Multi-mode magnetic field intelligent closed-loop plasma confinement system and method

Abstract

The invention discloses a multi-mode magnetic field intelligent closed-loop plasma confinement system and method. The system comprises a multi-mode magnetic field generation module, a multi-dimensional plasma diagnosis module and a layered intelligent closed-loop control module. The multi-mode magnetic field generation module generates a steady-state, pulse and alternating composite magnetic field to realize adjustable space gradient; the multi-dimensional diagnosis module synchronously acquires electrical, optical and particle parameters, and ensures the accuracy through data fusion; and the hierarchical intelligent closed-loop control module generates an adjusting instruction to realize closed-loop optimization based on dynamic threshold and multi-algorithm scheduling. According to the method, the constraint performance is improved through the steps of initialization, data acquisition, state evaluation and closed-loop adjustment in combination with parameter self-optimization and three-level fault protection. According to the scheme, the plasma constraint time exceeds 300 s, the escape rate is lower than 5%, the adjustment response is smaller than or equal to 10 ms, the method adapts to multiple working conditions and can be popularized to the fields of nuclear fusion experiments and plasma processing, and the problems that a traditional system is poor in stability and lags in response are solved.

Description

Technical Field

[0001] This invention belongs to the field of nuclear fusion energy technology, specifically relating to a multi-mode magnetic field intelligent closed-loop confinement system and method for magnetically confined plasma, applicable to scenarios requiring long-term stable confinement of plasma, such as nuclear fusion experimental devices and plasma processing equipment. Background Technology

[0002] In fields such as nuclear fusion experiments and plasma processing, plasma confinement is a core technological component, and its performance directly determines the success rate of experiments and the effectiveness of industrial applications. Currently, most mainstream confinement systems employ single-mode magnetic fields (such as steady-state magnetic fields or pulsed magnetic fields only), which have significant technical limitations. Furthermore, existing publicly available technical solutions have not yet addressed key challenges, as detailed below:

[0003] Single-mode magnetic field performance bottleneck: Chinese patent CN108766234A (title: A steady-state magnetic field plasma confinement device) discloses a single steady-state magnetic field confinement system, which relies on a single set of niobium-titanium alloy superconducting coils to generate a 0.5-5T constant magnetic field. Although it can achieve basic confinement, due to the lack of a dynamic magnetic field adjustment mechanism, the measured plasma confinement time is only 180s at most, the electron density fluctuation reaches ±12%, and there is no high-frequency disturbance suppression design, which cannot meet the requirements of nuclear fusion experiments for long-term stable confinement (reference [1]); US patent US20220122451A1 (title: Pulse Magnetic Plasma Confinement System) proposes a single-pulse magnetic field confinement device, which generates a 1-2T, 5-50kHz pulsed magnetic field through 6 sets of hollow copper coils. However, due to the single magnetic field shape, the particle escape rate is as high as 18%, and no cooling channel is designed. After the coil works continuously for 30 minutes, the temperature exceeds 120℃, and it needs to be stopped to cool down, which aggravates the damage to the boundary material (reference [2]).

[0004] Diagnostic and control hysteresis defects: Traditional systems often rely on a single detection method. Japanese patent JP2021523178A (title: Plasma Parameter Detection Device) uses only a single set of electrostatic probes to collect plasma parameters, resulting in a data error rate exceeding 5%. Furthermore, this patent and most existing technologies use fixed thresholds to determine constraint states (e.g., a fixed electron density threshold of 5 × 10⁻⁶). 17 m -3 It cannot adapt to the initial density of 10. 16 -10 18 m -3 Constrained area pressure 10 -3 -10 -5When the operating conditions of Pa change, the adjustment response lag exceeds 100ms. To address this deficiency, this invention uses the PID algorithm of the hierarchical intelligent closed-loop control module to compress the linear deviation adjustment response time to ≤10ms. At the same time, it combines a 3-layer BP neural network algorithm to deal with nonlinear deviations (such as escape rate mutations of more than 10%), thus solving the dual problems of operating condition adaptation and response speed.

[0005] Insufficient safety and reusability: Existing technologies mostly provide single-level safety protection. The German patent (title: Plasma Confinement Safety System) with publication number DE102020133211A1 only achieves fault protection by cutting off the high-voltage power supply, without a backup magnetic field unit or emergency pressure relief design. The untimely handling of faults leads to a system failure rate of over 8%. Moreover, customized designs are common. For example, the confinement system for nuclear fusion experiments (with an inner diameter of 150mm) cannot be directly adapted to plasma processing scenarios (which require an inner diameter of 300mm). The reuse rate of core components is less than 30%, and the adaptation cost exceeds 500,000 yuan per instance, which restricts the industry's promotion.

[0006] With the advancement of nuclear fusion energy development (such as China's CFETR demonstration reactor) and high-end plasma applications (such as semiconductor plasma etching), the requirements for "long-term stability, intelligent control, safety and reliability" of confinement systems have increased significantly. Traditional technologies can no longer meet the needs, and there is an urgent need to develop new confinement systems that are multimodal collaborative, intelligent closed-loop, and safe and reliable. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to overcome the defects of the above-mentioned technologies and provide a multi-mode magnetic field intelligent closed-loop plasma confinement system and method.

[0008] To solve the above-mentioned technical problems, the present invention provides a multimodal magnetic field intelligent closed-loop plasma confinement system and method: a multimodal magnetic field intelligent closed-loop plasma confinement system, characterized in that:

[0009] include:

[0010] The multi-mode magnetic field generating module contains at least three independent magnetic field generating units, which can generate at least two magnetic field modes among steady-state magnetic field, pulsed magnetic field, and alternating magnetic field. The spatial gradient of the composite magnetic field can be adjusted through the magnetic field parameter collaborative control logic to dynamically enclose the plasma.

[0011] The multi-dimensional plasma diagnostic module is precisely matched with the magnetic field coverage area of ​​the multi-mode magnetic field generating module. It integrates an electrical detection sub-module, an optical detection sub-module, and a particle detection sub-module to simultaneously collect the microscopic state parameters and macroscopic constraint parameters of the plasma.

[0012] The hierarchical intelligent closed-loop control module establishes bidirectional communication with the multi-mode magnetic field generation module and the multi-dimensional plasma diagnostic module through a high-speed data bus. The hierarchical intelligent closed-loop control module receives and integrates multi-source diagnostic data, generates magnetic field adjustment commands based on multi-level control algorithms, and feeds them back to the multi-mode magnetic field generation module to dynamically adjust the intensity gradient, frequency characteristics, and spatial distribution of the composite magnetic field, thereby realizing real-time closed-loop optimization of the plasma confinement state.

[0013] As an improvement, the magnetic field generating unit of the multimodal magnetic field generating module includes:

[0014] The superconducting steady-state magnetic field unit adopts a nested double solenoid structure. The inner solenoid and the outer solenoid are respectively wound with niobium-titanium alloy superconducting coils and yttrium barium copper oxide high-temperature superconducting coils, which can generate a constant magnetic field of 0.5-8T. The current of the inner and outer solenoids can be independently adjusted to achieve magnetic field gradient control.

[0015] The high-frequency pulsed magnetic field unit consists of at least 6 sets of hollow copper coils evenly distributed along the circumference of the confined area. The inner wall of each set of coils is provided with a spiral cooling channel, through which a mixed cooling medium of liquid nitrogen and deionized water is introduced, which can generate a periodic pulsed magnetic field with a peak intensity of 1-3T and a pulse frequency of 10-100kHz.

[0016] The alternating magnetic field unit adopts a toroidal Helmholtz coil structure, with the coil material being oxygen-free copper. Driven by a frequency-modulated power supply, it can generate a sinusoidal alternating magnetic field with a frequency of 0.1-10kHz and an intensity of 0.05-0.5T.

[0017] The superconducting steady-state magnetic field unit, high-frequency pulsed magnetic field unit, and alternating magnetic field unit achieve parameter linkage through magnetic field collaborative control logic, forming a spatially adjustable composite confinement magnetic field.

[0018] As an improvement, the multimodal magnetic field generating module also includes a magnetic field calibration submodule, which comprises a three-dimensional Hall sensor array and a magnetic field simulation unit. The three-dimensional Hall sensor array is deployed at the boundary and center of the plasma confinement region to collect actual spatial distribution data of the composite magnetic field. The magnetic field simulation unit constructs a magnetic field simulation model based on the finite element analysis algorithm, compares the deviation between the actual spatial distribution data and the simulation data, and generates calibration instructions to adjust the current parameters of each magnetic field generating unit so that the deviation between the actual spatial distribution of the composite magnetic field and the preset distribution is less than 5%.

