Reactor core thermal parameter prediction method based on improved GAN method and combined ANN model

Abstract

The invention belongs to the technical field of nuclear engineering, and particularly relates to a reactor core thermal parameter prediction method based on an improved GAN method and a combined ANN model. Determining the reactor core thermal parameters as input data, expanding samples of the reactor core thermal parameters, and preprocessing the input data; a thermal parameter prediction network model is established, the network model is an MLP-CNN-RNN combined neural network model, and the MLP-CNN-RNN combined neural network model is used for predicting the thermal parameters of the nuclear reactor; training and optimizing the established network model, and verifying; and predicting the thermal parameters of the nuclear reactor by using the verified network model to obtain a predicted value. Accurate prediction of the thermal parameters of the nuclear reactor is realized. The model combines the advantages of a generative model and a deep learning model, the limitation of a traditional method can be effectively overcome, and the prediction precision and performance are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of nuclear engineering, and in particular relates to a method for predicting core thermal parameters based on an improved GAN method and a combined ANN model. Background Art

[0002] During the operation of a nuclear power plant, it is necessary to monitor and predict important thermal parameters such as reactor power and core thermal conductivity in real time to ensure the safe and efficient operation of the nuclear reactor. 2 The thermal conductivity of the pellet is UO 2 It is an important parameter to be considered in component design. It determines the operating temperature, temperature distribution, thermal stress of the core and a series of important phenomena caused by it, such as core swelling, fission gas release, core and cladding wall stress and interaction, etc. Therefore, the research on core thermal conductivity has received much attention.

[0003] At present, the measured value of high-temperature thermal conductivity is not easy to obtain, especially the data of high-temperature thermal conductivity in the pile is relatively small. Because most experiments to measure thermal parameters are carried out under high temperature and high pressure conditions, the operation is complicated and the experimental cycle is long. The traditional calculation method based on empirical formula is not accurate enough and is not easy to adapt to changes in actual conditions. Therefore, a data-driven prediction method is combined with the prediction of core thermal conductivity to provide design data for in-depth exploration of high-temperature thermal conductivity mechanisms. The data-driven prediction method is based on existing data and establishes a nonlinear mapping relationship between data and thermal parameters through machine learning algorithms and neural network models to achieve the prediction of unknown data.

[0004] Neural network models have made significant progress in the field of prediction and modeling in recent years. They have powerful nonlinear modeling and data processing capabilities. Generative adversarial networks (GANs) are powerful generative models that can generate realistic data samples through adversarial training, but traditional GAN ​​methods have problems such as gradient disappearance and mode collapse. Summary of the invention

[0005] The present invention provides a method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, which realizes accurate prediction of thermal parameters of nuclear reactors. This model combines the advantages of generative models and deep learning models, can effectively overcome the limitations of traditional methods, and improve prediction accuracy and performance.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, the method for predicting core thermal parameters comprising:

[0008] Determine the core thermal parameters as input data, expand the sample of the core thermal parameters, and pre-process the input data;

[0009] Establishing a thermal parameter prediction network model, wherein the network model is an MLP-CNN-RNN combined neural network model, and the MLP-CNN-RNN combined neural network model is used for predicting thermal parameters of a nuclear reactor;

[0010] Train, optimize and verify the established network model;

[0011] The verified network model is used to predict the thermal parameters of the nuclear reactor and obtain the predicted values.

[0012] First, 17 sets of experimental data were collected through literature review, and the improved GAN was used to expand the data to generate additional data samples, expanding the data set to 2000 sets. Such data expansion can increase the number of samples and improve the training effect and prediction accuracy of the model.

[0013] Secondly, in order to better process the data, the present invention introduces a data preprocessing module, which performs preprocessing operations such as data cleaning and normalization on the experimental data and the generated data samples to ensure the quality and consistency of the data and provide accurate input data for the subsequent neural network model.

[0014] Then, the present invention constructs an MLP-CNN-RNN combined neural network model, which is composed of a multi-layer perceptron (MLP) module, a convolutional neural network (CNN) module and a recurrent neural network (RNN) module. These modules achieve accurate prediction of nuclear reactor thermal parameters through layer-by-layer connection and information transmission. The MLP module is used to process input features, the CNN module is used to extract spatial information of features, and the RNN module is used to process time series information of features, thereby fully mining the potential laws and relevance of data.

