Neural Network-Based Neutron Solving Method and System for Hybrid Energy Spectrum Reactors

Abstract

The present invention provides a method and system for solving neutrons in a hybrid energy spectrum reactor based on a neural network, which relates to the technical field of hybrid energy spectrum reactors. The method includes: obtaining spatio-temporal characteristic data of materials in different energy spectrum regions; inputting the spatio-temporal characteristic data into a pre-trained deep learning model to predict the coupling response data and neutron diffusion mode between the thermal spectrum region and the fast spectrum region. The embodiments of the present invention utilize a deep learning model combined with two deep neural networks, breaking through the limitations of traditional numerical simulation methods, having strong spatio-temporal characteristic capture ability, non-linear mapping ability and generalization ability, showing significant technical effects in accurately predicting the coupling response and neutron diffusion mode in a hybrid energy spectrum reactor, effectively reducing the computational complexity and experimental dependence, and enhancing the universality and real-time prediction ability of the application.

Description

Technical Field

[0001] The present invention relates to the technical field of hybrid energy spectrum reactors, and more particularly, to a method and system for solving neutrons in a hybrid energy spectrum reactor based on a neural network. Background Art

[0002] With the continuous development of nuclear energy technology, hybrid energy spectrum reactors have gradually become a key research direction. Such reactors have the advantages of thermal neutrons and fast neutron regions. However, in actual operation, there are complex neutron correlation relationships between the thermal spectrum region and the fast spectrum region, resulting in uneven power distribution and uneven fuel utilization, and the thermodynamic performance of the reactor core is restricted. In addition, in the energy spectrum mixing reaction of the reactor, the diffusion laws of materials (such as fissile materials and breeding materials) also directly affect the safety and efficiency of the reactor reaction.

[0003] Most of the existing technologies use numerical methods to solve hybrid energy spectrum reactors, with high computational complexity and difficult to predict in real time. Especially for Monte Carlo simulation, although it can accurately simulate the coupling response in the reactor, due to the need to handle complex equations in multi-dimensional space, the computational amount is extremely large and it is difficult to perform real-time prediction. For the actual reactor design and optimization requirements, this high computational complexity leads to a long calculation cycle and cannot quickly respond to design adjustments or parameter optimizations. The state changes during the reactor operation are dynamic, and the existing numerical simulations require a large amount of repeated calculations, which are time-consuming and have poor real-time performance. Summary of the Invention

[0004] To solve the above problems, an embodiment of the present invention provides a method for solving neutrons in a hybrid energy spectrum reactor based on a neural network, including: obtaining spatio-temporal characteristic data of materials in different energy spectrum regions; inputting the spatio-temporal characteristic data into a pre-trained deep learning model to predict the coupling response data and neutron diffusion patterns between the thermal spectrum region and the fast spectrum region; the deep learning model includes a first deep neural network and a second deep neural network; the first deep neural network is used to extract spatial features based on the spatio-temporal characteristic data, and the second deep neural network is used to extract temporal features based on the spatial features and the spatio-temporal characteristic data to predict the coupling response data and the neutron diffusion patterns.

[0005] The method for solving neutrons in a hybrid energy spectrum reactor based on a neural network provided by the embodiment of the present invention breaks through the limitations of traditional numerical simulation methods, has strong spatio-temporal feature capture ability, non-linear mapping ability and generalization ability, shows significant technical effects in accurately predicting the coupling response and neutron diffusion patterns in a hybrid energy spectrum reactor, and effectively reduces the computational complexity and experimental dependence, and improves the universality and real-time prediction ability of the application.

[0006] Optionally, the first deep neural network is a convolutional neural network, and the second deep neural network is a long short-term memory network.

[0007] In the embodiment of the present invention, a CNN-LSTM combined model is adopted. The CNN is responsible for extracting the patterns of spatial distribution in the reactor, and the LSTM is responsible for processing the connection responses at different time points in the time series and the evolution of diffusion patterns.

[0008] Optionally, the spatio-temporal feature data includes distribution data of temperature, pressure, and neutron density in different energy spectrum regions, and data recording the change of the distribution data over time.