[0019] As an improvement, the multi-dimensional plasma diagnostic module includes:

[0020] The electrical detection submodule includes a multi-probe array and an electrostatic probe. The multi-probe array has 5-10 Langmuir probes arranged radially along the confinement region to collect the electron density of the plasma (10). 15-10 19 The electrostatic probe is used to collect the current-voltage characteristic curve of the plasma and calculate the collision frequency of the plasma.

[0021] The optical detection submodule includes a hyperspectral imager and a laser interferometer. The hyperspectral imager has a spectral resolution of not less than 0.1 nm and collects emission spectrum data of plasma to analyze particle types (hydrogen, helium, oxygen ions, etc.) and energy level distribution. The laser interferometer uses a 10.6 μm wavelength carbon dioxide laser to measure the spatial distribution uniformity of electron density in plasma, with a measurement accuracy better than 0.1%.

[0022] The particle detection submodule includes a charge-coupled detector (CCD) and a mass spectrometer. The CCD is used to capture morphological images of the plasma and analyze the boundary contours of the plasma within the confined region. The mass spectrometer is used to detect the types and energy distribution of escaping particles in the plasma, with a detection energy range of 1-100 keV.

[0023] The electrical detection submodule, optical detection submodule, and particle detection submodule have a synchronous acquisition frequency of 1-5kHz, and the timestamp alignment and redundancy verification of multi-source diagnostic data are achieved through a data fusion algorithm.

[0024] As an improvement, the hierarchical intelligent closed-loop control module includes:

[0025] The data preprocessing layer includes a noise suppression unit, an outlier removal unit, and a data normalization unit. The noise suppression unit uses a wavelet transform algorithm to denoise the diagnostic data, reducing noise errors caused by electromagnetic interference. The outlier removal unit identifies and removes abnormal diagnostic data based on the 3σ criterion. The data normalization unit maps diagnostic data of different dimensions to the 0-1 interval, achieving data standardization.

[0026] The constraint state evaluation layer includes a dynamic threshold generation unit and a multi-index evaluation unit. The dynamic threshold generation unit generates dynamic constraint thresholds based on historical plasma constraint data (at least 100 sets of effective constraint cycle data) and real-time operating parameters (such as constraint region pressure and initial plasma density) using a support vector machine algorithm. The multi-index evaluation unit constructs an evaluation matrix including constraint time, density uniformity, and escape rate, and uses the analytic hierarchy process (AHP) to calculate a comprehensive score of the current plasma constraint state. When the comprehensive score is lower than a preset threshold (80 points out of 100), the constraint state is deemed unqualified.

[0027] The instruction generation and execution layer includes an algorithm scheduling unit and an instruction issuing unit. The algorithm scheduling unit selects a PID control algorithm, a neural network algorithm, or a model predictive control algorithm to generate magnetic field adjustment instructions based on the constraint state evaluation results. Specifically, the PID algorithm is used when linear deviation is dominant (adjustment response time ≤ 10ms), the neural network algorithm (using a 3-layer BP neural network with a training sample size ≥ 5000 groups) is used when nonlinear deviation is dominant, and the model predictive control algorithm (prediction step size 5-10 acquisition cycles) is used under complex dynamic conditions. The instruction issuing unit transmits the adjustment instructions to the multimodal magnetic field generation module via an Ethernet / fiber optic hybrid communication link, with an instruction transmission delay of less than 1ms.

[0028] A multimodal magnetic field intelligent closed-loop plasma confinement method, applied to the multimodal magnetic field intelligent closed-loop plasma confinement system according to any one of claims 1-5, the method comprising the following steps:

[0029] S1: System initialization. Each magnetic field generating unit of the multi-mode magnetic field generating module starts according to the preset initial parameters. The superconducting steady-state magnetic field unit generates the basic confinement magnetic field (intensity 2-5T). The high-frequency pulse magnetic field unit and the alternating magnetic field unit operate at the initial frequency (20-50kHz, 1-5kHz) and intensity (1-2T, 0.1-0.3T) to form an initial composite magnetic field to envelop the plasma in the plasma confinement area.

[0030] S2: The electrical detection submodule, optical detection submodule, and particle detection submodule of the multi-dimensional plasma diagnostic module start up simultaneously, collect state parameters such as electron density, electron temperature, particle type, boundary profile, and energy distribution of escaped particles of the plasma, realize timestamp alignment and redundancy verification of multi-source data through data fusion algorithm, and transmit the processed diagnostic data to the hierarchical intelligent closed-loop control module.

[0031] S3: The constraint state evaluation layer of the hierarchical intelligent closed-loop control module determines whether the current constraint state of the plasma meets the standard based on the real-time threshold of the dynamic threshold generation unit and the comprehensive score of the multi-index evaluation unit. If the comprehensive score is ≥80 points, it is determined that the constraint meets the standard, and the current composite magnetic field parameters are maintained and continuously monitored. If the comprehensive score is <80 points, it is determined that the constraint does not meet the standard, and proceed to step S4.

[0032] S4: The instruction generation and execution layer of the hierarchical intelligent closed-loop control module schedules the corresponding control algorithm according to the reasons for non-compliance with constraints (linear deviation / nonlinear deviation / complex dynamic working conditions), generates magnetic field adjustment instructions (including current, frequency, and phase adjustment parameters of each magnetic field generating unit), and transmits them to the multi-mode magnetic field generating module; the multi-mode magnetic field generating module adjusts the composite magnetic field parameters according to the adjustment instructions, returns to step S2, and forms a closed-loop control cycle.

[0033] As an improvement, in step S2, the multi-dimensional plasma diagnostic module also includes a data validity verification step: cross-validation is performed between the electron density data of the electrical detection submodule and the laser interferometer data of the optical detection submodule. When the deviation between the two exceeds 10%, the backup probe and backup laser channel are activated to re-acquire data. At the same time, the mass spectrometer data of the particle detection submodule is matched and verified with the particle type data of the hyperspectral imager. When the particle type identification deviation exceeds 5%, the data calibration procedure is triggered to ensure the validity of the diagnostic data.

[0034] As an improvement, in step S4, the magnetic field adjustment command also includes magnetic field collaborative compensation logic: when adjusting the current of the superconducting steady-state magnetic field unit, the pulse frequency of the high-frequency pulse magnetic field unit is adjusted synchronously (the frequency adjustment amount is linearly related to the change in steady-state magnetic field strength, with a correlation coefficient of 0.5-2kHz / T) to avoid abrupt changes in the spatial gradient of the composite magnetic field; when adjusting the phase of the alternating magnetic field unit, the pulse phase of the high-frequency pulse magnetic field unit is corrected synchronously (the phase correction amount is 1 / 2-1 / 3 of the alternating magnetic field phase adjustment amount) to ensure the constraint stability of the composite magnetic field.

[0035] As an improvement, a constraint parameter self-optimization step is also included: after every 100 closed-loop control cycles, the constraint state evaluation layer of the hierarchical intelligent closed-loop control module extracts the optimal constraint parameters (composite magnetic field parameters and corresponding diagnostic data) for that cycle and stores them in the historical database; when the number of optimal parameter samples in the historical database reaches 500, a genetic algorithm is used to perform cluster analysis on the optimal parameters to generate different initial operating conditions (such as initial plasma density 10). 16 -10 18 m -3 Constrained area pressure 10 -3 -10 -5 The optimal parameter template under (Pa) can be matched according to the current initial working conditions when the system starts up, thereby shortening the time to achieve the constraint state (the time to achieve the standard is shortened by 30%-50%).

[0036] As an improvement, multi-level fault protection steps are also included:

[0037] Level 1 protection: The layered intelligent closed-loop control module monitors the current of each coil of the multi-mode magnetic field generating module in real time (triggered when the deviation exceeds the rated value ±10%), the coil temperature (triggered when it exceeds the critical temperature of the superconducting coil by 10K or the copper coil by 80℃), and the data acquisition integrity of the diagnostic module (triggered when the data loss rate exceeds 5%). When an abnormality is detected, an early warning signal is generated and the control algorithm parameters are adjusted for compensation.

[0038] Secondary protection: If the abnormality persists for more than 3 acquisition cycles after primary protection, the high-voltage power supply circuit of the multi-mode magnetic field generation module is cut off (cut-off response time ≤ 50ms), and the backup magnetic field unit (if any) is started to maintain basic constraints.