[0015] Finally, by comparing and analyzing the prediction results of GAN and MLP-CNN-RNN models with real experimental data, the advantages of the proposed method in predicting nuclear reactor thermal parameters are verified. The experimental results show that the prediction accuracy of the model can be improved by using improved GAN to expand data, and the MLP-CNN-RNN combined neural network model has good generalization ability and prediction performance.

[0016] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the sample of the expanded core thermal parameters is specifically expanded by using an improved GAN method;

[0017] The improved GAN method specifically includes changing the maximum singular value of the discriminator weight matrix, spectrally normalizing the weights of each convolutional layer or fully connected layer of the discriminator before updating the discriminator network parameters each time; performing singular value decomposition on the weight matrix, and then dividing each singular value by the maximum singular value so that the spectral norm of the weight matrix does not exceed a given threshold.

[0018] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the improved GAN method comprises the following steps:

[0019] Step S1: Calculate the singular value decomposition of the matrix W;

[0020] Step S2: Calculate the spectral norm sigma_max of the matrix W;

[0021] Step S3: Normalize the matrix W, that is, divide each element in the matrix W by the maximum singular value sigma_max to ensure that the spectral norm of the matrix W does not exceed 1.

[0022] Through this spectral normalization process, the range of the discriminator network weights can be limited, enhancing the stability of the network.

[0023] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the MLP-CNN-RNN combined neural network model includes an MLP model, a CNN model and an RNN model;

[0024] The MLP model includes an input layer, a hidden layer and an output layer;

[0025] The input layer receives the original feature data, and the hidden layer transforms and extracts the input features through a nonlinear activation function to extract high-level feature representation;

[0026] The connections between hidden layers are realized through weight matrices to combine and weight features;

[0027] The output layer maps the output of the hidden layer to the final prediction result through the activation function.

[0028] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the CNN model includes a convolutional layer, a pooling layer, and a fully connected layer;

[0029] The convolution layer extracts features from the input data through a sliding window convolution operation and captures different local features through the convolution kernel;

[0030] The pooling layer reduces the size of the feature matrix through downsampling operations, retains the most significant features, and reduces the number of parameters and computational complexity;

[0031] The fully connected layer maps the output of the pooling layer to the final prediction result through the weight matrix.

[0032] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the RNN model includes an input layer, a circulation layer and an output layer;

[0033] The recurrent layers are connected through feedback in time to capture the correlation and long-term dependencies between stack power and other parameters when processing sequence data.

[0034] The input layer receives sequence data, and the recurrent layer combines the input of the current time step with the hidden state of the previous time step through the recurrent unit to update the hidden state.

[0035] The output layer maps the output of the loop layer to the final prediction result through an activation function.

[0036] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model is proposed. The expanded sample data is input into the input layer of the MLP model. After passing through the hidden layer, the predicted data is sent to the convolution layer of the CNN module through the output layer. After passing through the pooling layer, the predicted data is output to the input layer of the RNN module through the fully connected layer. Finally, after the warning loop layer, the final predicted data is sent to the output layer to obtain the final prediction result.

[0037] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the training and optimization of the network model are specifically as follows:

[0038] Construct loss function;

[0039] An L2 regularization term is added to the loss function to form a new loss function, the new loss function is optimized, and the weight parameters of the model are adjusted by minimizing the new loss function to reduce the total loss of the original loss function and the L2 regularization term.

[0040] An electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0041] Memory, used to store computer programs;

[0042] The processor is used to implement the above method steps when executing the program stored in the memory.

[0043] A computer-readable storage medium stores a computer program, and the computer program implements the above method steps when executed by a processor.

[0044] The beneficial effects of the present invention are:

[0045] The invention realizes accurate prediction of thermal parameters of nuclear reactor and has high prediction accuracy and generalization ability.

[0046] The invention can improve the prediction accuracy and generalization ability of thermal parameters of nuclear reactors.