[0009] In the embodiment of the present invention, the model is trained and the prediction based on the model is carried out using the spatial characteristic data and time series characteristic data of the material in different energy spectrum regions, realizing the efficient capture and comprehensive modeling of spatio-temporal features, with strong generalization ability and wide adaptability.

[0010] Optionally, the method further includes: obtaining multi-dimensional physical quantity data through a reactor physics simulation tool, where the multi-dimensional physical quantity data includes: distribution data of temperature, pressure, and neutron density in the thermal spectrum region and the fast spectrum region, and data recording the change of the distribution data over time; preprocessing the multi-dimensional physical quantity data; inputting the preprocessed data into the deep learning model for training, and adjusting the model parameters based on minimizing the loss function until the loss function drops to a set threshold to obtain a trained deep learning model.

[0011] In the embodiment of the present invention, training data is obtained based on an existing simulation tool, and training is carried out based on it until the model is completed. The coupling response and diffusion pattern of any material under pseudo-critical conditions can be predicted by inputting the element characteristic curve of the material, and there is no need to reconstruct a new numerical model.

[0012] Optionally, the preprocessing of the multi-dimensional physical quantity data includes: performing standardization or normalization processing on the multi-dimensional physical quantity data; representing the processed data as a multi-dimensional spatio-temporal tensor.

[0013] In the embodiment of the present invention, the standardization or normalization processing ensures that each physical quantity is within the same numerical range, avoiding the unbalanced influence of different physical quantities on model training. Representing it as a multi-dimensional spatio-temporal tensor can separate spatio-temporal features, facilitating the extraction of spatial and temporal features respectively.

[0014] Optionally, inputting the preprocessed data into the deep learning model for training includes: inputting the spatial tensor of the multi-dimensional spatio-temporal tensor into a convolutional neural network to output spatial features; converting the spatial features into time series features, and inputting the time series features into a long short-term memory network to output coupled response data; calculating a loss function based on the coupled response data and the true response data; and ending the training of the deep learning model after iteratively optimizing until the loss function drops to a set threshold.

[0015] The embodiment of the present invention provides a specific training method for a deep learning model, which can effectively fit the coupled response of the thermal spectrum region and the fast spectrum region and the neutron diffusion mode.

[0016] Optionally, the method further includes: acquiring multi-dimensional physical quantity data accumulated during the operation of the reactor, and training and updating the deep learning model according to the accumulated multi-dimensional physical quantity data.

[0017] After the training of the network in the embodiment of the present invention, the model can process the input data in real time, is suitable for real-time monitoring and prediction of the operating state in the reactor, and can achieve online learning and continuous optimization.

[0018] Optionally, the method further includes: using the cross-validation method to divide the multi-dimensional physical quantity data into a training set, a validation set, and a test set; the validation set is used to monitor the performance of the model during training, and the test set is used to evaluate the performance of the model on unseen data.

[0019] The embodiment of the present invention defines that dividing the training data obtained by the simulation tool into a training set, a validation set, and a test set can verify and evaluate the training effect of the model.

[0020] The embodiment of the present invention provides a hybrid energy spectrum reactor neutron solution system based on a neural network, including: a data acquisition module for acquiring spatio-temporal feature data of materials in different energy spectrum regions; a prediction module for inputting the spatio-temporal feature data into a pre-trained deep learning model to predict the coupled response data between the thermal spectrum region and the fast spectrum region and the neutron diffusion mode; the deep learning model includes a first deep neural network and a second deep neural network; the first deep neural network is used to extract spatial features based on the spatio-temporal feature data, and the second deep neural network is used to extract time features based on the spatial features and the spatio-temporal feature data to predict the coupled response data and the neutron diffusion mode.

[0021] Optionally, the first deep neural network is a convolutional neural network, and the second deep neural network is a long short-term memory network.

