Reactor core power distribution prediction method and device, equipment and storage medium
By performing refined segmentation of fuel rods and using graph neural network prediction, the problem that traditional segmentation methods cannot accurately reflect changes in fuel rod size has been solved, and refined prediction of nuclear reactor core power distribution has been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-10
- Publication Date
- 2026-03-13
AI Technical Summary
Traditional nodalization methods cannot accurately reflect the fine distribution changes at the fuel rod scale in nuclear reactors, resulting in inaccurate power distribution predictions.
By finely dividing the fuel rods into segments and constructing a graph structure, the power of each segment is predicted using a graph neural network. Combined with neutron flux density and spatial location information, refined power distribution prediction is achieved.
It improves the accuracy and efficiency of power distribution prediction, and can more accurately reflect the power changes inside the fuel rod.
Smart Images
Figure CN121662461A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of nuclear reactor technology, and more specifically, to a method, apparatus, equipment, and storage medium for predicting reactor core power distribution. Background Technology
[0002] In nuclear reactor safety analysis and operation management, accurate local power distribution is a key factor in assessing core thermal-hydraulic characteristics, identifying potential hotspots, and preventing fuel rod failure. The commonly used nodal method divides the core into macroscopic nodules, enabling the acquisition of neutronics parameter distributions across the entire reactor with reasonable computational resources, providing crucial information for reactor operation and safety reviews. However, the traditional nodal method, based on homogenization assumptions, only provides average physical quantities (such as power density and neutron flux) within the nodules, failing to accurately reflect the fine-grained distribution variations within the nodules, particularly at the fuel rod scale. Summary of the Invention
[0003] The purpose of this invention is to provide a method, apparatus, device, and storage medium for predicting the power distribution of a reactor core. By dividing the fuel rods into segments, the neutron flux density information of each segment is obtained. Then, based on a graph neural network, the power of each segment is predicted to obtain a refined power distribution.
[0004] In a first aspect, the present invention provides a method for predicting reactor core power distribution, comprising: Each fuel rod in the reactor core is divided into m segments, resulting in n×m segments; where m is an integer greater than 1, and n is the number of fuel rods in the reactor core. Based on the nodal method, the neutron flux density information of each nodal is determined; Based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes, a graph structure is constructed; the graph structure is used to characterize the neutron flux density distribution information among the nodes in the core. The graph structure is input into the power prediction model to obtain the power prediction value of each node output by the power prediction model; wherein, the power prediction model has learned to predict the power of each node based on the graph structure; the power prediction model is built based on a graph neural network; Based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes, the power distribution information of the reactor core is obtained.
[0005] In some embodiments, constructing a graph structure based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes includes: Using each of the aforementioned blocks as nodes, the initial feature information of the corresponding nodes is determined based on the neutron flux density information of each of the aforementioned blocks; Based on the spatial location information of each of the n×m nodes, the edge weights between each pair of nodes are determined. The graph structure is formed by n×m nodes, their initial feature information, and the edge weights between each pair of nodes.
[0006] In some embodiments, the power prediction model includes a graph neural network layer and a fully connected layer; the step of inputting the graph structure into the power prediction model to obtain the power prediction value of each node output by the power prediction model includes: The graph structure is input into the graph neural network layer, so that the graph neural network layer updates the feature information of each node based on the initial feature information of each of the n×m nodes and the edge weights between each pair of nodes, and outputs the updated feature information of each node as the feature information of the corresponding block. The feature information of each node is input into the fully connected layer for mapping to obtain the power prediction value of each node.
[0007] As one possible implementation, the graph neural network layer includes multiple consecutive graph neural network sub-layers; each graph neural network sub-layer is used to continue updating the feature information of each node based on the edge weights between each pair of nodes, the feature update results of each of the n×m nodes output by the previous graph neural network sub-layer, and the weight matrix of the graph neural network sub-layer.