[0047] This invention improves the prediction accuracy of nuclear reactor thermal parameters: By using an improved generative adversarial network to expand experimental data, researchers have successfully improved the accuracy of the nuclear reactor thermal parameter prediction model.

[0048] The present invention improves GAN by generating more data samples to help the model better understand and capture the distribution characteristics of the data, thereby improving the accuracy of the prediction.

[0049] This paper verifies the advantages of the proposed method: By comparing and analyzing the prediction results of the generative adversarial network and the MLP-CNN-RNN model, this study verifies the advantages of the proposed method in predicting nuclear reactor thermal parameters. This shows that the combined neural network model used performs well in capturing the complex correlation of input data and predicting target parameters.

[0050] The present invention has good generalization ability and prediction performance: the prediction results show that the prediction accuracy of the model can be improved by using the improved GAN to expand the data. In addition, the MLP-CNN-RNN combined neural network model also shows good generalization ability, that is, it can also show good prediction performance on unseen data.

[0051] The present invention constructs a MLP-CNN-RNN combined neural network model: This study uses a multi-layer perceptron-convolutional neural network-recurrent neural network (MLP-CNN-RNN) combined neural network model to predict nuclear reactor thermal parameters. The model combines different types of neural network layers and makes full use of their advantages in feature extraction, spatial modeling, and sequence modeling to improve prediction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of the method of the present invention.

[0053] Figure 2 It is the mean square error training graph of the present invention.

[0054] Figure 3 It is the mean square error regression graph of the present invention.

[0055] Figure 4 It is the mean square error fitting diagram of the present invention. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the 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.

[0057] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, the method for predicting core thermal parameters includes:

[0058] Determine the core thermal parameters as input data, expand the samples of the core thermal parameters, generate additional data samples by training and optimizing the collected experimental data, expand the data set to 2000 groups, and preprocess the input data, the preprocessing is used to preprocess the experimental data and the generated data samples, including data cleaning, normalization and other operations;

[0059] Establishing a thermal parameter prediction network model, wherein the network model is an MLP-CNN-RNN combined neural network model, and the MLP-CNN-RNN combined neural network model is used for predicting thermal parameters of a nuclear reactor;

[0060] Train and optimize the established network model to train and adjust the parameters of the MLP-CNN-RNN combined neural network model to improve the prediction accuracy and generalization ability of the model;

[0061] The verified network model is used to predict the thermal parameters of the nuclear reactor and obtain the predicted values.

[0062] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the sample of the expanded core thermal parameters is specifically expanded by using an improved GAN method;

[0063] The improved GAN method specifically includes changing the maximum singular value of the discriminator weight matrix, spectrally normalizing the weights of each convolutional layer or fully connected layer of the discriminator before updating the discriminator network parameters each time; performing singular value decomposition on the weight matrix, and then dividing each singular value by the maximum singular value so that the spectral norm of the weight matrix does not exceed a given threshold.

[0064] The GAN model uses an improved generator network and discriminator network. The traditional GAN ​​method has problems such as gradient disappearance and mode collapse. Therefore, this project uses methods such as changing the loss function, gradient penalty and spectral normalization to improve the GAN method to generate more experimental data, so as to improve the quality and accuracy of the data and thus improve the effect of the prediction model.

[0065] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the improved GAN method comprises the following steps:

[0066] Step S1: Calculate the singular value decomposition of the matrix W;

[0067] W=UΣV^T

[0068] Where U is an orthogonal matrix, Σ is a diagonal matrix, and V^T is the transposed matrix of V;

[0069] Step S2: Calculate the spectral norm sigma_max of the matrix W;

[0070] sigma_max=max(Σ)

[0071] Where max(Σ) represents the maximum singular value in Σ;

[0072] Step S3: normalize the matrix W, that is, divide each element in the matrix W by the maximum singular value sigma_max to ensure that the spectral norm of the matrix W does not exceed 1;

[0073] W_normalized = W / sigma_max.

