Rapid calculation method suitable for magnetic confinement nuclear fusion scraping layer target plate area
Through the combination of neural network model and base calculation program, the problem of slow calculation speed and lack of independent property rights in the existing technology is solved, and the rapid calculation and parameter prediction of the scraping layer target plate area is realized, forming closed-loop data Transmission improves computing efficiency.
Patent Information
- Application Number
- CN202411049567.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-15
AI Technical Summary
The existing magnetic constrained fusion integrated simulation platform lacks the program for fast calculation of scraping layer target plate area, and the calculation speed is slow, so it is impossible to achieve rapid simulation and real-time integrated calculation of scraping layer target plate area, and lacks independent property rights and automated processing.
The neural network model is combined with the base calculation program, and the neural network model parameters are adjusted by preprocessing experimental data until it meets the iteration conditions, and the rapid calculation and parameter prediction of the scraping layer target plate area are realized, forming a closed-loop data transmission.
The rapid automatic transmission and integration of plasma physical quantities in the target plate area of the scraping layer is realized, the calculation efficiency is improved, and the problems of slow operation speed and lack of independent property rights in the prior art are overcome.
Smart Images

Figure CN118965990A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of rapid calculation of scraped layer target plate area, and more specifically, it relates to a rapid calculation method for scraped layer target plate area suitable for magnetic confinement nuclear fusion. Background Art
[0002] Magnetic confinement fusion boundary physics plays an important role in magnetic confinement fusion technology. It is related to key issues such as plasma stability, impurity control, and heat emission, and is of great significance for achieving controlled nuclear fusion reactions. In magnetic confinement fusion research, high-energy plasma is accompanied by the outbreak of some magnetohydrodynamic instabilities (MHD), which leads to high-energy particles continuously bombarding the divertor target plate, thereby increasing the energy and heat load on the divertor target plate and causing damage to the target plate material. There is a scraping layer region (SOL) between the outermost closed magnetic surface and the first wall of the tokamak device, which can significantly reduce the contamination of the plasma core by impurities generated by the interaction between the plasma and the first wall material, and at the same time can better control the bombardment position of the discharged energy through the magnetic field of the scraping layer region. When the ions reach the surface of the divertor target plate, they recombine and neutralize with the surface electrons. According to the speed of the incoming ions, there are two main neutralization processes: fast ions are reflected into neutral atoms, and slow ions are absorbed and released again in the form of neutral molecules. These atoms and molecules are ionized in the boundary scraping layer and the cycle is repeated. In a stable state, the plasma is self-fueled through this recycling process, which mainly occurs in the divertor region. In addition, the recycling in the divertor region will form a dense cloud of neutral atoms and molecules in front of the target plate, thereby increasing the friction of neutral particles and the interaction between plasma neutral particles, thereby reducing the particle, momentum and energy flux flowing to the target plate. There is a high density of neutral particles in the divertor region, and the discharge of impurities and helium ash is crucial to controlling particles, because if these particles are not discharged in time, it will lead to dilution of deuterium-tritium DT fuel. Therefore, boundary physics plays a vital role in magnetic confinement fusion, and its importance is mainly reflected in the following aspects: 1) Plasma confinement: Boundary physics ensures that the plasma is stably confined in the magnetic confinement region by designing and optimizing the magnetic field structure. Reasonable boundary design can reduce the interaction between plasma and the wall, reduce energy loss and impurity generation, and improve the stability and duration of plasma; 2) Impurity control: Boundary physics can help control the impurity content in the plasma. By designing a suitable boundary structure and using appropriate wall materials, the process of impurities entering the plasma can be reduced, thereby improving the efficiency and controllability of the nuclear fusion reaction; 3) Heat and particle discharge: Boundary physics can also affect the discharge of heat and particles. Reasonable design of the boundary structure can effectively control the heat and particle flow at the plasma boundary, reduce damage and pollution to the wall, and extend the life of the device.