[0008] In some embodiments, the power prediction model is pre-trained in the following manner: Obtain training samples, which include graph structure samples and their corresponding power label values; The graph structure sample is input into the initial power prediction model to obtain the power prediction sample output by the initial power prediction model; Based on the power prediction sample and the power label value, a first loss value is determined; the first loss value is used to characterize the degree of difference between the power prediction sample and the power label value. A second loss value is determined based on the power prediction sample and the expected power distribution pattern of the reactor core; the expected power distribution pattern is determined based on the shape factor of the fuel assemblies in the reactor core, and the second loss value is used to characterize the degree of difference between the distribution of the power prediction sample and the expected power distribution pattern. Based on the first loss value and the second loss value, determine the total loss value; Based on the total loss value, the initial power prediction model is trained to obtain the trained power prediction model.
[0009] As an example, determining the total loss value based on the first loss value and the second loss value includes: The first loss value and the second loss value are weighted and summed, and the weighted sum is used as the total loss value.
[0010] In some embodiments, the neutron flux density information includes at least one of the following: average flux density, surface partial neutron flux density, and surface neutron flux density.
[0011] In a second aspect, the present invention provides a device for predicting the power distribution of a reactor core, comprising: The partitioning module is used to divide each fuel rod in the reactor core into m segments, resulting in n×m segments; The determination module is used to determine the neutron flux density information of each of the nodes based on the nodal method; A construction module is used to input the graph structure into a power prediction model to obtain the power prediction value of each node output by the power prediction model; wherein, the power prediction model has been learned to predict the power of each node based on the graph structure; the power prediction model is built based on a graph neural network; A prediction module is used to input the graph structure into a power prediction model to obtain the power prediction value of each node output by the power prediction graph neural network; wherein the power prediction model has been learned to predict the power of each node based on the graph structure; The acquisition module is used to obtain the power distribution information of the reactor core based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes.
[0012] A third aspect of the present invention provides an electronic device including a processor and a memory storing a computer program, wherein the processor executes the program to implement the reactor core power distribution prediction method described in the first aspect above.
[0013] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for predicting reactor core power distribution as described in the first aspect.
[0014] The method, apparatus, equipment, and storage medium for predicting reactor core power distribution provided by this invention have the following beneficial effects: This invention achieves refined core power distribution information by dividing the reactor into segments and obtaining the neutron flux density of each segment based on the segment method. Then, based on the neutron flux density and spatial location information of each segment, a graph structure is constructed. Through a power prediction model based on a graph neural network, the power prediction of each segment is accurately realized, thereby obtaining refined core power distribution information. This not only improves the efficiency of power distribution prediction but also improves the accuracy of power distribution prediction. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0016] Figure 1 A schematic flowchart of a method for predicting reactor core power distribution provided in an embodiment of the present invention; Figure 2 A schematic flowchart of another method for predicting reactor core power distribution provided in an embodiment of the present invention; Figure 3 A schematic diagram of a reactor core power distribution prediction device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0017] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] In related technologies, to improve the resolution of power distribution calculations, polynomial expansion modulation methods are typically employed. These methods reconstruct a fine power distribution by fitting the neutron flux density distribution within a node and combining it with a shape function or shape factor. However, polynomial fitting models cannot effectively handle complex nonlinear flux distributions, especially when there is significant non-uniformity in the neutron flux distribution. The fitting results are not ideal, and polynomial fitting modulation methods simplify local physical properties, failing to fully capture the more detailed power distribution of the fuel assembly.
[0019] To address the aforementioned problems, this invention provides a method, apparatus, device, and storage medium for predicting reactor core power distribution.
[0020] Figure 1 This is a schematic flowchart illustrating a method for predicting reactor core power distribution according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step 101: Divide each fuel rod in the reactor core into m segments to obtain n×m segments.
[0021] Where m is an integer greater than 1, and n is the number of fuel rods in the reactor core.
[0022] In some embodiments, each fuel rod is divided into m segments in the longitudinal direction, and the division can be an average division, dividing each fuel rod into m segments of the same length.
