A Full-Parameter Timing Optimization Method for the FRC Center Compression Coil Based on Adaptive Response Surface
By optimizing the geometric and current parameters of the FRC central compression coil using the adaptive response surface methodology and surrogate model, the problem of full parameter optimization in the central fusion region of the FRC device was solved. This enabled efficient magnetic field uniformity and adaptive magnetic field control under time-series conditions, improving computational efficiency and multi-physics coupling processing capabilities.
Patent Information
- Application Number
- CN202611258788.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-19
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies lack a systematic, full-parameter optimization design method for the compression coil in the central fusion region of FRC devices. They cannot achieve efficient magnetic field uniformity under multiple constraints and adaptive magnetic field control under time-sequence conditions. Furthermore, they have low computational efficiency and fail to effectively consider multi-physics coupling effects.
An adaptive response surface model for full-parameter timing optimization of the FRC center compression coil is adopted. By constructing a surrogate model, efficient global search and local optimization are performed. Combined with multi-scale optimization and coil mutual inductance coupling correction, the magnetic field uniformity and current optimization under timing conditions are achieved, satisfying multiple constraints.
Stable solutions for high uniformity coil layout in the central fusion region of the FRC device were achieved, improving computational efficiency and enabling adaptive control of magnetic field morphology under different discharge time states. This solved the computational feasibility and multi-physics coupling problems in the existing technology.
Smart Images

Figure CN123116573A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of coil optimization technology for magnetic confinement fusion devices, and more specifically, to a full-parameter timing optimization method for the FRC central compression coil based on adaptive response surface. Background Technology
[0002] Field-Reversed Configuration (FRC) is a compact magnetic confinement fusion device concept where the plasma is confined solely by a poloidal magnetic field. It offers advantages such as a fully axisymmetric structure, simple design, high specific pressure, and natural divertor geometry. An FRC device typically consists of a central fusion region (central straight tube section) and acceleration / compression sections at both ends. The central fusion region is the core area of the entire device, requiring the generation of a high-intensity, highly uniform axial magnetic field to maintain stable plasma confinement.
[0003] Unlike stellarators and tokamak devices that rely on complex three-dimensional magnetic confinement, the FRC uses axisymmetric coaxial compression coils to generate a nearly uniform axial magnetic field, with the plasma forming a field-reversing configuration through its own diamagnetic properties. This unique physical mechanism dictates the following special characteristics of the magnetic field design in the central fusion region of the FRC: extremely high axial uniformity is required to maintain plasma stability; the end magnetic field needs to form a specific gradient to prevent axial particle loss; and the same set of magnets needs to generate different magnetic field distributions at different stages of the discharge sequence to adapt to the plasma injection, confinement, and diffusion processes.
[0004] In terms of magnetic field optimization design methods, existing technologies mainly include: stellarator coil optimization methods, nuclear magnetic resonance (MRI) shimming coil design methods, tokamak coil optimization methods, and FRC coil empirical design methods. However, existing technologies have the following objective drawbacks: (1) Incompatibility of methods. The magnetic field optimization methods for stellarators, nuclear magnetic resonance and tokamak are developed for their own unique physical requirements, and their optimization objectives, variable types and constraints are not compatible with FRC devices. At present, there is no systematic full-parameter optimization design method specifically for the compression coil in the central fusion region of FRC devices.
[0005] (2) Full parameter optimization is not feasible. Traditional optimization algorithms converge very slowly in high-dimensional spaces. Each objective function evaluation requires time-consuming magnetic field calculations, and the total calculation time can reach several days to several weeks.
[0006] (3) Lack of time-series magnetic field control capability. Existing FRC coil design methods allocate a single fixed current to each coil to generate a static magnetic field distribution. However, the actual operation of FRC devices requires different magnetic field modes at different stages of the discharge sequence (injection state, confined state, diffusion state). Existing methods cannot achieve adaptive switching of multiple field modes of the same magnet structure under different time sequences.