[0003] In the field of theoretical simulation of boundary physics of magnetic confinement fusion, the more common programs currently used include BOUT++, SOLPS (Scrape-Off Layer Plasma Simulator), EMC3-EIRENE, etc. BOUT++ is a program used for plasma boundary physics research, mainly used for the simulation of magnetic confinement fusion plasma. Its characteristics are that BOUT++ uses finite difference method to solve magnetohydrodynamics and transport equations of plasma boundary physics, can simulate the boundary layer structure and behavior of plasma, and the program has modular design; users can choose different physical models and boundary conditions according to their needs, flexibly customize simulation schemes, support parallel computing, and can run efficiently on multi-core processors or cluster systems to speed up simulation. It is widely used to study the physical processes of plasma boundary layer, including heat and particle transport, impurity control, etc. SOLPS (Scrape-Off Layer Plasma Simulator) is a program used to simulate the plasma boundary layer. It is mainly used to study the process of interaction between plasma and wall. Its characteristics are that it uses magnetohydrodynamics and transport equations to simulate the behavior of plasma in the boundary layer. It can consider the transport process of heat, particles and momentum, and can simulate complex physical processes such as flow, turbulence, heat transfer and impurity transport in the boundary layer of plasma. It supports multiple boundary conditions and wall models, etc. Different simulation settings can be selected according to specific research needs. SOLPS simulates the transport and diffusion process of impurities in the boundary layer, evaluates the effects of different impurity control methods, improves the purity of plasma, optimizes the control strategy of plasma boundary layer, improves the stability and performance of plasma, and provides support for the effective operation of fusion reactions. EMC3-EIRENE is a program used to simulate the interaction between the plasma boundary layer and the wall. It combines the functions of EMC3 (Edge Monte Carlo Code for Energetic Particles) and EIRENE (European Integrated Code for Modeling Divertor Plasma Transport). The program combines the Monte Carlo method and the transport equation to simulate the behavior of plasma in the boundary layer. It can consider the transport process of heat, particles and radiation, and can simulate the complex physical processes such as the flow, turbulence, heat transfer, impurity transport and radiation of plasma in the boundary layer, and supports a variety of boundary conditions and wall models. EMC3-EIRENE is widely used to study the behavior of plasma in the boundary layer, including heat and particle transport, turbulence generation and development, and radiation loss. By simulating the interaction process between plasma and the wall, it evaluates the influence of different wall materials on plasma behavior, optimizes the selection of wall materials, optimizes the control strategy of the plasma boundary layer, improves the stability and performance of plasma, and improves the efficiency of fusion reactions.
[0004] Magnetic confinement fusion plasma is a non-equilibrium multi-body system. The core-base-boundary plasma physics mechanism inside the device is extremely complex. It is a typical complex system with extremely complex nonlinear self-organization phenomena, which are difficult to quantitatively describe with existing theoretical models. Introducing machine learning methods to derive the model of the system from data and establish a model mapping relationship between experimental parameters and high confinement discharge may break through the existing theoretical difficulties in the study of high confinement discharge. It is also one of the current popular technical routes. Neural networks are used to control the properties and operating state of plasma in real time, including the control of parameters such as plasma temperature, density, and stability. The control efficiency and accuracy can be improved by learning and optimizing the feedback mechanism of the plasma control system. It also plays an important role in optimizing the boundary layer control of magnetic confinement fusion plasma. Modeling and predicting the physical process of the boundary layer through neural networks can optimize the boundary layer control strategy and improve the stability and performance of the plasma. It is also used to optimize the magnetic field design, including Including magnetic field topology, coil layout and other aspects, it can help design more effective magnetic field structure by learning a large amount of data and simulation results, improve the efficiency and performance of magnetic confinement fusion system, and also has application potential in plasma diagnosis; by processing and analyzing plasma diagnostic data through neural networks, useful information can be extracted to help researchers better understand the role of plasma behavior and properties; currently, large-scale tokamak devices such as ASDEX-Upgrade, J-TEXT, EAST, HL-2A, DIII-D and JET at home and abroad have applied machine learning-neural network as a tool to carry out a large number of rupture prediction studies and other rapid prediction calculation studies. The relevant research progress is rapid and the application prospects are very considerable.
