A data-driven refined core physics-thermal coupling calculation method

Abstract

The present invention discloses a data-driven refined core physical-thermal coupling calculation method, which belongs to the field of nuclear power technology. The method comprises collecting nuclear reactor core data through a data acquisition module, and using a physical model and a high-fidelity CFD model in a model construction module to construct a physical and thermal model of the core. A coupling control program is used to realize data transmission and coupling calculation between the physical model and the CFD model. A multi-fidelity flow field data reconstruction model is adopted, and high-fidelity and low-fidelity flow field data are combined to further optimize the coupling calculation process and perform iterations. The method solves the technical problem of fast and stable refined thermal data acquisition, avoids repeated high-fidelity CFD model calculations, reduces computing costs, reduces the possibility of non-physical solutions, improves the accuracy of simulation results, improves the reliability of data reconstruction, and enhances the stability of coupling calculations.

Description

Technical Field

[0001] The present invention belongs to the field of nuclear power technology, and in particular relates to a data-driven refined core physics-thermal coupling calculation method. Background Art

[0002] In recent years, with the significant improvement of computer performance and model accuracy, the use of computational fluid dynamics (CFD) for thermal analysis and refined three-dimensional neutron transport physics programs for physical analysis, and coupling them, has become an important part of modern nuclear design and verification work.

[0003] Currently, coupled calculations are performed using the Picard (or fixed point) iteration method, which solves each physical field sequentially. While this method is relatively simple and can effectively use existing programs to accurately solve each independent physical field, it has the following two problems:

[0004] 1. Slow convergence. Especially under conditions of distorted power distribution, such as a rod-flicking accident, oscillations may occur during the iteration process, or even lead to non-convergence, seriously affecting computational efficiency.

[0005] 2. Using traditional numerical calculation methods to directly predict the temperature field is prone to non-numerical solutions, especially when using only a small amount of pre-calculated physical-thermal coupled calculation results. In this case, the temperature values ​​and distribution trends of the sample temperature field data and the final converged temperature field data may differ significantly, which may provide erroneous temperature field information and lead to errors in the coupled calculation results. Summary of the Invention

[0006] The purpose of the present invention is to provide a data-driven refined core physics-thermal coupling calculation method, which solves the technical problem of fast and stable refined thermal data acquisition.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A data-driven, refined core physics-thermal coupling calculation method includes the following steps:

[0009] Step 1: Establish a data acquisition module, which acquires the core data of the nuclear reactor and stores it in the database;

[0010] Step 2: Establish a model building module. The model building module obtains core data from the database and performs physical and thermal calculations of the core using the physical model and the high-fidelity CFD model, respectively. The physical model uses a structured grid, while the high-fidelity CFD model uses an unstructured grid.

[0011] The core neutron flux and core power distribution are obtained by physical calculations, and the coolant and fuel rod temperatures are obtained by thermal calculations;

[0012] Establish an interface between the physical model and the high-fidelity CFD model to transfer temperature and power data between the two models, and set the coupling control program between the two models, including controlling the physical model to transfer power distribution to the high-fidelity CFD model and controlling the high-fidelity CFD model to transfer temperature data to the physical model;

[0013] Depending on the complexity of the nuclear reactor, several coupled calculations are performed without converging. The calculation data of the high-fidelity CFD model and the power distribution calculated by the physical model are retained to obtain a high-fidelity flow field data set.

[0014] Step 3: Establish a simplified CFD module. The simplified CFD module obtains a low-fidelity CFD model by roughening the network of the high-fidelity CFD model. The low-fidelity CFD model is used to perform calculations based on the power distribution under the same boundary conditions to obtain a low-fidelity flow field dataset.

[0015] Step 4: Establish a flow field data processing module. The flow field data processing module uses a high-fidelity autoencoder, a low-fidelity autoencoder, and a BP neural network model to reduce the dimension and map the high-fidelity flow field dataset and the low-fidelity flow field dataset, and establish a multi-fidelity flow field data reconstruction model MFM.

