inversion;

[0138] Each parameter search range is used as the parameter range for inversion of the nuclear accident source term.

[0139] According to a preferred embodiment, the model optimization module 40000 includes:

[0140] After the hyperparameter optimization, the CatBoost ensemble learning model has optimal hyperparameters.

[0141] According to a preferred embodiment, the model evaluation module 50000 is specifically used for:

[0142] The CatBoost ensemble learning model after the hyperparameter optimization is tested and analyzed using the test set, and the CatBoost ensemble learning model after the hyperparameter optimization is compared with the ensemble tree model based on the ideas of Bagging and Boosting to evaluate the model performance of the nuclear accident source term inversion;

[0143] The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

[0144] It should be noted that the nuclear accident source term inversion device based on CatBoost ensemble learning algorithm provided in the above embodiment only uses the division of the above functional modules as an example when executing the nuclear accident source term inversion method based on CatBoost ensemble learning algorithm. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the nuclear accident source term inversion device based on CatBoost ensemble learning algorithm provided in the above embodiment and the nuclear accident source term inversion method based on CatBoost ensemble learning algorithm belong to the same concept. The implementation process is detailed in the method embodiment, which will not be repeated here.

[0145] In the present application, the nuclear accident source term inversion device based on the CatBoost ensemble learning algorithm obtains the target signal and input characteristic variables of the nuclear accident source term inversion; performs data preprocessing on the target signal and the input characteristic variables to obtain the training set and the test set of the nuclear accident source term inversion; determines the structural parameters and parameter ranges of the CatBoost ensemble learning model of the nuclear accident source term inversion according to the training set and the test set; obtains the hyperparameter-optimized CatBoost ensemble learning model of the nuclear accident source term inversion according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model; uses the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model. The present application realizes the automatic adjustment of the structural parameters of the CatBoost ensemble learning model through the joint application of HHO-CatBoost-SHAP. Compared with the prior art, the hyperparameters that the CatBoost ensemble learning model needs to optimize are simple, which can reduce the complexity of parameter adjustment. Combining with the Harris Eagle algorithm can reduce the uncertainty and time-consuming cost of manual parameter optimization. The model's own feature importance calculation, the CatBoost ensemble learning model after hyperparameter optimization and the SHAP calculation method are combined, which has higher interpretability while meeting the accuracy of source term inversion, and is of great significance for promoting the application of machine learning methods in the field of nuclear accident source term inversion and realizing the intelligent inversion of accident source terms.

[0146] The present invention also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implements the nuclear accident source term inversion method based on the CatBoost ensemble learning algorithm provided in the above-mentioned various method embodiments.

[0147] The present invention also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the nuclear accident source term inversion method based on the CatBoost ensemble learning algorithm of the above-mentioned various method embodiments.

[0148] See also Figure 8 , which is a schematic diagram of the structure of a terminal according to an embodiment of the present application. The terminal may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.

[0149] The communication bus 1002 is used to realize the connection and communication between these components.

[0150] The user interface 1003 may include a display screen (Display) and a camera (Camera), and the optional user interface 1003 may also include a standard wired interface and a wireless interface.

[0151] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0152] Among them, the processor 1001 may include one or more processing cores. The processor 1001 uses various interfaces and lines to connect various parts in the entire terminal, and executes various functions of the terminal and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 1005, and calling data stored in the memory 1005. Optionally, the processor 1001 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 1001 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 1001, and it can be implemented separately through a chip.

[0153] Among them, the memory 1005 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 1005 may optionally be at least one storage device located away from the aforementioned processor 1001. As Figure 8 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a nuclear accident source term inversion application based on the CatBoost ensemble learning algorithm.

[0154] exist Figure 8 In the terminal shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 1001 can be used to call the nuclear accident source term inversion application based on the CatBoost ensemble learning algorithm stored in the memory 1005, and specifically perform the following operations:

[0155] Obtain target signals and input characteristic variables for nuclear accident source term inversion;

[0156] Performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion;

[0157] Determining the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set;

[0158] According to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model, a hyperparameter-optimized CatBoost ensemble learning model for the nuclear accident source term inversion is obtained; wherein the hyperparameter-optimized CatBoost ensemble learning model has optimal hyperparameters;

[0159] The test set is used to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the CatBoost ensemble learning model after hyperparameter optimization.