[0039] Level 3 protection: If the backup magnetic field unit fails to start or becomes abnormally aggravated (e.g., the coil temperature exceeds 100°C), an emergency pressure relief procedure is triggered to release the plasma in the confinement area. At the same time, abnormal data (including diagnostic data and magnetic field parameters 10 seconds before and after the fault) is recorded, and a fault report is generated for subsequent analysis and system optimization.

[0040] The advantages of this invention compared to existing technologies are as follows: Compared to traditional single-mode magnetic field confinement systems, it significantly improves plasma confinement performance through coordinated control of steady-state, pulsed, and alternating composite magnetic fields, combined with multi-dimensional diagnostics and intelligent closed-loop adjustment. The confinement time is extended from 100-200s to over 300s; electron density fluctuations and temperature deviations are reduced from ±10% and ±15% to ±5% and ±8%, respectively; the escape particle rate is reduced from 15%-20% to within 5%; and the magnetic field spatial distribution deviation is ≤5%, meeting the high requirements of scenarios such as nuclear fusion experiments. The system's intelligence level is significantly upgraded. Based on SVM to generate dynamic thresholds, combined with AHP multi-index evaluation, the accuracy of confinement state determination exceeds 99%. Hierarchical scheduling of PID, neural network, and other algorithms improves adjustment accuracy from ±8% to ±3%. Stability can be restored within one acquisition cycle in sudden scenarios. Furthermore, parameter templates can be generated through genetic algorithms, shortening the start-up time by 50%. In terms of safety, the three-level fault protection reduces the processing response time from 100ms to within 50ms, lowering the system failure rate by 60%. Multi-source data cross-validation reduces the data anomaly rate from 5% to below 0.1%, avoiding erroneous adjustments. Furthermore, the modular design reduces adaptation costs by 30%-40%, enabling its application in multiple fields such as plasma processing. The core technology can also be transferred to equipment such as particle accelerators, driving industry upgrades and providing crucial support for the development of nuclear fusion energy. Attached Figure Description

[0041] Figure 1 This is a system composition framework diagram of a multimodal magnetic field intelligent closed-loop plasma confinement system and method according to the present invention.

[0042] Figure 2 This is a flowchart of the intelligent closed-loop control of a multimodal magnetic field intelligent closed-loop plasma confinement system and method according to the present invention. Detailed Implementation

[0043] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings, which illustrate embodiments of the present application. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of this application will be thorough and complete.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0045] It is understood that spatial relation terms such as "below," "under," "below," "below," "above," "over," etc., can be used here to describe the relationship between one element or feature shown in the figure and other elements or features. It should be understood that, in addition to the orientation shown in the figure, spatial relation terms also include different orientations of the device in use and operation. For example, if the device in the figure is flipped, the element or feature described as "below" or "under" or "below" of the other element or feature will be oriented "over" the other element or feature. Therefore, the exemplary terms "below" and "under" can include both upper and lower orientations. Furthermore, the device may also include other orientations, such as being rotated 90 degrees or other orientations, and the spatial descriptive terms used herein will be interpreted accordingly.

[0046] It should be noted that when one element is considered to be "connected" to another element, it can be directly connected to the other element or connected to the other element through an intermediary element. In the following embodiments, "connection" should be understood as "electrical connection," "communication connection," etc., if the connected circuits, modules, units, etc., have the transmission of electrical signals or data between them.

[0047] When used herein, the singular forms of “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising,” “including,” or “having,” etc., specify the presence of the stated feature, whole, step, operation, component, part, or combination thereof, but do not preclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof.

[0048] Referring to the accompanying drawings, a multimodal magnetic field intelligent closed-loop plasma confinement system and method are disclosed. The multimodal magnetic field intelligent closed-loop plasma confinement system is characterized by:

[0049] include:

[0050] The multi-mode magnetic field generating module contains at least three independent magnetic field generating units, which can generate at least two magnetic field modes among steady-state magnetic field, pulsed magnetic field, and alternating magnetic field. The spatial gradient of the composite magnetic field can be adjusted through the magnetic field parameter collaborative control logic to dynamically enclose the plasma.

[0051] The multi-dimensional plasma diagnostic module is precisely matched with the magnetic field coverage area of ​​the multi-mode magnetic field generating module. It integrates an electrical detection sub-module, an optical detection sub-module, and a particle detection sub-module to simultaneously collect the microscopic state parameters and macroscopic constraint parameters of the plasma.

[0052] The hierarchical intelligent closed-loop control module establishes bidirectional communication with the multi-mode magnetic field generation module and the multi-dimensional plasma diagnostic module through a high-speed data bus. The hierarchical intelligent closed-loop control module receives and integrates multi-source diagnostic data, generates magnetic field adjustment commands based on multi-level control algorithms, and feeds them back to the multi-mode magnetic field generation module to dynamically adjust the intensity gradient, frequency characteristics, and spatial distribution of the composite magnetic field, thereby realizing real-time closed-loop optimization of the plasma confinement state.

[0053] As an improvement, the magnetic field generating unit of the multimodal magnetic field generating module includes:

[0054] The superconducting steady-state magnetic field unit adopts a nested double solenoid structure. The inner solenoid and the outer solenoid are respectively wound with niobium-titanium alloy superconducting coils and yttrium barium copper oxide high-temperature superconducting coils, which can generate a constant magnetic field of 0.5-8T. The current of the inner and outer solenoids can be independently adjusted to achieve magnetic field gradient control.

[0055] The high-frequency pulsed magnetic field unit consists of at least 6 sets of hollow copper coils evenly distributed along the circumference of the confined area. The inner wall of each set of coils is provided with a spiral cooling channel, through which a mixed cooling medium of liquid nitrogen and deionized water is introduced, which can generate a periodic pulsed magnetic field with a peak intensity of 1-3T and a pulse frequency of 10-100kHz.

[0056] The alternating magnetic field unit adopts a toroidal Helmholtz coil structure, with the coil material being oxygen-free copper. Driven by a frequency-modulated power supply, it can generate a sinusoidal alternating magnetic field with a frequency of 0.1-10kHz and an intensity of 0.05-0.5T.

[0057] The superconducting steady-state magnetic field unit, high-frequency pulsed magnetic field unit, and alternating magnetic field unit achieve parameter linkage through magnetic field collaborative control logic, forming a spatially adjustable composite confinement magnetic field.

[0058] As an improvement, the multimodal magnetic field generating module also includes a magnetic field calibration submodule, which comprises a three-dimensional Hall sensor array and a magnetic field simulation unit. The three-dimensional Hall sensor array is deployed at the boundary and center of the plasma confinement region to collect actual spatial distribution data of the composite magnetic field. The magnetic field simulation unit constructs a magnetic field simulation model based on the finite element analysis algorithm, compares the deviation between the actual spatial distribution data and the simulation data, and generates calibration instructions to adjust the current parameters of each magnetic field generating unit so that the deviation between the actual spatial distribution of the composite magnetic field and the preset distribution is less than 5%.

[0059] As an improvement, the multi-dimensional plasma diagnostic module includes:

[0060] The electrical detection submodule includes a multi-probe array and an electrostatic probe. The multi-probe array has 5-10 Langmuir probes arranged radially along the confinement region to collect the electron density of the plasma (10). 15 -10 19 The electrostatic probe is used to collect the current-voltage characteristic curve of the plasma and calculate the collision frequency of the plasma.

[0061] The optical detection submodule includes a hyperspectral imager and a laser interferometer. The hyperspectral imager has a spectral resolution of not less than 0.1 nm and collects emission spectrum data of plasma to analyze particle types (hydrogen, helium, oxygen ions, etc.) and energy level distribution. The laser interferometer uses a 10.6 μm wavelength carbon dioxide laser to measure the spatial distribution uniformity of electron density in plasma, with a measurement accuracy better than 0.1%.

[0062] The particle detection submodule includes a charge-coupled detector (CCD) and a mass spectrometer. The CCD is used to capture morphological images of the plasma and analyze the boundary contours of the plasma within the confined region. The mass spectrometer is used to detect the types and energy distribution of escaping particles in the plasma, with a detection energy range of 1-100 keV.

[0063] The electrical detection submodule, optical detection submodule, and particle detection submodule have a synchronous acquisition frequency of 1-5kHz, and the timestamp alignment and redundancy verification of multi-source diagnostic data are achieved through a data fusion algorithm.