[0074] Through this spectral normalization process, the range of the discriminator network weights can be limited, and the stability of the network can be enhanced. Compared with ordinary GAN models, the data of thermal neutron fission heat generation, fast neutron fission heat generation, reactor gamma field heat generation, core thermal conductivity, fuel temperature at different positions, and reactor power can be generated in the table data.

[0075] Table 1 Original data

[0076]

[0077] Table 2. Training dataset of improved GAN method

[0078]

[0079] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the MLP-CNN-RNN combined neural network model comprises an MLP model, a CNN model and an RNN model; the MLP-CNN-RNN combined neural network model adopts a combination of a multi-layer perceptron (MLP) module, a convolutional neural network (CNN) module and a recurrent neural network (RNN) module, and realizes accurate prediction of thermal parameters of a nuclear reactor through layer-by-layer connection and information transmission.

[0080] The MLP model includes an input layer, a hidden layer and an output layer;

[0081] Input layer: receives input data, each input corresponds to a feature.

[0082] Hidden Layers: There are multiple hidden layers and each hidden layer consists of multiple neurons.

[0083] Output layer: The last layer, usually one or more neurons, each output neuron corresponds to a target value to be predicted;

[0084] The input layer receives the original feature data, and the hidden layer transforms and extracts the input features through a nonlinear activation function to extract high-level feature representation;

[0085] The connections between hidden layers are realized through weight matrices to combine and weight features;

[0086] The output layer maps the output of the hidden layer to the final prediction result through the activation function.

[0087] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the CNN model includes a convolutional layer, a pooling layer, and a fully connected layer;

[0088] Convolutional layer: Extracts spatial features of input data by applying convolution operations, using multiple convolution kernels to capture different features.

[0089] Pooling layer: Reduces the size of the feature matrix through downsampling operations and retains the most significant features.

[0090] Fully connected layers: The output of the pooling layer is connected to one or more fully connected layers to further process the features and generate prediction results.

[0091] The convolution layer extracts features from the input data through a sliding window convolution operation and captures different local features through the convolution kernel;

[0092] The pooling layer reduces the size of the feature matrix through downsampling operations, retains the most significant features, and reduces the number of parameters and computational complexity;

[0093] The fully connected layer maps the output of the pooling layer to the final prediction result through the weight matrix.

[0094] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the RNN model includes an input layer, a circulation layer and an output layer;

[0095] Input layer: receives the time series data of input data. Each time series corresponds to one input. Here, the stack power data can be regarded as sequence data.

[0096] Recurrent layer: It consists of one or more recurrent units and models sequence data by combining the current input with the previous hidden state.

[0097] Output layer: The last layer, usually one or more neurons, each output neuron corresponds to a target value to be predicted.

[0098] The recurrent layers are connected through feedback in time to capture the correlation and long-term dependencies between stack power and other parameters when processing sequence data.

[0099] The input layer receives sequence data, and the recurrent layer combines the input of the current time step with the hidden state of the previous time step through the recurrent unit to update the hidden state.

[0100] The output layer maps the output of the loop layer to the final prediction result through an activation function.

[0101] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model is proposed. The expanded sample data is input into the input layer of the MLP model. After passing through the hidden layer, the predicted data is sent to the convolution layer of the CNN module through the output layer. After passing through the pooling layer, the predicted data is output to the input layer of the RNN module through the fully connected layer. Finally, after the warning loop layer, the final predicted data is sent to the output layer to obtain the final prediction result.

[0102] During the prediction process, the model inputs the characteristic vector of the nuclear reactor thermal parameters to be predicted, and obtains the corresponding prediction results through forward propagation calculation of the neural network.

[0103] It can be applied to the prediction of thermal parameters of nuclear reactors in the field of nuclear engineering, providing accurate parameter prediction and technical support for nuclear engineers.

[0104] By comparing and analyzing the prediction results of GAN and MLP-CNN-RNN models, the advantages and effectiveness of the proposed method in predicting nuclear reactor thermal parameters are verified. Figure 3-4 shown.