[0160] In one embodiment, when the processor 1001 executes the method of obtaining the target signal and input characteristic variables for the nuclear accident source term inversion, the processor 1001 specifically performs the following operations:

[0161] Acquire the release rate of I-131 nuclides after the nuclear accident occurs in the inversion of the nuclear accident source term, and use the release rate of I-131 nuclides after the nuclear accident occurs as the target signal of the inversion of the nuclear accident source term;

[0162] According to the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables of the inversion of the nuclear accident source term.

[0163] In one embodiment, when the processor 1001 determines the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set, the processor 1001 specifically performs the following operations:

[0164] Determine the maximum number of trees, maximum depth, and learning rate in the decision tree parameters for the nuclear accident source term inversion according to the training set and the test set, and determine the search range of each parameter by a five-fold cross validation method;

[0165] The maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the nuclear accident source term inversion;

[0166] Each parameter search range is used as the parameter range for inversion of the nuclear accident source term.

[0167] In one embodiment, when the processor 1001 performs the model performance evaluation and interpretability analysis of the nuclear accident source term inversion using the hyperparameter-optimized CatBoost ensemble learning model using the test set, the processor 1001 specifically performs the following operations:

[0168] The CatBoost ensemble learning model after the hyperparameter optimization is tested and analyzed using the test set, and the CatBoost ensemble learning model after the hyperparameter optimization is compared with the ensemble tree model based on the ideas of Bagging and Boosting to evaluate the model performance of the nuclear accident source term inversion;

[0169] The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

[0170] In the present application, the method and device for nuclear accident source term inversion based on the CatBoost ensemble learning algorithm obtain the target signal and input characteristic variables for the nuclear accident source term inversion; perform data preprocessing on the target signal and the input characteristic variables to obtain the training set and test set for the nuclear accident source term inversion; determine the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion based on the training set and the test set; obtain the hyperparameter-optimized CatBoost ensemble learning model for the nuclear accident source term inversion based on the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model; use the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model. The present application realizes the automatic adjustment of the structural parameters of the CatBoost ensemble learning model through the joint application of HHO-CatBoost-SHAP. Compared with the prior art, the hyperparameters that the CatBoost ensemble learning model needs to optimize are simple, which can reduce the complexity of parameter adjustment. Combining with the Harris Eagle algorithm can reduce the uncertainty and time-consuming cost of manual parameter optimization. The model's own feature importance calculation, the CatBoost ensemble learning model after hyperparameter optimization and the SHAP calculation method are combined, which has higher interpretability while meeting the accuracy of source term inversion, and is of great significance for promoting the application of machine learning methods in the field of nuclear accident source term inversion and realizing the intelligent inversion of accident source terms.

[0171] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0172] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A method for inverting nuclear accident source terms based on CatBoost ensemble learning algorithm, characterized in that: The following steps are involved: Obtain target signals and input characteristic variables for nuclear accident source term inversion; Performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion; Determining the structural parameters and parameter ranges of the CatBoost ensemble learning model for nuclear accident source term inversion according to the training set and the test set; According to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model, a hyperparameter-optimized CatBoost ensemble learning model for the nuclear accident source term inversion is obtained; The test set is used to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the CatBoost ensemble learning model after hyperparameter optimization.

2. The method for inverting nuclear accident source terms based on CatBoost ensemble learning algorithm according to claim 1, characterized in that: The method of obtaining the target signal and input characteristic variables for nuclear accident source term inversion includes: Acquire the release rate of I-131 nuclides after the nuclear accident occurs in the inversion of the nuclear accident source term, and use the release rate of I-131 nuclides after the nuclear accident occurs as the target signal of the inversion of the nuclear accident source term; According to the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables of the inversion of the nuclear accident source term.