[0023] Step 102: Based on the nodal method, determine the neutron flux density information of each nodal.
[0024] The neutron flux density information for each node may include at least one of the following: average flux density, surface partial neutron flux density, and surface neutron flux density. As an example, the neutron flux density information for each node may include average flux density, surface partial neutron flux density, and surface neutron flux density.
[0025] It should be noted that the nodal method is a widely used method in nuclear reactor physics calculations to solve approximate solutions to neutron diffusion or transport equations in the reactor core. Related technologies employ the nodal method by dividing the reactor core into larger nodal units and solving for each unit to calculate macroscopic physical quantities such as flux density. However, this invention does not divide the reactor core into larger nodal units; instead, it divides each fuel rod into nodal units. Essentially, it first calculates an approximate solution related to the power of each nodal unit based on this refined nodal division using the nodal method.
[0026] In some embodiments, the specific process of determining the neutron flux density information of each node based on the nodal method can be calculated based on the nodal method in related technologies, or it can be implemented based on automated programs such as SIMULATE, PARCS, and DYN3D.
[0027] In some embodiments, within each node, assuming that the neutron flux and power distribution are uniform or change slowly, the physical parameters within the node (such as diffusion coefficient, flux density, etc.) can be averaged or processed by interpolation.
[0028] As an example, the specific solution process of the nodal method includes: assuming that the interior of each nodal is a uniform region; giving the initial average flux and surface flux (e.g., uniform distribution) of each nodal; calculating the surface flux density based on the initial flux gradient and diffusion coefficient; solving the neutron diffusion equation for each nodal, substituting it into the integral diffusion equation, and solving for the sum of flux densities to update the average flux; adjusting the surface flux by the continuity condition of adjacent nodals; and stopping the iteration when the relative error between adjacent iterations is less than a threshold and the effective multiplication coefficient is stable.
[0029] Step 103: Based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes, construct a graph structure; the graph structure is used to characterize the neutron flux density distribution information among the nodes in the core.
[0030] In some embodiments, the graph structure includes multiple nodes, edges between nodes, feature information of each node, and edge weights. Each node can be considered a node, the feature information of each node is used to characterize the neutron flux density information of the corresponding node, and the edge weights of each edge are used to characterize the spatial positional relationship between the nodes corresponding to the node.
[0031] In some embodiments, the process of constructing a graph structure based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes includes the following steps: Step S1: Using each block as a node, determine the initial feature information of the corresponding node based on the neutron flux density information of each block.
[0032] In some embodiments, the neutron flux density information of each block can be used as the initial feature information of the corresponding node.
[0033] In other embodiments, the neutron flux density information of each block can be spliced together, for example, the average flux density, surface partial neutron flux density, and surface neutron flux density can be spliced together, and then the spliced data can be encoded and converted to obtain the initial feature information of the corresponding node.
[0034] Step S2: Based on the spatial location information of each of the n×m nodes, determine the edge weights between each pair of nodes.
[0035] In some embodiments, the edge weights between pairs of nodes are used to characterize the spatial relationship between corresponding nodes. Since neutron flow or heat exchange exists between nearby fuel rods, edge weights can be defined based on the spatial relationship between the nodes connected by the edges and the corresponding nodes. For example, edge weights can be determined based on the geometric distance between the fuel rods containing the nodes; a larger edge weight indicates a greater mutual influence between the corresponding nodes.
[0036] Step S3: Construct a graph structure from the n×m nodes, their initial feature information, and the edge weights between each pair of nodes.
[0037] In other words, the graph structure includes n×m nodes, initial feature information of each node, and edge weights between any two nodes.
[0038] Step 104: Input the graph structure into the power prediction model to obtain the power prediction value of each node output by the power prediction model; wherein, the power prediction model has learned the ability to predict the power of each node based on the graph structure.
[0039] In some embodiments, the power prediction model is based on a graph neural network. Since graph neural networks can learn the complex relationships between nodes and better capture local non-uniform features, they can more accurately predict the power at different locations on the fuel rod during the power reconstruction process.