[0105] A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, wherein the training and optimization of the network model are specifically as follows:

[0106] Construct the loss function; define the original loss function according to the structure and task requirements of the MLP-CNN-RNN combined neural network model, assuming that the original loss function is L;

[0107] Adding L2 regularization term to the loss function forms a new loss function. L2 regularization is used to reduce the risk of overfitting of neural network models. The role is to add regularization term to the loss function to penalize the size of the model's weight parameters to limit their growth, thereby improving the generalization ability of the model.

[0108] L2 regularization adds a regularization term to the loss function. This term is the sum of the squares of the weight parameters multiplied by a regularization coefficient λ. The goal of the regularization term is to keep the weight parameters small to avoid overfitting of the model caused by excessive weights. The formula is as follows:

[0109] Loss function = original loss function + λ*sum of squares of weight parameters

[0110] Among them, the original loss function is the loss of the model on the training data, and the sum of squares of weight parameters refers to the sum of squares of all weight parameters of the model.

[0111] During training, gradient descent updates the weight parameters by minimizing the entire loss function. When the model's weight parameters are large, the regularization term comes into play, making the regularization term in the loss function larger, forcing the optimization algorithm to adjust the weight parameters to smaller values. This helps avoid overfitting and improves the model's ability to generalize to unseen data.

[0112] Specifically, the L2 regularization term is added to the loss function of the model, and then the gradient descent method is used to minimize the entire loss function. The strength of regularization can be controlled by adjusting the value of the regularization coefficient λ. A larger λ value imposes a greater penalty on the size of the weight parameter, making the model more inclined to choose a smaller weight value. L2 regularization helps reduce the overfitting phenomenon of the neural network model by limiting the size of the weight parameter, and improves the generalization ability and prediction performance of the model.

[0113] Optimize the new loss function and adjust the weight parameters of the model by minimizing the new loss function to reduce the total loss of the original loss function and the L2 regularization term.

[0114] Add L2 regularization term: Add L2 regularization term to the original loss function L to construct a new loss function. L2 regularization term is the sum of squares of weight parameters multiplied by a regularization coefficient λ. Let weight parameter be W, L2 regularization term be R, and new loss function be L'. Then L'=L+λ*||W||^2, where ||W||^2 represents the sum of squares of weight parameters.

[0115] Define the optimization goal: Take L' as the optimization goal,

[0116] Weight update rule: According to the gradient descent method, the weight update rule is used to update the weight parameters of the model. Usually, in each training iteration step, the value of the weight parameter is updated by calculating the gradient of the loss function with respect to the weight parameter and the derivative of the regularization term with respect to the weight parameter.

[0117] Adjustment of the regularization coefficient λ: The selection of the regularization coefficient λ is a hyperparameter tuning process. By trying different λ values ​​and observing the performance of the model on the validation set, a λ value that can prevent overfitting while maintaining good prediction performance is selected.

[0118] Through the above steps, the L2 regularization method can play a role in training the MLP-CNN-RNN combined neural network model. By controlling the size of the weight parameter, L2 regularization can reduce the overfitting risk of the model and improve the prediction accuracy and generalization ability of nuclear reactor thermal parameters.

[0119] Model evaluation: predict the 18th item of the original data set, input the reactor power of the 18th set of data, and compare the result with the true value. The relative error is shown in the following table. The relative error is small, indicating that the model prediction effect is good.

[0120]

[0121] Embodiment 2

[0122] The embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the memory is used to store software programs and modules, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory and the processor are connected via a bus. Specifically, the processor implements any of the steps of the above method by running the computer program stored in the memory.

[0123] It should be understood that in the embodiments of the present application, the processor referred to may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0124] The memory may include a read-only memory, a flash memory, and a random access memory, and provides instructions and data to the processor. A part or all of the memory may also include a nonvolatile random access memory.

[0125] It should be understood that the above-mentioned integrated module / unit, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the above-mentioned computer program includes computer program code, and the above-mentioned computer program code can be in source code form, object code form, executable file or some intermediate form. The above-mentioned computer-readable medium may include: any entity or device capable of carrying the above-mentioned computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the above-mentioned computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0126] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0127] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the above-mentioned device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0128] It should be noted that the methods and detailed examples provided in the above embodiments can be combined with the devices and equipment provided in the embodiments, and references can be made to each other, and no further details will be given.