[0005] The existing technology currently has the following problems:
[0006] (1) The existing integrated simulation platform for magnetic confinement fusion does not have a fast simulation program for the scraped layer target plate area. Numerical simulation programs such as the SOLPS program, an important tool for studying the boundary physics of the divertor of a fusion device, have a large demand for computer resources. Usually, one example requires 1,000 to 1,000,000 CPU hours of calculation, which cannot achieve fast calculation of the scraped layer target plate area.
[0007] (2) At present, my country lacks a boundary scraping layer simulation platform and program with independent intellectual property rights, and the calculation speed is slow, which is not conducive to the rapid integrated calculation and research of the integrated simulation of the entire magnetic confinement fusion device;
[0008] (3) At present, the input of physical parameters of the scraping layer target plate area program and the output of calculation results in the magnetic confinement fusion integrated simulation platform are mainly integrated by manually transmitting the input and output parameters required for the operation of the program in each area, which cannot achieve fast, efficient and real-time integration.
[0009] (4) At present, there is no proxy model for tokamak boundary physics and data development in the world, and it is impossible to achieve rapid feedback on physical problems at the boundary in experiments. Summary of the invention
[0010] The purpose of the present invention is to provide a fast calculation method for the target plate area of the scraped layer of magnetic confinement nuclear fusion, so as to overcome the shortcomings of the prior art such as slow operation speed, lack of independent property rights, and inability to be automated.
[0011] The above technical objectives of the present invention are achieved through the following technical solutions:
[0012] In a first aspect, the present application provides a fast calculation method for a target plate area of a scraped layer for magnetic confinement nuclear fusion, comprising the following specific steps:
[0013] S1, obtaining experimental data of boundary physical parameters of the scraped layer in the magnetic confinement nuclear fusion device in the past as first data, and preprocessing the first data to obtain second data corresponding to the first data;
[0014] S2, inputting the second data into a preset neural network model for processing to obtain third data corresponding to the second data;
[0015] S3, using the result of calculating the second data using the platform-based calculation program to verify the third data, and determining the verified third data as the fourth data;
[0016] S4, repeating steps S2-S3, and determining whether the two fourth data obtained in the last two iterations meet the preset iteration conditions. If the preset iteration conditions are not met, adjusting the model parameters of the neural network model, and continuing to repeat steps S2-S3 based on the neural network model after adjusting the model parameters, until the preset iteration conditions are met;
[0017] S5, determining the neural network model that meets the preset iteration conditions as the calculation model, and using the calculation model to perform parameter prediction of experimental data of the scraping layer target plate area in the magnetic confinement nuclear fusion device.
[0018] Based on the above technical solution, the present invention can also be improved as follows.
[0019] Furthermore, the above preprocessing includes data classification, label normalization and unified unit system processing.
[0020] Furthermore, the above preprocessing also includes:
[0021] Classifying the first data according to the presence or absence of impurity injection, the impurity injection position and the impurity injection rate;
[0022] The classified first data is scanned, and the impurity injection rate and the electron temperature and electron density at the bottom of the stage base within the scanning range are feature extracted using the clip algorithm, and linear interpolation fitting, similarity calculation and loss calculation are performed to obtain second data corresponding to the first data.
[0023] Furthermore, the above preset iteration condition is that the loss function is less than a preset order of magnitude, and the loss function is specifically:
[0024]
[0025] In the formula, C represents the value of the loss function, B represents the fourth data obtained in the last iteration, and A represents the fourth data obtained in an iteration adjacent to B.
[0026] Furthermore, the above preset order of magnitude is 10 -2 Order of magnitude.