[0064] As an improvement, the hierarchical intelligent closed-loop control module includes:

[0065] The data preprocessing layer includes a noise suppression unit, an outlier removal unit, and a data normalization unit. The noise suppression unit uses a wavelet transform algorithm to denoise the diagnostic data, reducing noise errors caused by electromagnetic interference. The outlier removal unit identifies and removes abnormal diagnostic data based on the 3σ criterion. The data normalization unit maps diagnostic data of different dimensions to the 0-1 interval, achieving data standardization.

[0066] The constraint state evaluation layer includes a dynamic threshold generation unit and a multi-index evaluation unit. The dynamic threshold generation unit generates dynamic constraint thresholds based on historical plasma constraint data (at least 100 sets of effective constraint cycle data) and real-time operating parameters (such as constraint region pressure and initial plasma density) using a support vector machine algorithm. The multi-index evaluation unit constructs an evaluation matrix including constraint time, density uniformity, and escape rate, and uses the analytic hierarchy process (AHP) to calculate a comprehensive score of the current plasma constraint state. When the comprehensive score is lower than a preset threshold (80 points out of 100), the constraint state is deemed unqualified.

[0067] The instruction generation and execution layer includes an algorithm scheduling unit and an instruction issuing unit. The algorithm scheduling unit selects a PID control algorithm, a neural network algorithm, or a model predictive control algorithm to generate magnetic field adjustment instructions based on the constraint state evaluation results. Specifically, the PID algorithm is used when linear deviation is dominant (adjustment response time ≤ 10ms), the neural network algorithm (using a 3-layer BP neural network with a training sample size ≥ 5000 groups) is used when nonlinear deviation is dominant, and the model predictive control algorithm (prediction step size 5-10 acquisition cycles) is used under complex dynamic conditions. The instruction issuing unit transmits the adjustment instructions to the multimodal magnetic field generation module via an Ethernet / fiber optic hybrid communication link, with an instruction transmission delay of less than 1ms.

[0068] A multimodal magnetic field intelligent closed-loop plasma confinement method, applied to the multimodal magnetic field intelligent closed-loop plasma confinement system according to any one of claims 1-5, the method comprising the following steps:

[0069] S1: System initialization. Each magnetic field generating unit of the multi-mode magnetic field generating module starts according to the preset initial parameters. The superconducting steady-state magnetic field unit generates the basic confinement magnetic field (intensity 2-5T). The high-frequency pulse magnetic field unit and the alternating magnetic field unit operate at the initial frequency (20-50kHz, 1-5kHz) and intensity (1-2T, 0.1-0.3T) to form an initial composite magnetic field to envelop the plasma in the plasma confinement area.

[0070] S2: The electrical detection submodule, optical detection submodule, and particle detection submodule of the multi-dimensional plasma diagnostic module start up simultaneously, collect state parameters such as electron density, electron temperature, particle type, boundary profile, and energy distribution of escaped particles of the plasma, realize timestamp alignment and redundancy verification of multi-source data through data fusion algorithm, and transmit the processed diagnostic data to the hierarchical intelligent closed-loop control module.

[0071] S3: The constraint state evaluation layer of the hierarchical intelligent closed-loop control module determines whether the current constraint state of the plasma meets the standard based on the real-time threshold of the dynamic threshold generation unit and the comprehensive score of the multi-index evaluation unit. If the comprehensive score is ≥80 points, it is determined that the constraint meets the standard, and the current composite magnetic field parameters are maintained and continuously monitored. If the comprehensive score is <80 points, it is determined that the constraint does not meet the standard, and proceed to step S4.

[0072] S4: The instruction generation and execution layer of the hierarchical intelligent closed-loop control module schedules the corresponding control algorithm according to the reasons for non-compliance with constraints (linear deviation / nonlinear deviation / complex dynamic working conditions), generates magnetic field adjustment instructions (including current, frequency, and phase adjustment parameters of each magnetic field generating unit), and transmits them to the multi-mode magnetic field generating module; the multi-mode magnetic field generating module adjusts the composite magnetic field parameters according to the adjustment instructions, returns to step S2, and forms a closed-loop control cycle.

[0073] As an improvement, in step S2, the multi-dimensional plasma diagnostic module also includes a data validity verification step: cross-validation is performed between the electron density data of the electrical detection submodule and the laser interferometer data of the optical detection submodule. When the deviation between the two exceeds 10%, the backup probe and backup laser channel are activated to re-acquire data. At the same time, the mass spectrometer data of the particle detection submodule is matched and verified with the particle type data of the hyperspectral imager. When the particle type identification deviation exceeds 5%, the data calibration procedure is triggered to ensure the validity of the diagnostic data.

[0074] As an improvement, in step S4, the magnetic field adjustment command also includes magnetic field collaborative compensation logic: when adjusting the current of the superconducting steady-state magnetic field unit, the pulse frequency of the high-frequency pulse magnetic field unit is adjusted synchronously (the frequency adjustment amount is linearly related to the change in steady-state magnetic field strength, with a correlation coefficient of 0.5-2kHz / T) to avoid abrupt changes in the spatial gradient of the composite magnetic field; when adjusting the phase of the alternating magnetic field unit, the pulse phase of the high-frequency pulse magnetic field unit is corrected synchronously (the phase correction amount is 1 / 2-1 / 3 of the alternating magnetic field phase adjustment amount) to ensure the constraint stability of the composite magnetic field.

[0075] As an improvement, a constraint parameter self-optimization step is also included: after every 100 closed-loop control cycles, the constraint state evaluation layer of the hierarchical intelligent closed-loop control module extracts the optimal constraint parameters (composite magnetic field parameters and corresponding diagnostic data) for that cycle and stores them in the historical database; when the number of optimal parameter samples in the historical database reaches 500, a genetic algorithm is used to perform cluster analysis on the optimal parameters to generate different initial operating conditions (such as initial plasma density 10). 16 -10 18 m -3 Constrained area pressure 10 -3 -10 -5 The optimal parameter template under (Pa) can be matched according to the current initial working conditions when the system starts up, thereby shortening the time to achieve the constraint state (the time to achieve the standard is shortened by 30%-50%).

[0076] As an improvement, multi-level fault protection steps are also included:

[0077] Level 1 protection: The layered intelligent closed-loop control module monitors the current of each coil of the multi-mode magnetic field generating module in real time (triggered when the deviation exceeds the rated value ±10%), the coil temperature (triggered when it exceeds the critical temperature of the superconducting coil by 10K or the copper coil by 80℃), and the data acquisition integrity of the diagnostic module (triggered when the data loss rate exceeds 5%). When an abnormality is detected, an early warning signal is generated and the control algorithm parameters are adjusted for compensation.

[0078] Secondary protection: If the abnormality persists for more than 3 acquisition cycles after primary protection, the high-voltage power supply circuit of the multi-mode magnetic field generation module is cut off (cut-off response time ≤ 50ms), and the backup magnetic field unit (if any) is started to maintain basic constraints.

[0079] Level 3 protection: If the backup magnetic field unit fails to start or becomes abnormally aggravated (e.g., the coil temperature exceeds 100°C), an emergency pressure relief procedure is triggered to release the plasma in the confinement area. At the same time, abnormal data (including diagnostic data and magnetic field parameters 10 seconds before and after the fault) is recorded, and a fault report is generated for subsequent analysis and system optimization.

[0080] I. Overview of Implementation Examples:

[0081] This embodiment addresses the plasma confinement scenario in nuclear fusion experiments by constructing a multi-modal magnetic field intelligent closed-loop plasma confinement system, achieving 10 17 -10 18 m -3 The plasma is stably confined with an electron density and an electron temperature of 5-8 eV, a confinement time of not less than 300 s, and an escape particle rate controlled within 5%. The system hardware selection, parameter configuration, and method execution flow are all based on the technical solutions of claims 1-10, which are described in detail below.

[0082] II. System Hardware Composition and Parameter Configuration:

[0083] 2.1 Multimodal magnetic field generation module:

[0084] 2.1.1 Superconducting steady-state magnetic field unit:

[0085] The system employs a nested double-sole structure. The inner solenoid is a 0.8mm diameter niobium-titanium alloy superconducting coil (critical temperature 9.2K), wound with 1200 turns, and has an inner diameter of 150mm and an outer diameter of 200mm. The outer solenoid is a 2mm wide, 0.5mm thick yttrium barium copper oxide (YBCO) high-temperature superconducting tape, wound with 800 turns, and has an inner diameter of 210mm and an outer diameter of 260mm. Powered by a superconducting power supply (model: Cryomagnetics LM5000), the inner solenoid current is adjustable from 0-300A, and the outer solenoid current is adjustable from 0-200A, generating a constant magnetic field of 0.5-8T. The initial basic constraint magnetic field is set to 3.5T (inner solenoid current 220A, outer solenoid current 150A).