3. The method for inverting nuclear accident source terms based on CatBoost ensemble learning algorithm according to claim 1, characterized in that: Determining the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set includes: Determine the maximum number of trees, maximum depth, and learning rate in the decision tree parameters for the nuclear accident source term inversion according to the training set and the test set, and determine the search range of each parameter by a five-fold cross validation method; The maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the nuclear accident source term inversion; Each parameter search range is used as the parameter range for inversion of the nuclear accident source term.

4. The method for inverting nuclear accident source terms based on CatBoost ensemble learning algorithm according to claim 1, characterized in that: The method comprises: obtaining the hyperparameter-optimized CatBoost ensemble learning model for nuclear accident source term inversion based on the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model, including: After the hyperparameter optimization, the CatBoost ensemble learning model has optimal hyperparameters.

5. The method for inverting nuclear accident source terms based on CatBoost ensemble learning algorithm according to claim 1, characterized in that: The use of the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the hyperparameter-optimized CatBoost ensemble learning model includes: The CatBoost ensemble learning model after the hyperparameter optimization is tested and analyzed using the test set, and the CatBoost ensemble learning model after the hyperparameter optimization is compared with the ensemble tree model based on the ideas of Bagging and Boosting to evaluate the model performance of the nuclear accident source term inversion; The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

6. A nuclear accident source term inversion device based on CatBoost ensemble learning algorithm, characterized in that: include: An acquisition module is used to obtain target signals and input characteristic variables for nuclear accident source term inversion; A preprocessing module, used for performing data preprocessing on the target signal and the input characteristic variable to obtain a training set and a test set for the nuclear accident source term inversion; A parameter determination module, used to determine the structural parameters and parameter ranges of the CatBoost ensemble learning model for the nuclear accident source term inversion according to the training set and the test set; A model optimization module, for obtaining a hyperparameter-optimized CatBoost ensemble learning model for inversion of nuclear accident source terms according to the training set, the structural parameters and parameter ranges of the CatBoost ensemble learning model, the Harris Eagle algorithm and the CatBoost ensemble learning model; A model evaluation module is used to use the test set to perform model performance evaluation and interpretability analysis of the nuclear accident source term inversion on the CatBoost ensemble learning model after hyperparameter optimization.

7. The nuclear accident source term inversion device based on CatBoost ensemble learning algorithm according to claim 6, characterized in that: The acquisition module is specifically used for: Acquire the release rate of I-131 nuclides after the nuclear accident occurs in the inversion of the nuclear accident source term, and use the release rate of I-131 nuclides after the nuclear accident occurs as the target signal of the inversion of the nuclear accident source term; According to the nuclear accident monitoring data and the Gaussian plume atmospheric diffusion model, the inversion factors affecting the source term inversion of the nuclear accident source term are obtained, and the inversion factors affecting the source term are used as the input characteristic variables of the inversion of the nuclear accident source term.

8. The nuclear accident source term inversion device based on CatBoost ensemble learning algorithm according to claim 6, characterized in that: The parameter determination module is specifically used for: Determine the maximum number of trees, maximum depth, and learning rate in the decision tree parameters for the nuclear accident source term inversion according to the training set and the test set, and determine the search range of each parameter by a five-fold cross validation method; The maximum number of trees, the maximum depth, and the learning rate in the decision tree parameters are used as the structural parameters of the CatBoost ensemble learning model for the nuclear accident source term inversion; Each parameter search range is used as the parameter range for inversion of the nuclear accident source term.

9. The nuclear accident source term inversion device based on CatBoost ensemble learning algorithm according to claim 6, characterized in that: The model optimization module comprises: After the hyperparameter optimization, the CatBoost ensemble learning model has optimal hyperparameters.

10. The nuclear accident source term inversion device based on CatBoost ensemble learning algorithm according to claim 6, characterized in that: The model evaluation module is specifically used for: The CatBoost ensemble learning model after the hyperparameter optimization is tested and analyzed using the test set, and the CatBoost ensemble learning model after the hyperparameter optimization is compared with the ensemble tree model based on the ideas of Bagging and Boosting to evaluate the model performance of the nuclear accident source term inversion; The test set is used to test and analyze the CatBoost ensemble learning model after hyperparameter optimization, and the interpretability analysis of the nuclear accident source term inversion is performed using its own feature importance calculation and SHAP calculation.

Images