[0027] In a second aspect, the present application provides a fast calculation system for a target plate area of a scraped layer for magnetic confinement nuclear fusion, which is applied to a fast calculation method for a target plate area of a scraped layer for magnetic confinement nuclear fusion in any one of the first aspects, comprising:
[0028] The first module is used to obtain the experimental data of the boundary physical parameters of the scraped layer in the magnetic confinement nuclear fusion device in the past as the first data, and pre-process the first data to obtain the second data corresponding to the first data;
[0029] The second module is used to input the second data into a preset neural network model for processing to obtain third data corresponding to the second data;
[0030] The third module is used to verify the third data by using the result of calculating the second data by the platform-based calculation program, and determine the verified third data as the fourth data;
[0031] The fourth module is used to repeatedly execute the second module and the third module, and determine whether the two fourth data obtained in the last two iterations meet the preset iteration conditions. If the preset iteration conditions are not met, the model parameters of the neural network model are adjusted, and the second module and the third module are continued to be repeatedly executed based on the neural network model after the model parameters are adjusted until the preset iteration conditions are met;
[0032] The fifth module is used to determine the neural network model that meets the preset iteration conditions as the calculation model, and use the calculation model to perform parameter prediction of the experimental data of the scraping layer target plate area in the magnetic confinement nuclear fusion device.
[0033] Furthermore, in the above-mentioned first module, preprocessing includes data classification, label normalization processing and unified unit system processing.
[0034] Furthermore, in the first module, the preprocessing further includes:
[0035] Classifying the first data according to the presence or absence of impurity injection, the impurity injection position and the impurity injection rate;
[0036] The classified first data is scanned, and the impurity injection rate and the electron temperature and electron density at the bottom of the stage base within the scanning range are feature extracted using the clip algorithm, and linear interpolation fitting, similarity calculation and loss calculation are performed to obtain second data corresponding to the first data.
[0037] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the methods in the first aspect when executing the computer program.
[0038] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any one of the methods in the first aspect.
[0039] Compared with the prior art, the present invention has at least the following beneficial effects:
[0040] In the present application, by adopting a reasonable and efficient neural network numerical algorithm, the KLBJ-NN program (computational model) for the scraping layer target plate area independently developed based on a neural network is used to solve the relevant parameters given in the boundary scraping layer area (such as the type of boundary impurity particle injection, injection rate, density, temperature, pressure, energy density, etc. of the boundary particles) as the source of plasma boundary data, and the output file of the KLBJ-NN program is used as the input file of the rapid integration calculation program of the base area, forming a closed loop with the physical quantities of the turbulence calculation and flux calculation in the core area and the boundary physical quantities selected in the scraping layer area, and the convolutional neural network method is used to quickly calculate the selected boundary key parameters, so as to achieve the purpose of quickly realizing the automatic and rapid transmission and integration of physical quantities such as plasma pressure, temperature, density, impurities, neutral particles, turbulence, heat flux, etc. in the scraping layer target area. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
[0042] Figure 1A method flow chart of a calculation method in an embodiment of the present invention;
[0043] Figure 2 A connection diagram of a computing system in an embodiment of the present invention;
[0044] Figure 3 Schematic diagram of the connection of electronic equipment in an embodiment of the present invention. DETAILED DESCRIPTION
[0045] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0046] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0047] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.
[0048] In the description of the embodiments of the present invention, "plurality" means at least 2.
[0049] Example 1: This example provides a fast calculation method for the target plate area of the scraping layer suitable for magnetic confinement nuclear fusion, such as Figure 1 As shown, the specific steps include:
[0050] S1, obtaining experimental data of boundary physical parameters of the scraped layer in the magnetic confinement nuclear fusion device in the past as first data, and preprocessing the first data to obtain second data corresponding to the first data.
[0051] Among them, the first data contains the characteristic parameters that have the greatest impact on the base structure: the electron density n at the top of the base e-ped , effective boundary charge number Z eff-ped , normalized specific pressure β N , plasma current I P , toroidal magnetic field strength B T , impurity injection type A, impurity injection rate υ, etc.
[0052] Optionally, the above preprocessing includes data classification, label normalization and unified unit system processing.
[0053] Among them, data classification includes the conversion of physical units corresponding to the input physical parameters into the International System of Units.
[0054] Furthermore, the above preprocessing also includes:
[0055] The first data is classified according to the rules of whether or not impurities are injected, the impurity injection position, and the impurity injection rate.