[0086] 2.1.2 High-frequency pulsed magnetic field unit:

[0087] It consists of 6 groups of hollow copper coils evenly distributed circumferentially along the constrained area (with an included angle of 60° between adjacent coils). Each group of coils is wound with oxygen-free copper tubing with a diameter of 5mm and 50 turns. The inner diameter of the coil is 280mm and the outer diameter is 320mm. The diameter of the spiral cooling channel on the inner wall is 2mm. Driven by a pulse power supply (model: TREK609B-6), the cooling system is circulated with a mixture of liquid nitrogen and deionized water at a volume ratio of 1:3 (flow rate 5L / min, inlet temperature -20℃), which can generate a periodic pulsed magnetic field with a peak intensity of 1-3T and a pulse frequency of 10-100kHz. The initial parameters are set to a peak intensity of 1.8T and a pulse frequency of 35kHz.

[0088] 2.1.3 Alternating magnetic field unit:

[0089] It employs a double-ring Helmholtz coil structure, with the coil material being oxygen-free copper (99.99% purity). Each coil has 200 turns, a coil radius of 180mm, and a spacing of 180mm between the two coils. Driven by a frequency-modulated power supply (model: Agilent 33522A), it can generate a sinusoidal alternating magnetic field with a frequency of 0.1-10kHz and an intensity of 0.05-0.5T. The initial parameters are set to a frequency of 3kHz and an intensity of 0.2T.

[0090] 2.1.4 Magnetic Field Calibration Submodule:

[0091] The three-dimensional Hall sensor array uses an HMC1043 magnetoresistive sensor with a total of 9 detection points (1 at the center of the constraint area and 8 at the boundary, spaced 50mm apart), a measurement range of ±10T, and an accuracy of ±0.1%. The magnetic field simulation unit is based on a finite element model constructed using COMSOL Multiphysics software, with a mesh generation accuracy of 1mm and a simulation step size of 0.1s. The actual data collected by the sensor is compared with the simulation data in real time. When the deviation exceeds 5%, a calibration command is generated to adjust the current of each magnetic field unit (such as the solenoid current ±5A and the pulse frequency ±1kHz) to ensure the accuracy of the spatial distribution of the composite magnetic field.

[0092] 2.2 Multi-dimensional Plasma Diagnostic Module:

[0093] 2.2.1 Electrical Detection Submodule:

[0094] A multi-probe array consists of seven Langmuir probes (model: DL-600) arranged radially along the constrained region, with a probe spacing of 30 mm and a probe tip diameter of 0.5 mm. Electron density (10⁻⁶) is collected via a data acquisition card (NIPCIe-6363). 15 -10 19 m -3 Electron temperature (1-10 eV) and space potential distribution were measured at a sampling rate of 1 MHz. An ESP-1000 electrostatic probe was used to acquire plasma current-voltage characteristic curves (voltage range -50V to +50V, current range -10mA to +10mA). The collision frequency (10 eV) was calculated using MATLAB software. 6 -10 8 Hz).

[0095] 2.2.2 Optical Inspection Submodule:

[0096] The hyperspectral imager used is the Hyperspec VNIR, with a spectral range of 400-1000 nm, a spectral resolution of 0.1 nm, and a spatial resolution of 0.5 mm. Plasma emission spectra are acquired using a fiber optic probe (2 mm diameter) to analyze the energy level distribution of hydrogen (Hα 656.3 nm), helium (HeI 587.6 nm), and oxygen (OI 777.2 nm) ions. A 10.6 μm wavelength carbon dioxide laser (10 W power) is used for the laser interferometer. The spatial uniformity of electron density distribution is calculated using interference fringe analysis software, with a measurement accuracy better than 0.1%.

[0097] 2.2.3 Particle Detection Submodule:

[0098] The CCD used is a Baslerac A2500-14uc, with a resolution of 2592×1944 pixels and a frame rate of 30fps. It captures plasma morphology images using a neutral density filter (ND1000) and extracts boundary contours (error ±1mm). The mass spectrometer used is a HidenHPR-60, with a detection energy range of 1-100keV and a mass resolution of m / Δm=500, recording escaped particles (such as H+) in real time. + He 2+ The types and energy distribution of ).

[0099] All detection submodules achieve synchronous acquisition of 1-5kHz (3kHz in this embodiment) through a clock synchronizer (model SRSDS345), and use Kalman filtering algorithm for data fusion to complete timestamp alignment (deviation ≤1μs) and redundancy verification (abnormal data rejection rate ≤0.1%).

[0100] 2.3 Hierarchical Intelligent Closed-Loop Control Module:

[0101] The hardware utilizes an industrial control computer (CPU: Intel Core i7-12700K, 32GB RAM, 1TB SSD), connected to the magnetic field generation module and diagnostic module via a high-speed data bus (Ethernet + fiber optic hybrid link, 10Gbps bandwidth), with a command transmission latency of less than 1ms. The software is developed based on the LabVIEW platform, and the functional layers are implemented as follows:

[0102] 2.3.1 Data Preprocessing Layer:

[0103] The noise suppression unit uses a wavelet transform algorithm (db4 wavelet basis, 5 decomposition levels) to reduce noise in the diagnostic data, mitigating errors caused by electromagnetic interference (such as power supply noise from the magnetic field unit) (the signal-to-noise ratio is improved by 20dB after noise reduction). The outlier removal unit identifies and removes outlier data (such as abrupt changes caused by poor probe contact) that exceed the range [μ-3σ, μ+3σ] based on the 3σ criterion (σ is the data standard deviation). The data normalization unit uses the min-max normalization formula (x'=(x-min) / (max-min)) to map data of different dimensions such as electron density and temperature to the 0-1 interval.

[0104] 2.3.2 Constraint State Evaluation Layer:

[0105] The dynamic threshold generation unit is based on 100 sets of historical constraint periodic data (covering an initial density of 10). 16 -10 18 m -3 Pressure 10 -3 -10 -5(Pa operating condition), the model is trained using the Support Vector Machine (SVM) algorithm to generate dynamic thresholds in real time (such as the electron density threshold of 1.2 × 10⁻⁶). 17 -8×10 17 m -3 The escape rate threshold is 3%-5%. A multi-indicator evaluation matrix is ​​constructed (constraint time weight 0.4, density uniformity weight 0.3, escape rate weight 0.3). The analytic hierarchy process (AHP) is used to calculate the comprehensive score (out of 100). When the score is ≥80, the constraint is considered to be met.

[0106] 2.3.3 Instruction Generation and Execution Layer:

[0107] The algorithm scheduling unit selects the appropriate algorithm based on the type of deviation: when linear deviation (e.g., electron density deviating from the threshold by 5%-10%) dominates, the PID algorithm is activated (proportional coefficient Kp = 2.5, integral coefficient Ki = 0.8, derivative coefficient Kd = 0.3, adjustment response time ≤ 10ms); when nonlinear deviation (e.g., escape rate mutation exceeding 10%) dominates, a 3-layer BP neural network algorithm is activated (5 nodes in the input layer, 10 nodes in the hidden layer, 3 nodes in the output layer, 5000 training samples, 1000 iterations); when complex dynamic conditions occur (e.g., pressure fluctuation in the constrained area ±20%), a model predictive control algorithm is activated (prediction step size of 8 acquisition cycles, control time domain of 5 cycles). The command issuing unit transmits adjustment commands (e.g., internal solenoid current +3A, pulse frequency +2kHz) to the magnetic field generating module via the bus to achieve parameter adjustment.

[0108] III. Constraint Method Execution Flow:

[0109] 3.1 Step S1: System Initialization:

[0110] The multi-mode magnetic field generation module is activated. The superconducting steady-state magnetic field unit generates a 3.5T basic magnetic field according to the initial parameters (inner solenoid 220A, outer solenoid 150A); the high-frequency pulsed magnetic field unit operates at a peak intensity of 1.8T and a frequency of 35kHz; the alternating magnetic field unit operates at a frequency of 3kHz and an intensity of 0.2T. The three units are superimposed through coordinated control logic to form an initial composite magnetic field, which generates a plasma (initial density 5×10⁻⁶) within the confinement region (diameter 150mm, length 300mm). 17 m -3 The coils were enclosed in a confined manner at a temperature of 6 eV. Simultaneously, a superconducting cooling system (model: Sumitomo F-70) was activated to lower the temperature of the niobium-titanium coil to 4.2 K and the YBCO coil to 77 K.