[0022] The neutron solution system of the hybrid energy spectrum reactor based on neural network according to the embodiment of the present invention can achieve the same technical effects as the above-mentioned neutron solution method of the hybrid energy spectrum reactor based on neural network. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0024] Figure 1 It is a schematic flowchart of the neutron solution method of the hybrid energy spectrum reactor based on neural network provided by the embodiment of the present invention;

[0025] Figure 2 It is a schematic diagram of the principle of the CNN-LSTM combined model in the embodiment of the present invention;

[0026] Figure 3 It is a schematic structural diagram of the neutron solution system of the hybrid energy spectrum reactor based on neural network provided by the embodiment of the present invention. Detailed Embodiments

[0027] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will describe in detail the specific embodiments of the present invention with reference to the drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] The embodiment of the present invention utilizes the high-efficiency computing power of the deep neural network. After one-time training, it can quickly perform forward propagation after inputting the material characteristics and output the coupled response and diffusion mode prediction results in real time, greatly reducing the computing time.

[0029] In this embodiment, the deep learning model combines two deep neural networks. Exemplarily, a model that combines a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) is used. CNN is suitable for processing spatial features. For the data in a hybrid energy spectrum reactor, such as the neutron density distribution, which has significant spatial correlation, CNN can effectively capture these local spatial features. It is responsible for extracting the patterns of the spatial distribution (such as the neutron density field) in the reactor. LSTM is suitable for processing hybrid data with time dependence. For a hybrid energy spectrum reactor, the relationship response between the thermal spectrum region and the fast spectrum region not only has spatial distribution characteristics but also changes dynamically over time. LSTM can effectively capture this information. It is responsible for processing the connection response and the evolution of the diffusion pattern at different time points in the time series.

[0030] Figure 1 FIG. shows a schematic flowchart of a method for solving neutrons in a hybrid energy spectrum reactor based on a neural network provided by an embodiment of the present invention. The method includes:

[0031] S102, obtaining spatio-temporal feature data of materials in different energy spectrum regions.

[0032] Wherein, the material refers to fission materials, breeding materials, etc. in the reactor. The spatio-temporal feature data includes distribution data of temperature, pressure, and neutron density of the above materials in different energy spectrum regions (such as the thermal spectrum region and the fast spectrum region), and data recording the change of the above distribution data over time. The above spatio-temporal feature data can be data obtained in real time during the monitoring of the operation of the hybrid energy spectrum reactor. The pre-trained deep learning model can make predictions based on it to obtain coupled response data and neutron diffusion patterns.

[0033] S104, inputting the spatio-temporal feature data into a pre-trained deep learning model to predict the coupled response data and neutron diffusion patterns between the thermal spectrum region and the fast spectrum region.

[0034] The deep learning model includes a first deep neural network and a second deep neural network. The first deep neural network is used to extract spatial features based on the spatio-temporal feature data, and the second deep neural network is used to extract time features based on the spatial features and the above spatio-temporal feature data to predict the coupled response data and neutron diffusion patterns between the thermal spectrum region and the fast spectrum region. In this embodiment, response data and diffusion patterns are obtained through spatio-temporal feature - classification (mapping). Among them, the neutron diffusion pattern refers to the phenomenon of neutron migration in a medium, which can be determined by the neutron distribution.

[0035] Specifically, the first deep neural network can be a convolutional neural network, and the second deep neural network can be a long short-term memory network.

[0036] Exemplarily, the CNN-LSTM neural network architecture is designed as follows:

[0037] 1. CNN module (for extracting spatial features):

[0038] Input: Multidimensional tensor (such as the neutron density distribution in the core space).

[0039] Structure layer: Set several matrix layers, each layer using different matrix structure sizes to capture different structure layers.

[0040] Pooling layer: Use max pooling or average pooling to reduce the dimension of the feature map and retain the most important spatial information.

[0041] Waveform output: The feature map after passing through the subsequent surface layer will be shallowed and used as the input for the subsequent LSTM module.

[0042] 2. LSTM module (for extracting temporal features):

[0043] Input: Spatial features from the CNN, as well as other physical quantities in the time series (such as the response change between time steps). This response refers to the change process of parameters such as neutron density inside the reactor.

[0044] LSTM layer: Set multiple LSTM networks to capture the unit correlation relationships at different time steps through its memory.

[0045] Dropout layer: Prevent the model from overfitting and improve the generalization ability of the model.

[0046] Fully connected layer:

[0047] The spatio-temporal features after being processed by the LSTM are mapped to the output space through the fully connected layer.