[0056] The classified first data is scanned, and the impurity injection rate and the electron temperature and electron density at the bottom of the stage base within the scanning range are feature extracted using the clip algorithm, and linear interpolation fitting, similarity calculation and loss calculation are performed to obtain second data corresponding to the first data.
[0057] S2, inputting the second data into a preset neural network model for processing to obtain third data corresponding to the second data.
[0058] S3, using the result of calculating the second data using a platform-based computing program to verify the third data, and determining the verified third data as the fourth data.
[0059] Among them, the construction of the neural network structure and the training of the data include: building a model framework based on the pytorch environment of the python programming language, using a convolutional neural network structure to train the second data, through specific feature integration, using the softmax activation function to generate the probability of the corresponding label, and adjusting the parameters through the back propagation algorithm and the gradient descent algorithm, learning the appropriate weights and biases, and to prevent overfitting, using the regularization strategy: adding the L1 regularization term to penalize the size of the weight to reduce the complexity of the neural network model, and the result of the model calculation output is used as the third data.
[0060] S4, repeat steps S2-S3, and determine whether the two fourth data obtained in the last two iterations meet the preset iteration conditions. If not, adjust the model parameters of the neural network model, and continue to repeat steps S2-S3 based on the neural network model after adjusting the model parameters until the preset iteration conditions are met.
[0061] Optionally, the above preset iteration condition is that the loss function is less than a preset order of magnitude, and the loss function is specifically:
[0062]
[0063] In the formula, C represents the value of the loss function, B represents the fourth data obtained in the last iteration, and A represents the fourth data obtained in an iteration adjacent to B.
[0064] Optionally, the above preset order of magnitude is 10 -2 Order of magnitude.
[0065] Among them, the calculation results of the neural network calculation program are checked with the calculation results of the platform-based calculation program, and the calculation result data that meets the set error is saved as the fourth data.
[0066] Specifically, the fourth data is used as an input file to use the neural network model to quickly solve and calculate the effective charge number Z at the fractal surface eff-sep and the electron temperature T e-sep And the effective charge number Z on the top of the base eff-core and the electron temperature T e-core And other physical parameters, adjust the hyperparameters of the neural network and update the fourth data. The hyperparameters include learning rate, regularization parameters, etc. During the training process, the best hyperparameter configuration is selected by the performance of the validation set data to avoid overfitting or underfitting problems.
[0067] S5, determining the neural network model that meets the preset iteration conditions as the calculation model, and using the calculation model to perform parameter prediction of experimental data of the scraping layer target plate area in the magnetic confinement nuclear fusion device.
[0068] Specifically, this embodiment adopts the scraping layer target plate area boundary neural network fast calculation program KLBJ-NN in the independent research and development platform, and the easily accessible and highly relevant and representative parameter experimental data that can be selected at the scraping layer include: macro parameters (B, Ip); off-target parameters (inflation position, particle type, inflation rate, inflation pulse time, exhaust position, exhaust rate) for training output; parameter impurity information: Z_eff, n_imp; main plasma: ne, Te, Ti. Geometric parameters; initial conditions (mid-plane plasma parameters and transport coefficients), DSOL zone plasma parameters (wall circulation coefficient, gas pressure, target plate temperature, etc.).
[0069] According to the off-target experimental results of impurity injection into the divertor of the HL-2A device, the radiation divertor experiments under N and Ne injection conditions were simulated and calculated using boundary programs such as SOLP, BOUT++, and EMC3-EIRENE. Then, some parameter scanning and predictive studies were carried out by changing physical parameters such as the type of injected impurities, injection rate, amount of injected particles, core boundary plasma density, upstream boundary plasma temperature, plasma transport coefficient, target plate particle recycling coefficient, etc., and a data set consisting of the corresponding numerical simulation results and the device discharge experimental measurement data was obtained.