[0111] 3.2 Step S2: Multi-dimensional diagnostic data acquisition and processing:

[0112] The multi-dimensional plasma diagnostic module was activated, and the seven Langmuir probes of the electrical detection submodule collected the electron density (real-time value 5.2 × 10⁻⁶). 17 m -3 The electron temperature (6.1 eV) was measured, and the IV curve was acquired using an electrostatic probe to calculate the collision frequency (2.5 × 10⁻⁶). 7 The optical detection submodule uses a hyperspectral imager to acquire the intensity of Hα and HeI spectral lines, and a laser interferometer to measure density uniformity (deviation 2.3%). The particle detection submodule uses a CCD to capture the plasma boundary profile (diameter 148 mm), and a mass spectrometer to detect the escape particle rate (3.8%). All data are fused using a Kalman filter algorithm, and after timestamp alignment, they are transmitted to the control module.

[0113] 3.3 Step S3: Constraint State Evaluation:

[0114] The constraint state evaluation layer of the control module is based on SVM dynamic threshold (electron density threshold 5×10). 17 -7×10 17 m -3 The current constraint status is deemed to meet the standard, based on the escape rate threshold of 4%, combined with the AHP comprehensive score (40 points for constraint time of 10s, 28 points for density uniformity deviation of 2.3%, 29 points for escape rate of 3.8%, and a total score of 97 points). The current magnetic field parameters are maintained and continuous monitoring is performed.

[0115] 3.4 Step S4: Closed-loop adjustment (example of a scenario where constraints are not met):

[0116] After 150 seconds of the experiment, the diagnostic module detected that the electron density had dropped to 4.8 × 10⁻⁶. 17 m -3 (below the lower limit of the dynamic threshold 5×10) 17 m -3 The escape rate rose to 5.2% (above the threshold of 4%), and the overall score dropped to 75 points (unsatisfactory). The control module analyzed the deviation type (linear deviation was dominant) and activated the PID algorithm to generate adjustment commands: the solenoid current was increased by 3A (to 223A) to increase the steady-state magnetic field strength to 3.6T, the high-frequency pulse frequency was increased by 2kHz (to 37kHz) to enhance local confinement, and the alternating magnetic field strength was increased by 0.05T (to 0.25T) to optimize spatial distribution. After receiving the commands, the magnetic field generation module adjusted the parameters and returned to step S2 to re-acquire data. After adjustment, the electron density rose back to 5.1×10⁻⁶. 17 m -3 The escape rate dropped to 3.9%, the overall score rose to 88 points, and the restraint status was restored to the standard.

[0117] 3.5 Data Validation:

[0118] During data acquisition, the electron density of the electrical detection submodule was compared (5.2 × 10⁻⁶). 17 m -3 ) and the laser interferometer data of the optical detection submodule (5.1×10 17 m -3 The deviation is 1.9% (less than 10%), so no re-acquisition is required; the mass spectrometer in the particle detection submodule identifies H. + 92% of the total, He 2+ The percentage was 8%, consistent with the spectral analysis results of the hyperspectral imager (deviation 0), and the validity of the data was verified.

[0119] 3.6 Self-optimization of constraint parameters:

[0120] Every 100 closed-loop control cycles (approximately 33 seconds) are completed, the control module extracts the optimal parameters for that cycle (e.g., a steady-state magnetic field of 3.6T, a pulse frequency of 36kHz, and a comprehensive score of 95) and stores them in the historical database. When the sample size reaches 500 groups, a genetic algorithm (population size 50, crossover probability 0.8, mutation probability 0.05) is used for cluster analysis to generate an initial density of 5×10⁻⁶. 17 m -3 Pressure 5×10 -4 The optimal parameter template under Pa operating conditions. When the system starts up subsequently, it directly matches the template parameters, reducing the constraint compliance time from the initial 8s to 4s (a 50% reduction).

[0121] 3.7 Multi-level fault protection:

[0122] 3.7.1 Level 1 Protection Triggered:

[0123] When the experiment lasted 200 seconds, the control module detected a current deviation of +12% (exceeding ±10%) in the high-frequency pulse coil, generated an early warning signal, and enabled PID algorithm compensation (increasing the output current of the pulse power supply by 5A). After one acquisition cycle, the current deviation dropped to 8%, and the fault was resolved.

[0124] 3.7.2 Level 2 protection triggered:

[0125] When the experiment lasted for 250 seconds, the temperature of the superconducting coil rose to 14.2K (exceeding the critical temperature by 10K). After the first-level protection (increasing the cooling power) was activated, the temperature continued to rise for 3 acquisition cycles (1 second). The control module cut off the high-voltage power supply circuit (response time 45ms) and started the backup superconducting coil (with parameters consistent with the main coil) to maintain the 3.5T basic magnetic field and prevent plasma diffusion.

[0126] 3.7.3 Level 3 Protection Trigger:

[0127] If the backup coil fails to start and the coil temperature rises to 105℃ (exceeding 100℃), the control module triggers an emergency pressure relief procedure, opens the vacuum valve in the confinement area (pressure relief time 2s), releases plasma, and records diagnostic data (such as temperature change curves and current fluctuation values) for 10s before and after the fault, generating a fault report (fault type: superconducting cooling system failure) for subsequent maintenance and system optimization.

[0128] IV. Verification of Implementation Results:

[0129] This embodiment achieves stable plasma confinement using the aforementioned system and method: the confinement time reaches 320s (exceeding the 300s target), and the electron density is stabilized at 5×10⁻⁶. 17 -6×10 17 m -3 (Fluctuation ≤ 5%), electron temperature maintained at 6-7 eV (deviation ≤ 8%), escape particle rate controlled at 3.5%-4.8% (below the 5% target), and spatial distribution deviation of the composite magnetic field ≤ 4% (better than the 5% requirement). Compared with traditional single-mode magnetic field confinement systems, this scheme improves confinement stability by 40% and reduces escape rate by 60%, verifying the advanced nature and practicality of the technical solution.

[0130] This technical solution solves the core problems of traditional plasma confinement systems, such as poor confinement stability, slow response, and weak fault tolerance, through an innovative combination of multi-modal magnetic field collaborative control, multi-dimensional diagnostic fusion, and hierarchical intelligent closed-loop algorithm. It demonstrates significant advantages in scenarios such as nuclear fusion experiments and plasma processing, with the following specific benefits:

[0131] Significantly improved plasma confinement performance to meet demanding application scenarios: Breakthroughs in both confinement strength and stability: A composite magnetic field structure is employed, consisting of a steady-state magnetic field (0.5-8T), a pulsed magnetic field (1-3T, 10-100kHz), and an alternating magnetic field (0.05-0.5T, 0.1-10kHz). Through magnetic field collaborative control logic, the spatial gradient is adjustable. Compared to traditional single-mode magnetic field confinement systems, the plasma confinement time is increased from 100-200s to over 300s (reaching 320s in this example), and the electron density fluctuation is reduced from ±10% to ±5% (stabilizing at 5×10⁻⁶). 17 -6×10 17 m -3 The electron temperature deviation was optimized from ±15% to ±8% (maintained at 6-7 eV), meeting the core requirement of long-term stable confinement of plasma in nuclear fusion experiments.

[0132] Precise control of escape particle rate: Real-time monitoring of escape particle characteristics through multi-dimensional diagnostic modules (electrical + optical + particle detection), combined with intelligent closed-loop adjustment, reduces the escape particle rate from 15%-20% in traditional systems to within 5% (as low as 3.5% in this embodiment), reducing particle loss while avoiding damage to the boundary material of the confinement area caused by high-energy particle bombardment, thus extending the service life of the device.

[0133] Spatial distribution uniformity optimization: With the help of the magnetic field calibration submodule (three-dimensional Hall sensor array + finite element simulation), the deviation between the actual spatial distribution of the composite magnetic field and the preset distribution is ≤5% (only 4% in the example). Combined with the density uniformity monitoring of the laser interferometer (measurement accuracy better than 0.1%), the deviation of plasma spatial distribution uniformity is reduced from ±5% to ±2.3%, providing a more uniform experimental environment for plasma physics research.