[0016] The multi-fidelity flow field data reconstruction model MFM is trained using high-fidelity flow field datasets and low-fidelity flow field datasets.

[0017] Step 5: Establish a multi-fidelity coupling calculation module. The multi-fidelity coupling calculation module embeds the low-fidelity CFD model and the multi-fidelity flow field data reconstruction model (MFM) into the coupling calculation process, thereby replacing the high-fidelity CFD model to calculate the flow field data. During the coupling calculation process, the power distribution is transferred to the low-fidelity CFD model to calculate the low-fidelity flow field data. The low-fidelity flow field data is processed using the multi-fidelity flow field data reconstruction model (MFM) to obtain the predicted data of the high-fidelity flow field data.

[0018] Step 6: The back-substitution calculation module back-substitutes the predicted data into the high-fidelity CFD model and calculates the temperature field data based on the power distribution;

[0019] Step 7: The iterative module uses the coupling control program according to the method in step 2 to transfer the temperature field data to the physical model and perform a new round of coupling calculation;

[0020] Repeat steps 5 and 6 until the coupled calculation reaches convergence.

[0021] Preferably, when executing step 2, the core data includes core structure, fuel rod physical parameters, coolant temperature and fuel rod temperature;

[0022] The physical model is a refined three-dimensional neutron transport physical model;

[0023] Physical calculations include calculating the core neutron flux and power distribution based on the core structure, fuel rod physical parameters, coolant temperature, and fuel rod temperature;

[0024] Thermal calculations include the calculation of coolant temperature and fuel rod temperature based on the core structure, flow field boundary conditions and power distribution.

[0025] Preferably, when executing step 2, the coupling process specifically includes the following steps:

[0026] Step 2-1: The high-fidelity CFD model integrates and averages the temperature data of the CFD calculation grid overlaid on the physical grid according to the grid position of the physical model to obtain the average value;

[0027] Step 2-2: The high-fidelity CFD model transfers the average value to the corresponding grid of the physical model through the interface;

[0028] Step 2-3: The physical model transfers the power data of each grid to the corresponding grid of the high-fidelity CFD model through the interface.

[0029] Preferably, when executing step 3, the degree of roughening is determined according to calculation accuracy and cost.

[0030] Preferably, when executing step 4, the following steps are specifically included:

[0031] Step 4-1: Establish a high-fidelity autoencoder and a low-fidelity autoencoder respectively. Perform dimensionality reduction on the high-fidelity flow field dataset obtained in step 2 and the low-fidelity flow field dataset obtained in step 3 respectively.

[0032] Step 4-2: Establish a BP neural network model to map the high-fidelity flow field data and the low-fidelity flow field data after dimensionality reduction;

[0033] Step 4-3: Combine the high-fidelity autoencoder, the low-fidelity autoencoder, and the BP neural network model into a multi-fidelity flow field data reconstruction model MFM, where the input of the multi-fidelity flow field data reconstruction model MFM is the low-fidelity flow field data and the output is the predicted data of the high-fidelity flow field data;

[0034] Step 4-4: Use the high-fidelity flow field dataset and the low-fidelity flow field dataset to train the multi-fidelity flow field data reconstruction model MFM.

[0035] Preferably, when executing step 6, it specifically includes substituting the predicted data back into the energy equation of the high-fidelity CFD model, and calculating the temperature field data using the energy equation and power distribution.