[0070] Furthermore, in order to study the rapid calculation of the data set composed of the above-mentioned numerical simulation results and the device discharge experiment measurement data by the neural network fast calculation program of the scraping layer target plate area boundary, the numerical calculation results with high physical reduction are output, and the macro parameters (B, Ip) and off-target parameters (inflation position, particle type, inflation rate, inflation pulse time, exhaust position, exhaust rate) are classified and allocated as training set, validation set, and test set in a ratio of 10:1:10. The model is trained using the training set data. Through the back propagation algorithm, the neural network continuously adjusts the weights and biases according to the training set data to minimize the loss function to improve the accuracy of the model, and the data is smoothed to obtain a smooth fitting curve. In order to avoid overfitting and underfitting problems, the validation set data is used to adjust the hyperparameters of the neural network model. During the training process, the performance of the validation set data is used to select the best hyperparameter configuration. Finally, the test set data is used to evaluate the performance of the neural network model on unseen data. After the training is completed, the test set data is input into the model. By calculating the accuracy, precision, recall and other indicators of the model, the generalization ability of the neural network prediction fast calculation program of the scraping layer target plate area boundary is evaluated to ensure the accuracy of the model's calculation of parameters.
[0071] Embodiment 2: The present application embodiment provides a fast calculation system for the scraping layer target plate area of magnetic confinement nuclear fusion, which is applied to a fast calculation method for the scraping layer target plate area of magnetic confinement nuclear fusion in any one of the embodiments 1, such as Figure 2 As shown, including:
[0072] The first module is used to obtain the experimental data of the boundary physical parameters of the scraped layer in the magnetic confinement nuclear fusion device in the past as the first data, and pre-process the first data to obtain the second data corresponding to the first data.
[0073] Optionally, in the above-mentioned first module, preprocessing includes data classification, label normalization and unified unit system processing.
[0074] Optionally, in the first module, the preprocessing further includes:
[0075] The first data is classified according to the rules of whether or not impurities are injected, the impurity injection position, and the impurity injection rate.
[0076] The classified first data is scanned, and the impurity injection rate and the electron temperature and electron density at the bottom of the stage base within the scanning range are feature extracted using the clip algorithm, and linear interpolation fitting, similarity calculation and loss calculation are performed to obtain second data corresponding to the first data.
[0077] The second module is used to input the second data into a preset neural network model for processing to obtain third data corresponding to the second data.
[0078] The third module is used to verify the third data by using the result of calculating the second data by the platform-based computing program, and determine the verified third data as the fourth data.
[0079] The fourth module is used to repeatedly execute the second module and the third module, and determine whether the two fourth data obtained in the last two iterations meet the preset iteration conditions. If the preset iteration conditions are not met, the model parameters of the neural network model are adjusted, and the second module and the third module are continued to be repeatedly executed based on the neural network model after the model parameters are adjusted until the preset iteration conditions are met.
[0080] The fifth module is used to determine the neural network model that meets the preset iteration conditions as the calculation model, and use the calculation model to perform parameter prediction of the experimental data of the scraping layer target plate area in the magnetic confinement nuclear fusion device.
[0081] Embodiment 3: This embodiment of the present application provides an electronic device, such as Figure 3 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any method in Embodiment 1 is implemented.
[0082] Embodiment 4: The embodiment of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any method in Embodiment 1.
[0083] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A fast calculation method for the target plate area of the scraped layer suitable for magnetic confinement nuclear fusion, characterized in that: The specific steps include: S1, obtaining experimental data of boundary physical parameters of the scraped layer in the magnetic confinement nuclear fusion device in the past as first data, and preprocessing the first data to obtain second data corresponding to the first data; S2, inputting the second data into a preset neural network model for processing to obtain third data corresponding to the second data; S3, using a result of calculating the second data using a platform-based computing program to verify the third data, and determining the verified third data as the fourth data; S4, repeating steps S2-S3, and determining whether the two fourth data obtained in the last two iterations meet the preset iteration conditions, if not, adjusting the model parameters of the neural network model, and continuing to repeat steps S2-S3 based on the neural network model after adjusting the model parameters, until the preset iteration conditions are met; S5, determining the neural network model that meets the preset iteration conditions as the calculation model, and using the calculation model to perform parameter prediction of experimental data of the scraping layer target plate area in the magnetic confinement nuclear fusion device.