[0134] The system's intelligence level has been upgraded to achieve adaptive control throughout the entire process: dynamic thresholds and multi-index evaluation avoid "one-size-fits-all" judgments; dynamic constraint thresholds (such as the electron density threshold of 1.2 × 10⁻⁶) are generated based on the Support Vector Machine (SVM) algorithm. 17 -8×10 17 m -3 A multi-index evaluation matrix was constructed using the Analytic Hierarchy Process (AHP) (constraint time weight 0.4, density uniformity weight 0.3, escape rate weight 0.3). Compared to traditional fixed threshold judgment, this matrix can adapt to initial plasma density (10... 16 -10 18 m -3 ), constrained area pressure (10 -3 -10 -5 With changes in operating conditions (Pa), the accuracy of constraint state determination has been improved to over 99%.

[0135] The hierarchical algorithm scheduling balances response speed and adjustment accuracy: when linear deviation is dominant, the PID algorithm is activated (response time ≤ 10ms); when nonlinear deviation is activated, the BP neural network algorithm is activated (5000 training samples, 1000 iterations); and when complex operating conditions are met, the model predictive control algorithm is activated (prediction step size 8 cycles). Compared with single PID control, the adjustment accuracy is improved from ±8% to ±3%, and in sudden scenarios such as sudden drop in electron density and pressure fluctuation of ±20%, it can still quickly restore constraint stability (deviation correction is completed within one acquisition cycle in the example).

[0136] Parameter self-optimization shortens start-up time: Genetic algorithm is used to perform cluster analysis on 500 sets of optimal constraint parameters to generate parameter templates for different initial operating conditions. The system can directly match the templates when starting up, reducing the constraint compliance time from 8s to 4s (a 50% reduction), reducing experimental preparation time and improving the operating efficiency of the device.

[0137] Multi-level protection and data verification ensure safe and reliable system operation: Three-level fault protection enables risk gradient control: Level 1 protection uses algorithm compensation to resolve minor anomalies (such as current deviation ±10%), Level 2 protection cuts off high-voltage power supply and activates backup units to deal with moderate faults (such as superconducting coil overheating), and Level 3 protection triggers emergency pressure relief to avoid extreme risks (such as coil temperature exceeding 100℃). Compared with traditional single-level protection, the fault handling response time is reduced from 100ms to less than 50ms (power supply cut off in 45ms in the example), and the abnormal data recording function (data before and after the fault in 10s) provides accurate basis for subsequent maintenance, reducing the system failure rate by 60%.

[0138] Cross-validation of multi-source data ensures diagnostic accuracy: cross-validation of electron density data from electrical detection (Langmuir probe) and optical detection (laser interferometer) (re-acquisition if deviation exceeds 10%), and particle species matching verification between particle detection (mass spectrometer) and optical detection (hyperspectral imager) (calibration if deviation exceeds 5%). Compared with a single diagnostic method, the data anomaly rate is reduced from 5% to below 0.1%, avoiding incorrect adjustments due to diagnostic errors and improving the reliability of system operation.

[0139] With outstanding technological innovation and broad application value: Modular design, adaptable to different application scenarios: The multi-modal magnetic field generation module and the multi-dimensional diagnostic module can be flexibly expanded according to needs (such as adding a microwave-assisted heating module to adapt to higher temperature plasma, or replacing the large-diameter coil to adapt to large volume confinement areas). Compared with customized systems, the adaptation cost is reduced by 30%-40%, and it can be promoted to different fields such as plasma torch, magnetic confinement fusion reactor, and plasma etching.

[0140] Core technologies are reusable and drive industry technology upgrades: Core technologies such as hierarchical intelligent closed-loop control logic, magnetic field coordinated adjustment algorithm, and multi-source data fusion method can be transferred to other high-energy physics devices (such as particle accelerators) or industrial equipment (such as high-precision temperature control systems) that require closed-loop control, providing technical references for intelligent upgrades in related fields and having a significant technology radiation effect.

[0141] In summary, this technical solution achieves breakthroughs in confinement performance, intelligence level, safety, and innovation through the combination of "multi-modal magnetic field + intelligent closed loop + multi-dimensional diagnosis". It not only solves the pain points of traditional plasma confinement systems, but also provides key technical support for nuclear fusion energy development and plasma application research, and has important practical value and industrialization potential.

[0142] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A multimodal magnetic field intelligent closed-loop plasma confinement system, characterized in that: include: The multi-mode magnetic field generating module contains at least three independent magnetic field generating units, which can generate at least two magnetic field modes among steady-state magnetic field, pulsed magnetic field, and alternating magnetic field. The spatial gradient of the composite magnetic field can be adjusted through the magnetic field parameter collaborative control logic to dynamically enclose the plasma. The multi-dimensional plasma diagnostic module is precisely matched with the magnetic field coverage area of ​​the multi-mode magnetic field generating module. It integrates an electrical detection sub-module, an optical detection sub-module, and a particle detection sub-module to simultaneously collect the microscopic state parameters and macroscopic constraint parameters of the plasma. The hierarchical intelligent closed-loop control module establishes bidirectional communication with the multi-mode magnetic field generation module and the multi-dimensional plasma diagnostic module through a high-speed data bus. The hierarchical intelligent closed-loop control module receives and integrates multi-source diagnostic data, generates magnetic field adjustment commands based on multi-level control algorithms, and feeds them back to the multi-mode magnetic field generation module to dynamically adjust the intensity gradient, frequency characteristics, and spatial distribution of the composite magnetic field, thereby realizing real-time closed-loop optimization of the plasma confinement state.

2. The multimodal magnetic field intelligent closed-loop plasma confinement system according to claim 1, characterized in that: The magnetic field generating unit of the multimodal magnetic field generating module includes: The superconducting steady-state magnetic field unit adopts a nested double solenoid structure. The inner solenoid and the outer solenoid are respectively wound with niobium-titanium alloy superconducting coils and yttrium barium copper oxide high-temperature superconducting coils, which can generate a constant magnetic field of 0.5-8T. The current of the inner and outer solenoids can be independently adjusted to achieve magnetic field gradient control. The high-frequency pulsed magnetic field unit consists of at least 6 sets of hollow copper coils evenly distributed along the circumference of the confined area. The inner wall of each set of coils is provided with a spiral cooling channel, through which a mixed cooling medium of liquid nitrogen and deionized water is introduced, which can generate a periodic pulsed magnetic field with a peak intensity of 1-3T and a pulse frequency of 10-100kHz. The alternating magnetic field unit adopts a toroidal Helmholtz coil structure, with the coil material being oxygen-free copper. Driven by a frequency-modulated power supply, it can generate a sinusoidal alternating magnetic field with a frequency of 0.1-10kHz and an intensity of 0.05-0.5T. The superconducting steady-state magnetic field unit, high-frequency pulsed magnetic field unit, and alternating magnetic field unit achieve parameter linkage through magnetic field collaborative control logic, forming a spatially adjustable composite confinement magnetic field.

3. The multimodal magnetic field intelligent closed-loop plasma confinement system according to claim 2, characterized in that: The multimodal magnetic field generating module also includes a magnetic field calibration submodule, which comprises a three-dimensional Hall sensor array and a magnetic field simulation unit. The three-dimensional Hall sensor array is deployed at the boundary and center of the plasma confinement region to collect actual spatial distribution data of the composite magnetic field. The magnetic field simulation unit constructs a magnetic field simulation model based on the finite element analysis algorithm, compares the deviation between the actual spatial distribution data and the simulation data, and generates calibration instructions to adjust the current parameters of each magnetic field generating unit so that the deviation between the actual spatial distribution of the composite magnetic field and the preset distribution is less than 5%.