[0036] The data-driven refined core physics-thermal coupling calculation method described in the present invention solves the technical problem of fast and stable refined thermal data acquisition. By reconstructing the calculation results of the low-fidelity CFD model, high-fidelity flow field data is obtained, thereby avoiding repeated high-fidelity CFD model calculations and reducing computing costs. The temperature field is reconstructed using high-fidelity energy equations, which reduces the possibility of non-physical solutions and improves the accuracy of simulation results. The method of predicting the flow velocity field that is less affected by power changes improves the reliability of data reconstruction and enhances the stability of the coupling calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is the main flow chart of the present invention;

[0038] Figure 2 Schematic diagram of the physical model and mesh model of the high-fidelity CFD model of the present invention;

[0039] Figure 3 Schematic diagram of the high-fidelity CFD grid model of the present invention that should be verified and analyzed by experiments;

[0040] Figure 4 Schematic diagram of a mesh mapping solution between a physical model and a CFD model of a 3×3 subassembly of the present invention;

[0041] Figure 5 Schematic diagram of the rough CFD meshing of the 3×3 subassembly of the present invention;

[0042] Figure 6 A schematic diagram of the process of online training of the data-driven model and the calculation process coupled with the physical model of the present invention;

[0043] Figure 7 Schematic diagram of the calculation process of coolant and fuel rod temperature data during the post-processing process of the MFM model of the present invention. DETAILED DESCRIPTION

[0044] like Figure 1-Figure 7 A data-driven, refined core physics-thermal coupling calculation method is shown, comprising the following steps:

[0045] Step 1: Establish a data acquisition module, which acquires the core data of the nuclear reactor and stores it in the database;

[0046] Step 2: Establish a model building module. The model building module obtains core data from the database and performs physical and thermal calculations of the core using the physical model and the high-fidelity CFD model, respectively. The physical model uses a structured grid, while the high-fidelity CFD model uses an unstructured grid.

[0047] The core neutron flux and core power distribution are obtained by physical calculations, and the coolant and fuel rod temperatures are obtained by thermal calculations;

[0048] The core data includes core structure, fuel rod physical parameters, coolant temperature and fuel rod temperature;

[0049] The physical model is a refined three-dimensional neutron transport physical model;

[0050] Physical calculations include calculating the core neutron flux and power distribution based on the core structure, fuel rod physical parameters, coolant temperature, and fuel rod temperature;

[0051] Thermal calculations include the calculation of coolant temperature and fuel rod temperature based on the core structure, flow field boundary conditions and power distribution.

[0052] In this embodiment, a three-dimensional refined neutron transport physical model is established for the target object. The physical model is usually established using a structured grid. The grid size and density are determined based on the calculation accuracy and cost requirements. The grid can usually be divided into vertical, fan, ring, row, and column modes, such as Figure 2 shown.

[0053] The core physical model can calculate the core neutron flux and core power distribution based on the fuel rod physical parameters, coolant temperature, fuel rod temperature, etc.; a refined high-fidelity CFD model is established for the target object. The high-fidelity CFD model is established using an unstructured grid with a very small grid size based on the calculation accuracy and cost requirements, such as Figure 3 As shown in Figure 1, high-fidelity CFD mesh models should be validated and analyzed experimentally. The CFD model can calculate coolant temperature and fuel rod temperature based on the target structure, flow field boundary conditions, and core power distribution.

[0054] Establish an interface between the physical model and the high-fidelity CFD model to transfer temperature and power data between the two models, and set the coupling control program between the two models, including controlling the physical model to transfer power distribution to the high-fidelity CFD model and controlling the high-fidelity CFD model to transfer temperature data to the physical model;

[0055] Depending on the complexity of the nuclear reactor, several coupled calculations are performed without converging. The calculation data of the high-fidelity CFD model and the power distribution calculated by the physical model are retained to obtain a high-fidelity flow field data set.

[0056] The coupling process specifically includes the following steps:

[0057] Step 2-1: The high-fidelity CFD model integrates and averages the temperature data of the CFD calculation grid overlaid on the physical grid according to the grid position of the physical model to obtain the average value;

[0058] Step 2-2: The high-fidelity CFD model transfers the average value to the corresponding grid of the physical model through the interface;

[0059] Step 2-3: The physical model transfers the power data of each grid to the corresponding grid of the high-fidelity CFD model through the interface.