[0048] Output layer: Generate prediction results, specifically the correlation response data between the thermal spectrum region and the fast spectrum region, as well as the neutron diffusion mode.

[0049] 3. Data interaction and flow

[0050] Spatial feature extraction and flattening: The original input data is processed by the CNN to obtain spatial features. After extracting the spatial features, these features are flattened into a one-dimensional feature vector. This process can convert the output of the convolutional layer into the LSTM input format through the Flatten operation. At this stage, the CNN converts the spatial distribution data of neutrons into time series features for the subsequent LSTM to capture time-dependent information.

[0051] Temporal Feature Capture and Output: The LSTM module learns the temporal dynamics from the flattened features and captures the complex dependencies between time steps through the structure of the recurrent neural network. Finally, the output of the LSTM will be passed to the fully connected layer to generate the final prediction result, completing the mapping from space to time and then to the final physical quantity.

[0052] Figure 2 The schematic diagram of the CNN-LSTM combined model provided by the embodiment of the present invention is shown.

[0053] When dealing with the deep learning model, simulation tools can be used to obtain the training data. The input data can be collected and preprocessed. Specifically, the data collection is as follows: Through high-precision reactor physics simulation tools such as Monte Carlo, multi-dimensional physical quantity data is obtained, including: the spatial distribution of temperature, neutron flux density, etc. in the thermal spectrum region and the fast spectrum region; time series data, recording the response of each physical quantity changing with time.

[0054] The data preprocessing is as follows: The multi-dimensional physical quantity data is standardized or normalized; the processed data is represented as a multi-dimensional spatio-temporal tensor. The first step, standardization / normalization: The input data is standardized or normalized to ensure that each physical quantity is within the same numerical range, avoiding the unbalanced influence of different physical quantities on model training. The second step, spatio-temporal feature separation: The data in the reactor is represented as a multi-dimensional spatio-temporal tensor (such as three-dimensional space plus time) so that CNN and LSTM can extract spatial and temporal features respectively.

[0055] As Figure 2 Exemplarily shown, the neutron density distribution (multi-dimensional tensor) in the core space is input into the CNN module, and the CNN module outputs spatial features. The response change between time steps (one-dimensional feature vector) is input into the LSTM module, and temporal features are output.

[0056] Optionally, the training process of the above deep learning model is as follows:

[0057] First, multi-dimensional physical quantity data is obtained through the reactor physics simulation tool. The multi-dimensional physical quantity data can include: the distribution data of temperature, pressure, neutron density in the thermal spectrum region and the fast spectrum region, and the data recording the change of the distribution data with time;

[0058] Secondly, the multi-dimensional physical quantity data is preprocessed;

[0059] Then, input the preprocessed data into the deep learning model for training, and adjust the model parameters based on minimizing the loss function until the loss function drops to a set threshold to obtain the trained deep learning model. Specifically, the spatial tensor of the multi-dimensional spatio-temporal tensor can be input into a convolutional neural network to output spatial features; then, the spatial features are converted into time series features, and the time series features are input into a long short-term memory network to output coupled response data; the loss function is calculated based on the coupled response data and the true response data; after iterative optimization until the loss function drops to the set threshold, the training of the deep learning model ends.

[0060] The embodiment of the present invention uses a deep learning model combining two deep neural networks, breaking through the limitations of traditional numerical simulation methods, having strong spatio-temporal feature capture ability, non-linear mapping ability and generalization ability, showing significant technical effects in accurately predicting the coupled response and neutron diffusion mode in a hybrid energy spectrum reactor, and effectively reducing the computational complexity and experimental dependence, improving the universality and real-time prediction ability of the application.

[0061] In order to effectively fit the coupled response and neutron diffusion mode in the thermal spectrum region and the fast spectrum region, the embodiment of the present invention can adopt a CNN-LSTM combined model for training and optimization. The specific steps are as follows:

[0062] First, input data

[0063] Input of the CNN branch network: Spatial characteristic data of the material in different energy spectrum regions (such as temperature field, pressure field and neutron density distribution). These features are used as the input of the CNN branch network, and the spatial distribution features are extracted through multiple convolutional layers.