4. The multimodal magnetic field intelligent closed-loop plasma confinement system according to claim 1, characterized in that: The multi-dimensional plasma diagnostic module includes: The electrical detection submodule includes a multi-probe array and an electrostatic probe. The multi-probe array has 5-10 Langmuir probes arranged radially along the confinement region to collect the electron density of the plasma (10). 15 -10 19 The electrostatic probe is used to collect the current-voltage characteristic curve of the plasma and calculate the collision frequency of the plasma. The optical detection submodule includes a hyperspectral imager and a laser interferometer. The hyperspectral imager has a spectral resolution of not less than 0.1 nm and collects emission spectrum data of plasma to analyze particle types (hydrogen, helium, oxygen ions, etc.) and energy level distribution. The laser interferometer uses a 10.6 μm wavelength carbon dioxide laser to measure the spatial distribution uniformity of electron density in plasma, with a measurement accuracy better than 0.1%. The particle detection submodule includes a charge-coupled detector (CCD) and a mass spectrometer. The CCD is used to capture morphological images of the plasma and analyze the boundary contours of the plasma within the confined region. The mass spectrometer is used to detect the types and energy distribution of escaping particles in the plasma, with a detection energy range of 1-100 keV. The electrical detection submodule, optical detection submodule, and particle detection submodule have a synchronous acquisition frequency of 1-5kHz, and the timestamp alignment and redundancy verification of multi-source diagnostic data are achieved through a data fusion algorithm.

5. The multimodal magnetic field intelligent closed-loop plasma confinement system according to claim 1, characterized in that: The hierarchical intelligent closed-loop control module includes: The data preprocessing layer includes a noise suppression unit, an outlier removal unit, and a data normalization unit. The noise suppression unit uses a wavelet transform algorithm to denoise the diagnostic data, reducing noise errors caused by electromagnetic interference. The outlier removal unit identifies and removes abnormal diagnostic data based on the 3σ criterion. The data normalization unit maps diagnostic data of different dimensions to the 0-1 interval, achieving data standardization. The constraint state evaluation layer includes a dynamic threshold generation unit and a multi-index evaluation unit. The dynamic threshold generation unit generates dynamic constraint thresholds based on historical plasma constraint data (at least 100 sets of effective constraint cycle data) and real-time operating parameters (such as constraint region pressure and initial plasma density) using a support vector machine algorithm. The multi-index evaluation unit constructs an evaluation matrix including constraint time, density uniformity, and escape rate, and uses the analytic hierarchy process (AHP) to calculate a comprehensive score of the current plasma constraint state. When the comprehensive score is lower than a preset threshold (80 points out of 100), the constraint state is deemed unqualified. The instruction generation and execution layer includes an algorithm scheduling unit and an instruction issuing unit. The algorithm scheduling unit selects a PID control algorithm, a neural network algorithm, or a model predictive control algorithm to generate magnetic field adjustment instructions based on the constraint state evaluation results. Specifically, the PID algorithm is used when linear deviation is dominant (adjustment response time ≤ 10ms), the neural network algorithm (using a 3-layer BP neural network with a training sample size ≥ 5000 groups) is used when nonlinear deviation is dominant, and the model predictive control algorithm (prediction step size 5-10 acquisition cycles) is used under complex dynamic conditions. The instruction issuing unit transmits the adjustment instructions to the multimodal magnetic field generation module via an Ethernet / fiber optic hybrid communication link, with an instruction transmission delay of less than 1ms.

6. A multi-modal magnetic field intelligent closed-loop plasma confinement method, characterized in that: The method, applied to a multimodal magnetic field intelligent closed-loop plasma confinement system according to any one of claims 1-5, comprises the following steps: S1: System initialization. Each magnetic field generating unit of the multi-mode magnetic field generating module starts according to the preset initial parameters. The superconducting steady-state magnetic field unit generates the basic confinement magnetic field (intensity 2-5T). The high-frequency pulse magnetic field unit and the alternating magnetic field unit operate at the initial frequency (20-50kHz, 1-5kHz) and intensity (1-2T, 0.1-0.3T) to form an initial composite magnetic field to envelop the plasma in the plasma confinement area. S2: The electrical detection submodule, optical detection submodule, and particle detection submodule of the multi-dimensional plasma diagnostic module start up simultaneously, collect state parameters such as electron density, electron temperature, particle type, boundary profile, and energy distribution of escaped particles of the plasma, realize timestamp alignment and redundancy verification of multi-source data through data fusion algorithm, and transmit the processed diagnostic data to the hierarchical intelligent closed-loop control module. S3: The constraint state evaluation layer of the hierarchical intelligent closed-loop control module determines whether the current constraint state of the plasma meets the standard based on the real-time threshold of the dynamic threshold generation unit and the comprehensive score of the multi-index evaluation unit. If the comprehensive score is ≥80 points, it is determined that the constraint meets the standard, and the current composite magnetic field parameters are maintained and continuously monitored. If the comprehensive score is <80 points, it is determined that the constraint does not meet the standard, and proceed to step S4. S4: The instruction generation and execution layer of the hierarchical intelligent closed-loop control module schedules the corresponding control algorithm according to the reasons for non-compliance with constraints (linear deviation / nonlinear deviation / complex dynamic working conditions), generates magnetic field adjustment instructions (including current, frequency, and phase adjustment parameters of each magnetic field generating unit), and transmits them to the multi-mode magnetic field generating module; the multi-mode magnetic field generating module adjusts the composite magnetic field parameters according to the adjustment instructions, returns to step S2, and forms a closed-loop control cycle.

7. The intelligent closed-loop plasma confinement method with multimodal magnetic field according to claim 6, characterized in that: In step S2, the multi-dimensional plasma diagnostic module also includes a data validity verification step: cross-validation is performed between the electron density data of the electrical detection submodule and the laser interferometer data of the optical detection submodule. When the deviation between the two exceeds 10%, the backup probe and backup laser channel are activated to re-acquire data. At the same time, the mass spectrometer data of the particle detection submodule is matched and verified with the particle type data of the hyperspectral imager. When the particle type identification deviation exceeds 5%, the data calibration procedure is triggered to ensure the validity of the diagnostic data.

8. The intelligent closed-loop plasma confinement method with multimodal magnetic field according to claim 6, characterized in that: In step S4, the magnetic field adjustment command also includes magnetic field collaborative compensation logic: when adjusting the current of the superconducting steady-state magnetic field unit, the pulse frequency of the high-frequency pulse magnetic field unit is adjusted synchronously (the frequency adjustment amount is linearly related to the change in steady-state magnetic field strength, and the correlation coefficient is 0.5-2kHz / T) to avoid abrupt changes in the spatial gradient of the composite magnetic field; when adjusting the phase of the alternating magnetic field unit, the pulse phase of the high-frequency pulse magnetic field unit is corrected synchronously (the phase correction amount is 1 / 2-1 / 3 of the alternating magnetic field phase adjustment amount) to ensure the constraint stability of the composite magnetic field.

9. The intelligent closed-loop plasma confinement method with multimodal magnetic field according to claim 6, characterized in that: It also includes a constraint parameter self-optimization step: after every 100 closed-loop control cycles, the constraint state evaluation layer of the hierarchical intelligent closed-loop control module extracts the optimal constraint parameters (composite magnetic field parameters and corresponding diagnostic data) within that cycle and stores them in the historical database; when the number of optimal parameter samples in the historical database reaches 500 sets, a genetic algorithm is used to perform cluster analysis on the optimal parameters to generate different initial operating conditions (such as initial plasma density 10). 16 -10 18 m -3 Constrained area pressure 10 -3 -10 -5 The optimal parameter template under (Pa) can be matched according to the current initial working conditions when the system starts up, thereby shortening the time to achieve the constraint state (the time to achieve the standard is shortened by 30%-50%).

10. The intelligent closed-loop plasma confinement method with multimodal magnetic field according to claim 6, characterized in that: It also includes multi-level fault protection steps: Level 1 protection: The layered intelligent closed-loop control module monitors the current of each coil of the multi-mode magnetic field generating module in real time (triggered when the deviation exceeds the rated value ±10%), the coil temperature (triggered when it exceeds the critical temperature of the superconducting coil by 10K or the copper coil by 80℃), and the data acquisition integrity of the diagnostic module (triggered when the data loss rate exceeds 5%). When an abnormality is detected, an early warning signal is generated and the control algorithm parameters are adjusted for compensation. Secondary protection: If the abnormality persists for more than 3 acquisition cycles after primary protection, the high-voltage power supply circuit of the multi-mode magnetic field generation module is cut off (cut-off response time ≤ 50ms), and the backup magnetic field unit (if any) is started to maintain basic constraints. Level 3 protection: If the backup magnetic field unit fails to start or becomes abnormally aggravated (e.g., the coil temperature exceeds 100°C), an emergency pressure relief procedure is triggered to release the plasma in the confinement area. At the same time, abnormal data (including diagnostic data and magnetic field parameters 10 seconds before and after the fault) is recorded, and a fault report is generated for subsequent analysis and system optimization.

Images