[0040] In some embodiments, the power prediction model can update the feature information of each node based on the initial feature information of each node in the graph structure and the edge weights of each pair of nodes, so that the updated node features contain both local and global features, and then perform mapping based on the updated node features to obtain the power prediction value of each block.
[0041] As one possible implementation, the power prediction model includes graph neural network layers and fully connected layers. Inputting the graph structure into the power prediction model to obtain the power prediction value for each node in the output of the power prediction model includes the following steps: Step S4: Input the graph structure into the graph neural network layer, so that the graph neural network layer updates the feature information of each node based on the initial feature information of each of the n×m nodes and the edge weights between each pair of nodes, and outputs the updated feature information of each node.
[0042] As an example, a graph neural network layer can perform a weighted calculation on each node, taking the initial feature information of that node, the initial feature information of other nodes, and the edge weights, and outputting the result as the updated feature information of that node. The weight matrix used for the weighted calculation is obtained based on training.
[0043] In some embodiments, the graph neural network layer includes multiple consecutive graph neural network sub-layers; each graph neural network sub-layer is used to continue updating the feature information of each node based on the edge weights between each pair of nodes, the feature update results of each of the n×m nodes output by the previous graph neural network sub-layer, and the weight matrix of the graph neural network sub-layer.
[0044] As an example, the graph structure is input into the first layer of the graph neural network, which outputs the first updated feature information of each node. The edge weights of each pair of nodes and the first updated feature information of each node are input into the second layer of the graph neural network, which outputs the second updated feature information of each node. The edge weights of each pair of nodes and the second updated feature information of each node are input into the third layer of the graph neural network, which outputs the third updated feature information of each node. This process continues until the last layer of the graph neural network outputs the updated feature information of each node that it has calculated.
[0045] The execution process of each graph neural network sub-layer includes: for each node, updating the feature information of the node by taking the feature update result output by the previous neural network sub-layer, the edge weights between the node and other nodes, the feature update result output by the previous neural network sub-layer, and the weight matrix of the layer, and obtaining the feature update result of the node as shown in the following formula (1).
[0046] (1); in, For the first The feature information of the i-th node output by a sublayer of a layered graph neural network. It is an integer, and Less than or equal to the total number of sub-layers in the graph neural network; For the first Weight matrices of sublayers in a layered graph neural network; For the first The feature information of the i-th node output by a sublayer of a layered graph neural network; For the first The feature information of nodes other than the i-th node output by the sublayer of the layer graph neural network; Let be the edge weight between the i-th node and the j-th node; For bias terms; This is the activation function.
[0047] As can be seen, graph neural network layers can update the features of each node through message passing mechanisms and by exchanging information between adjacent nodes, so that the features of each node contain both rich global information and the node's own local features.
[0048] Step S5: Input the feature information of each node into the fully connected layer for mapping to obtain the power prediction value of each node.
[0049] Step 105: Based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes, obtain the power distribution information of the reactor core.
[0050] In some embodiments, based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes, the power distribution information of each fuel rod can be obtained, and then based on the spatial location information of each fuel rod, the power distribution information of the reactor core can be obtained.
[0051] The reactor core power distribution prediction method according to embodiments of the present invention obtains the neutron flux density of each node by refining the node segmentation and based on the node method. Then, based on the neutron flux density of each node and the spatial location information of the node, a graph structure is constructed. Through a power prediction model based on a graph neural network, the power prediction of each node is accurately realized, thereby obtaining refined core power distribution information. This method can not only improve the efficiency of power distribution prediction, but also improve the accuracy of power distribution prediction.
[0052] Next, we will provide a detailed introduction to the training process of the power prediction model.
[0053] Figure 2 A schematic flowchart illustrating another method for predicting reactor core power distribution provided in an embodiment of the present invention. Figure 2 As shown, the power prediction model in the above embodiment was pre-trained in the following manner: Step 201: Obtain training samples, which include graph structure samples and their corresponding power label values.