[0064] Input of the LSTM backbone network: Time series characteristic data (such as the response data of the reactor at different time points), and the LSTM captures the time evolution mode of the material through its memory unit.

[0065] Second, the specific training process

[0066] Training of the CNN module: First, train the convolutional layer of the convolutional neural network to capture the spatial local features of the material through different convolutional kernels. The output after convolution is flattened and used as the input of the LSTM backbone network.

[0067] Training of the LSTM module: The LSTM network processes time series data to capture the temporal dependence of the material during operation. The LSTM learns the temporal coupled response and diffusion mode by gradually adjusting the state of the memory unit.

[0068] Loss Function and Optimization: The output of the model is compared with the true response curve in the experimental data, and the Mean-Square Error (MSE) is calculated as the loss function. The backpropagation algorithm (such as the Adam optimizer) is used to adjust the weights of the CNN and LSTM to reduce the prediction error.

[0069] Third, Iteration and Convergence of Training

[0070] As the loss function is iteratively optimized, the network weights gradually converge, and the prediction accuracy of the model continuously improves. The batch gradient descent method is used during training to accelerate convergence and ensure that the model can efficiently capture the coupled patterns in space and time.

[0071] Fourth, Model Validation

[0072] The cross-validation method is used to divide the data into a training set, a validation set, and a test set to ensure that the model has good generalization ability. The validation set is used to monitor the performance of the model during training to prevent overfitting, while the test set is used to evaluate the performance of the model on unseen data.

[0073] Fifth, Model Optimization:

[0074] When the loss function of the model drops to the set threshold, the training process of the model ends, and the model can accurately predict the coupled responses and diffusion patterns of different materials. By further adjusting hyperparameters such as the convolution kernel size, the number of LSTM layers, the learning rate, etc., the fitting performance of the model can be further improved.

[0075] In practical applications, the trained CNN-LSTM model can be deployed in the core calculation system to monitor the responses in the thermal spectrum region and the fast spectrum region during operation and predict the neutron diffusion pattern in the subsequent time.

[0076] The above CNN-LSTM model can also perform online learning. As new data accumulates during the reactor operation, the model can continue to perform online learning, update the weights of the neural network, and further improve the prediction accuracy.

[0077] Specific application examples based on the above CNN-LSTM model are as follows:

[0078] First, conduct numerical simulation experiments. Using Monte Carlo, a reactor physics simulation tool, select 12 core power levels such as 0%, 3%, 5%, 10%, 20%, 30%, 50%, 70%, 90%, 95%, 100%, 105% as the initial core nuclear power required by the existing high-precision core steady-state calculation software, and calculate the spatial distribution of data such as temperature and neutron current density in the thermal spectrum region and the fast spectrum region, as well as their changes over time.

[0079] Secondly, perform deep neural network training and fitting. Input the data into the CNN-LSTM network for network training.

[0080] Then, deploy the trained CNN-LSTM model in the core calculation system, which can monitor the responses in the thermal spectrum region and the fast spectrum region during the operation process and predict the neutron diffusion mode in the subsequent time.

[0081] The neutron solution method for the hybrid energy spectrum reactor based on neural network provided by the embodiment of the present invention has the following beneficial effects:

[0082] First, efficient capture and comprehensive modeling of spatio-temporal features. Most of the existing technologies rely on single-dimensional models, such as using numerical solution methods to simulate spatial and temporal features respectively. Such methods often have problems of high computational complexity and limited accuracy when dealing with complex coupled responses. The present invention combines the advantages of CNN and LSTM to achieve global modeling, enabling the model to accurately reflect the complex coupling relationship between the thermal spectrum region and the fast spectrum region.

[0083] Second, strong generalization ability and wide adaptability. Existing physics-based models usually need to rely on a large amount of experimental data for tuning, and complex parameter adjustments need to be made one by one for the characteristics of materials, so the applicable range is limited. The present invention is trained through a neural network, which can learn and extract the characteristics of different materials in different spectrum regions and time periods, thus having strong generalization ability. Once the model is trained, the coupling response and diffusion mode of any material under the quasi-critical condition can be predicted by inputting the element characteristic curve of the material, and there is no need to reconstruct a new numerical model.