[0060] Figure 4 A grid space mapping scheme for the physical model and high-fidelity CFD model of a 3×3 component is demonstrated. Based on this, a coupled data interface is established between the obtained physical model and the high-fidelity CFD model, and the temperature data from the thermal calculation and the power data from the physical calculation are transferred to each other.

[0061] The high-fidelity CFD model integrates and averages the temperature data of all CFD calculation grids covered by a grid of the physical program according to the corresponding position of the physical model grid to obtain the temperature data T of the physical program grid. nu，i ;

[0062]

[0063] Among them, T th，i，j is the temperature data of the jth CFD grid covered by the ith physical grid, T nu，i is the temperature data of the physical model grid.

[0064] The temperature data output interface transfers the temperature data of the physical program grid after CFD calculation to the corresponding grid of the physical program, while the power data output interface transfers the power of each grid of the physical program to the corresponding grid of the CFD program. The transfer method is lumped parameter:

[0065]

[0066] Among them, P th，i，j is the power data of the jth CFD fuel rod grid covered by the ith physical grid, P nu，i Power data for the physics program mesh.

[0067] In this embodiment, Figure 4 As shown in the figure, the coupling control program is used to control the data transmission during physical calculation and thermal calculation. In actual operation, it controls the physical calculation to transmit power distribution to the thermal calculation, and controls the thermal calculation to transmit fuel rod temperature and coolant temperature to the physical calculation.

[0068] Step 3: Establish a simplified CFD module. The simplified CFD module obtains a low-fidelity CFD model by roughening the network of the high-fidelity CFD model. The low-fidelity CFD model is used to perform calculations under the same boundary conditions according to the power distribution to obtain a low-fidelity flow field data set. The degree of roughening is determined according to the calculation accuracy and cost. Figure 5 Shown is the coarse CFD meshing of the 3×3 subassembly.

[0069] In this embodiment, the low-fidelity CFD model uses the same grid mapping scheme as the high-fidelity CFD model and performs low-fidelity CFD calculations based on the power distribution. The calculations use the same flow field boundary conditions as the high-fidelity CFD model to obtain the corresponding low-fidelity CFD flow field dataset.

[0070] Step 4: Establish a flow field data processing module. The flow field data processing module uses a high-fidelity autoencoder, a low-fidelity autoencoder, and a BP neural network model to reduce the dimension and map the high-fidelity flow field dataset and the low-fidelity flow field dataset, and establish a multi-fidelity flow field data reconstruction model MFM.

[0071] The multi-fidelity flow field data reconstruction model MFM is trained using high-fidelity flow field datasets and low-fidelity flow field datasets.

[0072] The specific steps include:

[0073] Step 4-1: Establish a high-fidelity autoencoder and a low-fidelity autoencoder respectively. Perform dimensionality reduction on the high-fidelity flow field dataset obtained in step 2 and the low-fidelity flow field dataset obtained in step 3 respectively.

[0074] Step 4-2: Establish a BP neural network model to map the high-fidelity flow field data and the low-fidelity flow field data after dimensionality reduction;

[0075] Step 4-3: Combine the high-fidelity autoencoder, the low-fidelity autoencoder, and the BP neural network model into a multi-fidelity flow field data reconstruction model MFM, where the input of the multi-fidelity flow field data reconstruction model MFM is the low-fidelity flow field data and the output is the predicted data of the high-fidelity flow field data;

[0076] Step 4-4: Use the high-fidelity flow field dataset and the low-fidelity flow field dataset to train the multi-fidelity flow field data reconstruction model MFM.

[0077] Step 5: Establish a multi-fidelity coupling calculation module. The multi-fidelity coupling calculation module embeds the low-fidelity CFD model and the multi-fidelity flow field data reconstruction model (MFM) into the coupling calculation process, thereby replacing the high-fidelity CFD model to calculate the flow field data. During the coupling calculation process, the power distribution is transferred to the low-fidelity CFD model to calculate the low-fidelity flow field data. The low-fidelity flow field data is processed using the multi-fidelity flow field data reconstruction model (MFM) to obtain the predicted data of the high-fidelity flow field data.