[0054] The power tag value refers to the power tag value of each node sample in the graph structure sample, which can be obtained based on the existing database.
[0055] Step 202: Input the graph structure sample into the initial power prediction model to obtain the power prediction sample output by the initial power prediction model.
[0056] Step 203: Determine the first loss value based on the power prediction sample and the power label value; the first loss value is used to characterize the degree of difference between the power prediction sample and the power label value.
[0057] As an example, the first loss value can be calculated using the following formula (2).
[0058] (2); in, For sample size; This is the power prediction sample for the i-th segment; The power tag value for the i-th segment; This is the first loss value.
[0059] Step 204: Determine the second loss value based on the power prediction sample and the expected power distribution pattern of the reactor core; the expected power distribution pattern is determined based on the shape factor of the fuel assemblies in the reactor core, and the second loss value is used to characterize the degree of difference between the distribution of the power prediction sample and the expected power distribution pattern.
[0060] The shape factor of the fuel assemblies within the reactor core mainly refers to the physical geometry of the fuel assemblies and the distribution of fuel enrichment within them. The shape factor primarily affects the power distribution within the reactor core through the neutron flux density distribution; therefore, the expected power distribution pattern can be determined by using the shape factor of the fuel assemblies within the reactor core. Incorporating this expected power distribution pattern into the loss value calculation ensures that the power distribution predicted by the model conforms to the expected power distribution pattern, thereby improving the model's training efficiency and effectiveness.
[0061] In some embodiments, based on the power prediction samples, the power distribution samples predicted by the model can be obtained, and the power distribution samples and the expected power distribution pattern are used to calculate the loss function to obtain the second loss value.
[0062] As an example, the expected power distribution pattern can be calculated as the ratio of the power of each node to the average power. Based on the power prediction samples of each node within the core, the average power sample within the core is calculated. The ratio of the power prediction sample to the average power sample of each node is then used to determine the power ratio sample for each node. The difference between the power ratio sample of each node and the power ratio of each node in the expected power distribution pattern is used to calculate the loss value, resulting in a second loss value. For example, the second loss value can be calculated using the root mean square method.
[0063] Step 205: Determine the total loss value based on the first loss value and the second loss value.
[0064] In some embodiments, the first loss value and the second loss value can be weighted to obtain the total loss value, or the first loss value and the second loss value can be nonlinearly mapped to obtain the total loss value.
[0065] Step 206: Based on the total loss value, train the initial power prediction model to obtain the trained power prediction model.
[0066] According to the reactor core power distribution prediction method of the present invention, during the power prediction model training, a loss value calculation is introduced based on the expected power distribution law determined by the shape factor, so that the model prediction result can not only approach the label value, but also conform to the expected power distribution law, thereby improving the training effect of the model training efficiency.
[0067] To achieve the above embodiments, the present invention also provides a device for predicting reactor core power distribution.
[0068] Figure 3 This is a schematic diagram of a reactor core power distribution prediction device provided in an embodiment of the present invention. Figure 3 As shown, the device includes a division module 310, a determination module 320, a construction module 330, a prediction module 340, and an acquisition module 350.
[0069] The partitioning module 310 is used to divide each fuel rod in the reactor core into m segments, resulting in n×m segments; where m is an integer greater than 1, and n is the number of fuel rods in the reactor core. The determination module 320 is used to determine the neutron flux density information of each of the nodes based on the nodal method; The construction module 330 is used to construct a graph structure based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes; the graph structure is used to characterize the neutron flux density distribution information among the nodes in the core. The prediction module 340 is used to input the graph structure into the power prediction model to obtain the power prediction value of each node output by the power prediction model; wherein, the power prediction model has learned the ability to predict the power of each node based on the graph structure; the power prediction model is built based on a graph neural network; The module 350 is used to obtain the power distribution information of the reactor core based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes.