[0084] Third, accurate non-linear mapping ability. In existing numerical simulations and experience-based modeling methods, there are often large errors in the processing of non-linear responses. Especially when the coupled response and diffusion mode change non-linearly with material characteristics and time, traditional methods are difficult to capture these complex relationships. The present invention has the non-linear mapping ability of a deep learning model, and this non-linear mapping ability enables the model to have higher accuracy in dealing with complex coupled reactions. Especially when there is a complex non-linear relationship between material characteristics and responses, it can provide more accurate prediction results than traditional models.

[0085] Fourth, it has high computational efficiency and strong real-time prediction ability. Existing technologies usually rely on complex numerical solution methods of finite elements or differential equations for simulation. These methods involve large amounts of calculations. Especially when dealing with spatio-temporal coupling characteristics, long-term iterative calculations may be required, making it difficult to apply in real time. Through the forward propagation process of the deep neural network, the present invention can complete calculations in a relatively short time and generate prediction results of coupling responses and diffusion patterns. After training the network, the model can process input data in real time, is suitable for real-time monitoring and predicting the operating state in the reactor, and can achieve online learning and continuous optimization.

[0086] In summary, the embodiments of the present invention utilize a model combining a deep neural network and CNN-LSTM, breaking through the limitations of traditional numerical simulation methods, and possessing strong spatio-temporal feature capture ability, non-linear mapping ability, and generalization ability. It shows significant technical effects in accurately predicting coupling responses and neutron diffusion patterns in a hybrid energy spectrum reactor, effectively reducing computational complexity and experimental dependence, and enhancing the universality of application and real-time prediction ability.

[0087] Figure 3 The schematic structural diagram of the neutron solving system for a hybrid energy spectrum reactor based on a neural network provided by the embodiments of the present invention is shown. The system includes:

[0088] A data acquisition module 301, configured to acquire spatio-temporal feature data of materials in different energy spectrum regions;

[0089] A prediction module 302, configured to input the spatio-temporal feature data into a pre-trained deep learning model, and predict coupling response data and neutron diffusion patterns between the thermal spectrum region and the fast spectrum region; the deep learning model includes a first deep neural network and a second deep neural network;

[0090] The first deep neural network is used to extract spatial features based on the spatio-temporal feature data, and the second deep neural network is used to extract temporal features based on the spatial features and the spatio-temporal feature data, and predict the coupling response data and the neutron diffusion patterns.

[0091] The embodiments of the present invention utilize a deep learning model combining two deep neural networks, breaking through the limitations of traditional numerical simulation methods, possessing strong spatio-temporal feature capture ability, non-linear mapping ability, and generalization ability, showing significant technical effects in accurately predicting coupling responses and neutron diffusion patterns in a hybrid energy spectrum reactor, effectively reducing computational complexity and experimental dependence, and enhancing the universality of application and real-time prediction ability.

[0092] Optionally, the first deep neural network is a convolutional neural network, and the second deep neural network is a long short-term memory network.

[0093] An embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is read and run by a processor, the method provided in the above embodiment is implemented and the same technical effects can be achieved. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0094] Of course, those skilled in the art can understand that all or part of the processes in implementing the method of the above embodiment can be completed by a computer program instructing a control device. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a memory, a magnetic disk, an optical disc, etc.

[0095] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.