[0078] like Figure 6 As shown, the low-fidelity CFD calculation model and the multi-fidelity flow field data reconstruction model MFM are embedded in the coupled calculation process and replace the calculation process of the high-fidelity CFD model.

[0079] In the coupling process after embedding, after the physical calculation is completed, the power distribution is passed to the low-fidelity CFD model. The low-fidelity CFD model calculates the corresponding low-fidelity flow field data based on the power distribution and inputs the low-fidelity flow field data into the multi-fidelity flow field data reconstruction model MFM, thereby achieving the purpose of predicting high-fidelity flow field data.

[0080] Step 6: The back-substitution calculation module back-substitutes the predicted data into the high-fidelity CFD model and calculates the temperature field data based on the power distribution. This specifically includes back-substituting the predicted data into the energy equation of the high-fidelity CFD model and calculating the temperature field data using the energy equation and power distribution.

[0081] Figure 7 The calculation flow chart of coolant and fuel rod temperature data during the MFM post-processing process is shown.

[0082] Step 7: The iterative module uses the coupling control program according to the method in step 2 to transfer the temperature field data to the physical model and perform a new round of coupling calculation;

[0083] Repeat steps 5 and 6 until the coupled calculation reaches convergence.

[0084] In this embodiment, the data acquisition module, model construction module, simplified CFD module, flow field data processing module, multi-fidelity coupling calculation module, back-substitution calculation module and iteration module are all deployed in a cloud server cluster, and the data acquisition module, model construction module, simplified CFD module, flow field data processing module, multi-fidelity coupling calculation module, back-substitution calculation module and iteration module communicate with each other through the Internet.

[0085] The data-driven refined core physics-thermal coupling calculation method described in the present invention solves the technical problem of fast and stable refined thermal data acquisition. By reconstructing the calculation results of the low-fidelity CFD model, high-fidelity flow field data is obtained, thereby avoiding repeated high-fidelity CFD model calculations and reducing computing costs. The temperature field is reconstructed using high-fidelity energy equations, which reduces the possibility of non-physical solutions and improves the accuracy of simulation results. The method of predicting the flow velocity field that is less affected by power changes improves the reliability of data reconstruction and enhances the stability of the coupling calculation.

[0086] In the present invention, any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0087] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a model for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be a paper or other suitable medium on which the model can be printed, since the model can be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0088] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0089] Those skilled in the art will understand that all or part of the steps in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a model, and the model can be stored in a computer-readable storage medium. When the model is executed, it includes one or a combination of the steps of the method embodiment.