[0070] In some embodiments, the construction module 330 is specifically used for: Using each of the aforementioned blocks as nodes, the initial feature information of the corresponding nodes is determined based on the neutron flux density information of each of the aforementioned blocks; Based on the spatial location information of each of the n×m nodes, the edge weights between each pair of nodes are determined. The graph structure is formed by n×m nodes, their initial feature information, and the edge weights between each pair of nodes.
[0071] In some embodiments, the power prediction model includes a graph neural network layer and a fully connected layer; the prediction module 340 is specifically used for: The graph structure is input into the graph neural network layer, so that the graph neural network layer updates the feature information of each node based on the initial feature information of each of the n×m nodes and the edge weights between each pair of nodes, and outputs the updated feature information of each node. The updated feature information of each node is input into the fully connected layer for mapping to obtain the power prediction value of each node block.
[0072] As one possible implementation, the graph neural network layer includes multiple consecutive graph neural network sub-layers; each graph neural network sub-layer is used to continue updating the feature information of each node based on the edge weights between each pair of nodes, the feature update results of each of the n×m nodes output by the previous graph neural network sub-layer, and the weight matrix of the graph neural network sub-layer.
[0073] In some embodiments, the device further includes a training module 360, which is used for: Obtain training samples, which include graph structure samples and their corresponding power label values; The graph structure sample is input into the initial power prediction model to obtain the power prediction sample output by the initial power prediction model; Based on the power prediction sample and the power label value, a first loss value is determined; the first loss value is used to characterize the degree of difference between the power prediction sample and the power label value. A second loss value is determined based on the power prediction sample and the expected power distribution pattern of the reactor core; the expected power distribution pattern is determined based on the shape factor of the fuel assemblies in the reactor core, and the second loss value is used to characterize the degree of difference between the distribution of the power prediction sample and the expected power distribution pattern. Based on the first loss value and the second loss value, determine the total loss value; Based on the total loss value, the initial power prediction model is trained to obtain the trained power prediction model.
[0074] As an example, the training module 360 is also used for: The first loss value and the second loss value are weighted and summed, and the weighted sum is used as the total loss value.
[0075] As an example, the neutron flux density information includes at least one of the following: average flux density, surface partial neutron flux density, and surface neutron flux density.
[0076] It should be noted that the explanations and descriptions in the above embodiments regarding the method for predicting reactor core power distribution can also be applied to the device for predicting reactor core power distribution in the embodiments of the present invention, and will not be repeated here.
[0077] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call the computer program in the memory 430 to execute the steps of the reactor core power distribution prediction method provided in the above embodiments.
[0078] For example, the method includes: dividing each fuel rod in the reactor core into m segments to obtain n×m segments; where m is an integer greater than 1, and n is the number of fuel rods in the reactor core; determining the neutron flux density information of each segment based on the segment method; constructing a graph structure based on the neutron flux density information of each of the n×m segments and the spatial location information of each of the n×m segments; the graph structure is used to characterize the neutron flux density distribution information among the segments in the reactor core; inputting the graph structure into a power prediction model to obtain the power prediction value of each segment output by the power prediction model; wherein the power prediction model has learned the ability to predict the power of each segment based on the graph structure; the power prediction model is constructed based on a graph neural network; and obtaining the power distribution information of the reactor core based on the power prediction values of each of the n×m segments and the spatial location information of each of the n×m segments.
[0079] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0080] On the other hand, embodiments of the present invention also provide a computer program product, the computer program product including a computer program that can be stored on a computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to perform the steps of the reactor core power distribution prediction method provided in the above embodiments.
[0081] On the other hand, embodiments of the present invention also provide a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being used to cause a processor to execute the reactor core power distribution prediction method provided in the above embodiments.
[0082] The non-transitory computer-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).