[0096] In this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0097] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0098] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily 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 invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A neutron solution method for a hybrid energy spectrum reactor based on a neural network, characterized in that, Including: Obtaining spatio-temporal characteristic data of materials in different energy spectrum regions; the spatio-temporal characteristic data includes distribution data of temperature, pressure, and neutron density in different energy spectrum regions, and data recording the change of the distribution data over time; Inputting the spatio-temporal characteristic data into a pre-trained deep learning model to predict the coupling response data and neutron diffusion mode between the thermal spectrum region and the fast spectrum region; the deep learning model includes a first deep neural network and a second deep neural network; the first deep neural network is a convolutional neural network, and the second deep neural network is a long short-term memory network; The first deep neural network is used to extract spatial features based on the spatio-temporal characteristic data, and the second deep neural network is used to extract temporal features based on the spatial features and the spatio-temporal characteristic data to predict the coupling response data and the neutron diffusion mode; The convolutional neural network is responsible for extracting the patterns of spatial distribution in the reactor, and the long short-term memory network is responsible for processing the connection responses and the evolution of diffusion modes at different time points in the time series; The method further includes: obtaining multi-dimensional physical quantity data through a reactor physics simulation tool, the multi-dimensional physical quantity data including: distribution data of temperature, pressure, and neutron density in the thermal spectrum region and the fast spectrum region, and data recording the change of the distribution data over time; preprocessing the multi-dimensional physical quantity data; inputting the preprocessed data into the deep learning model for training, and adjusting the model parameters based on minimizing the loss function until the loss function drops to a set threshold to obtain a trained deep learning model; The preprocessing of the multi-dimensional physical quantity data includes: performing standardization or normalization processing on the multi-dimensional physical quantity data; representing the processed data as a multi-dimensional spatio-temporal tensor; The inputting the preprocessed data into the deep learning model for training includes: inputting the spatial tensor of the multi-dimensional spatio-temporal tensor into the convolutional neural network to output spatial features; converting the spatial features into time series features, and inputting the time series features into the long short-term memory network to output coupling response data; calculating the loss function according to the coupling response data and the true response data; after iteratively optimizing until the loss function drops to a set threshold, the training of the deep learning model ends.

2. The method according to claim 1, wherein The method further includes: Obtaining the multi-dimensional physical quantity data accumulated during the operation of the reactor, and training and updating the deep learning model according to the accumulated multi-dimensional physical quantity data.

3. The method according to claim 1, wherein The method further includes: Using the cross-validation method to divide the multi-dimensional physical quantity data into a training set, a validation set, and a test set; the validation set is used to monitor the performance of the model during training, and the test set is used to evaluate the performance of the model on unseen data.

4. A neutron solving system for a hybrid energy spectrum reactor based on a neural network, characterized in that, Including: A data acquisition module for obtaining spatio-temporal characteristic data of materials in different energy spectrum regions; the spatio-temporal characteristic data includes distribution data of temperature, pressure, and neutron density in different energy spectrum regions, and data recording the change of the distribution data over time; A prediction module, configured to input the spatio-temporal feature data into a pre-trained deep learning model, and predict the coupling response data and the neutron diffusion mode between the thermal spectrum region and the fast spectrum region; the deep learning model includes a first deep neural network and a second deep neural network; the first deep neural network is a convolutional neural network, and the second deep neural network is a long short-term memory network; The first deep neural network is configured to extract spatial features based on the spatio-temporal feature data, and the second deep neural network is configured to extract temporal features based on the spatial features and the spatio-temporal feature data, and predict the coupling response data and the neutron diffusion mode; The convolutional neural network is responsible for extracting the patterns of spatial distribution in the reactor, and the long short-term memory network is responsible for processing the connection responses at different time points in the time series and the evolution of the diffusion mode; The system further includes a training module, configured to: obtain multi-dimensional physical quantity data through a reactor physics simulation tool, where the multi-dimensional physical quantity data includes: distribution data of temperature, pressure, and neutron density in the thermal spectrum region and the fast spectrum region, and data recording the change of the distribution data over time; preprocess the multi-dimensional physical quantity data; input the preprocessed data into the deep learning model for training, and adjust the model parameters based on minimizing a loss function until the loss function drops to a set threshold, so as to obtain a trained deep learning model; The preprocessing of the multi-dimensional physical quantity data includes: performing standardization or normalization processing on the multi-dimensional physical quantity data; representing the processed data as a multi-dimensional spatio-temporal tensor; The inputting the preprocessed data into the deep learning model for training includes: inputting the spatial tensor of the multi-dimensional spatio-temporal tensor into the convolutional neural network to output spatial features; converting the spatial features into time series features, and inputting the time series features into the long short-term memory network to output coupling response data; calculating a loss function according to the coupling response data and the true response data; after iteratively optimizing until the loss function drops to a set threshold, the training of the deep learning model ends.

Images