[0090] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0091] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A data-driven, refined core physics-thermal coupling calculation method, characterized by: The steps include: Step 1: Establish a data acquisition module, which acquires and stores the core data of the nuclear reactor; Step 2: Establish a model building module. The model building module obtains the core data from the data acquisition module and performs physical and thermal calculations of the core using the physical model and the high-fidelity CFD model, respectively. The physical model uses a structured grid, while the high-fidelity CFD model uses an unstructured grid. The core neutron flux and core power distribution are obtained by physical calculations, and the coolant and fuel rod temperatures are obtained by thermal calculations; Establish an interface between the physical model and the high-fidelity CFD model to transfer temperature and power data between the two models, and set the coupling control program between the two models, including controlling the physical model to transfer power distribution to the high-fidelity CFD model and controlling the high-fidelity CFD model to transfer temperature data to the physical model; Depending on the complexity of the nuclear reactor, several coupled calculations are performed without converging. The calculation data of the high-fidelity CFD model and the power distribution calculated by the physical model are retained to obtain a high-fidelity flow field data set and store it in the database. Step 3: Establish a simplified CFD module. The simplified CFD module obtains a low-fidelity CFD model by roughening the network of the high-fidelity CFD model. The low-fidelity CFD model is used to perform calculations based on the power distribution under the same boundary conditions to obtain a low-fidelity flow field dataset and store it in the database. Step 4: Establish a flow field data processing module. The flow field data processing module uses a high-fidelity autoencoder, a low-fidelity autoencoder, and a BP neural network model to reduce the dimension and map the high-fidelity flow field dataset and the low-fidelity flow field dataset, and establish a multi-fidelity flow field data reconstruction model MFM. The multi-fidelity flow field data reconstruction model MFM is trained using high-fidelity flow field datasets and low-fidelity flow field datasets. Step 5: Establish a multi-fidelity coupling calculation module. The multi-fidelity coupling calculation module embeds the low-fidelity CFD model and the multi-fidelity flow field data reconstruction model (MFM) into the coupling calculation process, thereby replacing the high-fidelity CFD model to calculate the flow field data. During the coupling calculation process, the power distribution is transferred to the low-fidelity CFD model to calculate the low-fidelity flow field data. The low-fidelity flow field data is processed using the multi-fidelity flow field data reconstruction model (MFM) to obtain the predicted data of the high-fidelity flow field data. Step 6: The back-substitution calculation module back-substitutes the predicted data into the high-fidelity CFD model and calculates the temperature field data based on the power distribution; Step 7: The iterative module uses the coupling control program according to the method in step 2 to transfer the temperature field data to the physical model and perform a new round of coupling calculation; Repeat steps 5 and 6 until the coupled calculation reaches convergence.

2. The data-driven refined core physics-thermal coupling calculation method according to claim 1, characterized in that: When executing step 2, the core data includes core structure, fuel rod physical parameters, coolant temperature, and fuel rod temperature; The physical model is a refined three-dimensional neutron transport physical model; Physical calculations include calculating the core neutron flux and power distribution based on the core structure, fuel rod physical parameters, coolant temperature, and fuel rod temperature; Thermal calculations include the calculation of coolant temperature and fuel rod temperature based on the core structure, flow field boundary conditions and power distribution.

3. The data-driven refined core physics-thermal coupling calculation method according to claim 1, characterized in that: When executing step 2, the coupling process specifically includes the following steps: Step 2-1: The high-fidelity CFD model integrates and averages the temperature data of the CFD calculation grid overlaid on the physical grid according to the grid position of the physical model to obtain the average value; Step 2-2: The high-fidelity CFD model transfers the average value to the corresponding grid of the physical model through the interface; Step 2-3: The physical model transfers the power data of each grid to the corresponding grid of the high-fidelity CFD model through the interface.

4. The data-driven refined core physics-thermal coupling calculation method according to claim 1, characterized in that: When executing step 3, the degree of roughening is determined based on the calculation accuracy and cost.

5. The data-driven refined core physics-thermal coupling calculation method according to claim 1, characterized in that: When executing step 4, the specific steps include: Step 4-1: Establish a high-fidelity autoencoder and a low-fidelity autoencoder respectively, and perform dimensionality reduction on the high-fidelity flow field dataset obtained in step 2 and the low-fidelity flow field dataset obtained in step 3 respectively; Step 4-2: Establish a BP neural network model to map the high-fidelity flow field data and the low-fidelity flow field data after dimensionality reduction; Step 4-3: Combine the high-fidelity autoencoder, the low-fidelity autoencoder, and the BP neural network model into a multi-fidelity flow field data reconstruction model MFM, where the input of the multi-fidelity flow field data reconstruction model MFM is the low-fidelity flow field data and the output is the predicted data of the high-fidelity flow field data; Step 4-4: Use the high-fidelity flow field dataset and the low-fidelity flow field dataset to train the multi-fidelity flow field data reconstruction model MFM.

6. The data-driven refined core physics-thermal coupling calculation method according to claim 1, characterized in that: When executing step 6, it specifically includes substituting the predicted data back into the energy equation of the high-fidelity CFD model, and calculating the temperature field data using the energy equation and power distribution.

Images