2. A fast calculation method for the target plate area of the scraped layer of magnetic confinement nuclear fusion according to claim 1, characterized in that: The preprocessing includes data classification, label normalization and unified unit system processing.
3. A fast calculation method for the target plate area of the scraped layer of magnetic confinement nuclear fusion according to claim 2, characterized in that: The pre-processing further comprises: Classifying the first data according to the presence or absence of impurity injection, the impurity injection position and the impurity injection rate; The first classified data is scanned, and the impurity injection rate and the electron temperature and electron density at the bottom of the stage base within the scanning range are feature extracted using the clip algorithm, and linear interpolation fitting, similarity calculation and loss calculation are performed to obtain second data corresponding to the first data.
4. A fast calculation method for the target plate area of the scraped layer of magnetic confinement nuclear fusion according to claim 1, characterized in that: The preset iteration condition is that the loss function is less than a preset order of magnitude, and the loss function is specifically: In the formula, C represents the value of the loss function, B represents the fourth data obtained in the last iteration, and A represents the fourth data obtained in an iteration adjacent to B.
5. A fast calculation method for the target plate area of the scraped layer of magnetic confinement nuclear fusion according to claim 4, characterized in that: The preset order of magnitude is 10 -2 Order of magnitude.
6. A fast calculation system for a scraped layer target plate area for magnetic confinement nuclear fusion, applied to a fast calculation method for a scraped layer target plate area for magnetic confinement nuclear fusion according to any one of claims 1 to 5, characterized in that: include: The first module is used to obtain experimental data of boundary physical parameters of the scraped layer in the magnetic confinement nuclear fusion device in the past as first data, and pre-process the first data to obtain second data corresponding to the first data; A second module is used to input the second data into a preset neural network model for processing to obtain third data corresponding to the second data; A third module is used to verify the third data by using the result of calculating the second data by the platform-based calculation program, and determine the verified third data as the fourth data; The fourth module is used to repeatedly execute the second module and the third module, and determine whether the two fourth data obtained in the last two iterations meet the preset iteration conditions. If the preset iteration conditions are not met, the model parameters of the neural network model are adjusted, and the second module and the third module are continued to be repeatedly executed based on the neural network model after the model parameters are adjusted until the preset iteration conditions are met; The fifth module is used to determine the neural network model that meets the preset iteration conditions as the calculation model, and use the calculation model to perform parameter prediction of the experimental data of the scraping layer target plate area in the magnetic confinement nuclear fusion device.
7. A fast calculation system for the target plate area of the scraped layer of magnetic confinement nuclear fusion according to claim 6, characterized in that: In the first module, the preprocessing includes data classification, label normalization and unified unit system processing.
8. A fast calculation method for the target plate area of the scraped layer of magnetic confinement nuclear fusion according to claim 6, characterized in that: In the first module, the preprocessing further includes: Classifying the first data according to the presence or absence of impurity injection, the impurity injection position and the impurity injection rate; The first classified data is scanned, and the impurity injection rate and the electron temperature and electron density at the bottom of the stage base within the scanning range are feature extracted using the clip algorithm, and linear interpolation fitting, similarity calculation and loss calculation are performed to obtain second data corresponding to the first data.
9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the method according to any one of claims 1 to 5 is implemented when the processor executes the computer program.
10. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable a computer to execute the method of any one of claims 1-5.
Citation Information
Cited By
Coil in vacuum chamber of magnetic confinement fusion device
CN116598025A
Internal Coils of the Vacuum Chamber of a Magnetic Confinement Fusion Device
CN116598025B
Laser-driven multi-shell fusion target design method and device based on hybrid optimization
CN119378273A
Intelligent control method for realizing operation of high-confinement mode of magnetic confinement nuclear fusion plasma
CN119670423A
SOLPS simulation method suitable for variable specific impulse magnetic plasma electric propulsion device
CN121211780A