[0129] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0130] In the embodiments provided in the present application, it should be understood that the disclosed apparatus / terminal equipment and method can be implemented in other ways. For example, the apparatus / equipment embodiments described above are only schematic, for example, the division of the above modules or units is only a logical function division, and in actual implementation, other division methods can be used, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0131] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for predicting core thermal parameters based on an improved GAN method and a combined ANN model, characterized in that: include Determine the core thermal parameters as input data, expand the sample of the core thermal parameters, and pre-process the input data; Establishing a thermal parameter prediction network model, wherein the network model is an MLP-CNN-RNN combined neural network model, and the MLP-CNN-RNN combined neural network model is used for predicting thermal parameters of a nuclear reactor; Train, optimize and verify the established network model; The verified network model is used to predict the thermal parameters of the nuclear reactor and obtain the predicted values.

2. The method for predicting core thermal parameters according to claim 1, characterized in that: The sample of expanding the core thermal parameters is specifically expanded by using an improved GAN method; The improved GAN method specifically includes changing the maximum singular value of the discriminator weight matrix, and performing spectral normalization on the weights of each convolutional layer or fully connected layer of the discriminator before updating the discriminator network parameters each time; Perform singular value decomposition on the weight matrix and then divide each singular value by the largest singular value so that the spectral norm of the weight matrix does not exceed a given threshold.

3. The method for predicting core thermal parameters according to claim 2, characterized in that: The improved GAN method comprises the following steps: Step S1: Calculate the singular value decomposition of the matrix W; Step S2: Calculate the spectral norm sigma_max of the matrix W; Step S3: Normalize the matrix W, that is, divide each element in the matrix W by the maximum singular value sigma_max to ensure that the spectral norm of the matrix W does not exceed 1. Through this spectral normalization process, the range of the discriminator network weights can be limited, enhancing the stability of the network.

4. The method for predicting core thermal parameters according to claim 1, characterized in that: The MLP-CNN-RNN combined neural network model includes an MLP model, a CNN model and an RNN model; The MLP model includes an input layer, a hidden layer and an output layer; The input layer receives the original feature data, and the hidden layer transforms and extracts the input features through a nonlinear activation function to extract high-level feature representation; The connections between hidden layers are realized through weight matrices to combine and weight features; The output layer maps the output of the hidden layer to the final prediction result through the activation function.

5. The method for predicting core thermal parameters according to claim 4, characterized in that: The CNN model includes a convolutional layer, a pooling layer and a fully connected layer; The convolution layer extracts features from the input data through a sliding window convolution operation and captures different local features through the convolution kernel; The pooling layer reduces the size of the feature matrix through downsampling operations, retains the most significant features, and reduces the number of parameters and computational complexity; The fully connected layer maps the output of the pooling layer to the final prediction result through the weight matrix.

6. The method for predicting core thermal parameters according to claim 4, characterized in that: The RNN model includes an input layer, a recurrent layer and an output layer; The recurrent layers are connected through feedback in time to capture the correlation and long-term dependencies between stack power and other parameters when processing sequence data. The input layer receives sequence data, and the recurrent layer combines the input of the current time step with the hidden state of the previous time step through the recurrent unit to update the hidden state. The output layer maps the output of the loop layer to the final prediction result through an activation function.

7. The method for predicting core thermal parameters according to claim 5 or 6, characterized in that: The expanded sample data is input into the input layer of the MLP model. After passing through the hidden layer, the predicted data is sent to the convolution layer of the CNN module through the output layer. After passing through the pooling layer, the predicted data is output to the input layer of the RNN module through the fully connected layer. Finally, after the warning loop layer, the final predicted data is sent to the output layer to obtain the final prediction result.

8. The method for predicting core thermal parameters according to claim 1, characterized in that: The training and optimization of the network model is specifically as follows: Construct loss function; An L2 regularization term is added to the loss function to form a new loss function, the new loss function is optimized, and the weight parameters of the model are adjusted by minimizing the new loss function to reduce the total loss of the original loss function and the L2 regularization term.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 8 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any one of claims 1 to 8 are implemented.

Images