[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0084] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting reactor core power distribution, characterized in that, include: Each fuel rod in the reactor core is divided into m segments, resulting in n×m segments; where m is an integer greater than 1, and n is the number of fuel rods in the reactor core. Based on the nodal method, the neutron flux density information of each nodal is determined; Based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes, a graph structure is constructed; the graph structure is used to characterize the neutron flux density distribution information among the nodes in the core. The graph structure is input into the power prediction model to obtain the power prediction value of each node output by the power prediction model; wherein, the power prediction model has learned to predict the power of each node based on the graph structure; the power prediction model is built based on a graph neural network; Based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes, the power distribution information of the reactor core is obtained.
2. The method according to claim 1, characterized in that, The construction of the graph structure based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes includes: Using each of the aforementioned blocks as nodes, the initial feature information of the corresponding nodes is determined based on the neutron flux density information of each of the aforementioned blocks; Based on the spatial location information of each of the n×m nodes, the edge weights between each pair of nodes are determined. The graph structure is formed by n×m nodes, their initial feature information, and the edge weights between each pair of nodes.
3. The method according to claim 2, characterized in that, The power prediction model includes a graph neural network layer and a fully connected layer; the step of inputting the graph structure into the power prediction model to obtain the power prediction value of each node output by the power prediction model includes: The graph structure is input into the graph neural network layer, so that the graph neural network layer updates the feature information of each node based on the initial feature information of each of the n×m nodes and the edge weights between each pair of nodes, and outputs the updated feature information of each node. The updated feature information of each node is input into the fully connected layer for mapping to obtain the power prediction value of each node block.
4. The method according to claim 3, characterized in that, The graph neural network layer includes multiple consecutive graph neural network sub-layers; each graph neural network sub-layer is used to update the feature information of each node based on the edge weights between each pair of nodes, the feature update results of each of the n×m nodes output by the previous graph neural network sub-layer, and the weight matrix of the graph neural network sub-layer.
5. The method according to claim 1, characterized in that, The power prediction model was pre-trained in the following manner: Obtain training samples, which include graph structure samples and their corresponding power label values; The graph structure sample is input into the initial power prediction model to obtain the power prediction sample output by the initial power prediction model; Based on the power prediction sample and the power label value, a first loss value is determined; the first loss value is used to characterize the degree of difference between the power prediction sample and the power label value. A second loss value is determined based on the power prediction sample and the expected power distribution pattern of the reactor core; the expected power distribution pattern is determined based on the shape factor of the fuel assemblies in the reactor core, and the second loss value is used to characterize the degree of difference between the distribution of the power prediction sample and the expected power distribution pattern. Based on the first loss value and the second loss value, determine the total loss value; Based on the total loss value, the initial power prediction model is trained to obtain the trained power prediction model.
6. The method according to claim 5, characterized in that, The step of determining the total loss value based on the first loss value and the second loss value includes: The first loss value and the second loss value are weighted and summed, and the weighted sum is used as the total loss value.
7. The method according to any one of claims 1 to 6, characterized in that, The neutron flux density information includes at least one of the following: average flux density, surface partial neutron flux density, and surface neutron flux density.
8. A device for predicting reactor core power distribution, characterized in that, include: The partitioning module is used to divide each fuel rod in the reactor core into m segments, resulting in n×m segments; where m is an integer greater than 1, and n is the number of fuel rods in the reactor core. The determination module is used to determine the neutron flux density information of each of the nodes based on the nodal method; A construction module is used to construct a graph structure based on the neutron flux density information of each of the n×m nodes and the spatial location information of each of the n×m nodes; the graph structure is used to characterize the neutron flux density distribution information among the nodes in the core. A prediction module is used to input the graph structure into a power prediction model to obtain the power prediction value of each node output by the power prediction model; wherein the power prediction model has been learned to predict the power of each node based on the graph structure; the power prediction model is built based on a graph neural network; The acquisition module is used to obtain the power distribution information of the reactor core based on the power prediction values of each of the n×m nodes and the spatial location information of each of the n×m nodes.
9. An electronic device, characterized in that, It includes a processor and a memory storing a computer program, wherein the processor executes the program to implement the method for predicting the reactor core power distribution according to any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the reactor core power distribution as described in any one of claims 1 to 7.