[0007] (4) The constraint treatment is too simplistic. Existing methods usually only consider the geometric constraints related to the magnetic field, without systematically considering the multi-physics coupling effects in the operation of the FRC device, such as the temperature rise constraint caused by the Joule heating effect of the coil, and the constraint of the electromagnetic force between the coils on the structural strength. Summary of the Invention
[0008] This application provides a full-parameter time-series optimization method for FRC center compression coils based on adaptive response surfaces. It aims to meet the requirements of multiple constraints by accelerating high-dimensional search through surrogate models to achieve computational feasibility. This method achieves ultra-high magnetic field uniformity while enabling adaptive field mode control of the same magnet structure under injection, constraint, and diffusion time-series states.
[0009] This invention provides a method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface, including: Step S1: Determine the number of single-sided optimized coils and construct a design space for the geometric parameters of each single-sided optimized coil; Step S2: Sample the sample within the design space to generate a sample set, evaluate the magnetic field uniformity of all sample points in the sample set, and train the surrogate model based on the evaluation results and the sample set. Step S3: Iteratively search for new sample points on the surrogate model and update the sample set. Reconstruct the surrogate model based on the updated sample set until convergence, and obtain the initial solution of the geometric parameters. Step S4: Perform multi-scale optimization on the initial solution and introduce coil mutual inductance coupling correction to eliminate electromagnetic coupling deviation, thereby obtaining the optimal geometric parameters; Step S5: Fix the optimal geometric parameters, use only current as the optimization variable, and perform current optimization for each time state of FRC discharge to obtain the optimal current vector corresponding to each time state. Step S6: Combine the optimal geometric parameters with the optimal current vector to form coil configuration schemes for each timing state. Verify the coil configuration schemes for each timing state. If any timing state does not meet the preset index, then optimize again.
[0010] Preferably, the design space for geometric parameters is constructed for each of the single-sided optimized coils, specifically as follows: The axial position, inner diameter, outer diameter, and winding width of the single-sided optimization coil are used as optimization variables of geometric parameters. The range of values for each optimization variable under the adaptation constraint is determined, and the design space of geometric parameters is obtained. The adaptation constraints include axial position constraints, adjacent spacing constraints, inner diameter constraints, outer diameter constraints, and width constraints.
[0011] Preferably, step S2 specifically includes: A sample set is formed by extracting several samples within the design space, and each sample point corresponds to a set of geometric parameters. Set preset current parameters for each sample point, calculate the axial magnetic field distribution of the central fusion region corresponding to each sample point, and calculate the root mean square relative deviation of the magnetic field based on the axial magnetic field distribution, which is used as the magnetic field uniformity corresponding to the sample point. A Kriging surrogate model is constructed, with the geometric parameters of each sample point as input and the corresponding magnetic field uniformity as output, to complete the training of the Kriging surrogate model.
[0012] Preferably, the process involves iteratively searching for new sample points on the proxy model and updating the sample set, specifically as follows: A global search is performed using the expected improvement criterion, expressed as:
[0013] In the formula, for Expected improvement at the point For any point within the design space, This represents the minimum objective function value in the current sample set, i.e., the optimal magnetic field uniformity among all currently evaluated coil configurations. For the proxy model in The predicted value of the objective function at the point output. The cumulative distribution function of the standard normal distribution. To predict the mean square error, The probability density function is the standard normal distribution. Using the maximization of the expected improvement as the selection criterion, new sample points are obtained, and these new sample points and their corresponding magnetic field uniformity are added to the sample set.
[0014] Preferably, the proxy model is reconstructed based on the updated sample set until convergence, specifically as follows: Based on the updated sample set, the hyperparameters of the Kriging model are re-estimated, and the surrogate model is reconstructed. Repeat the steps of iteratively searching for new sample points, updating the sample set, and reconstructing the surrogate model until the following convergence criteria are met: in five consecutive reconstruction iterations, the minimum change in the magnetic field uniformity of the new sample points is less than the first preset threshold; and the maximum prediction mean square error in the design space of the geometric parameters is less than the second preset threshold. If the convergence condition is not met even after the preset maximum number of iterations, the iteration is stopped and the current optimal solution is output.
[0015] Preferably, the initial solution is optimized at multiple scales, specifically as follows: Using the initial solution of the geometric parameters as the initial point, a coarse-grained global search of the differential evolution algorithm is performed on the Kriging surrogate model to obtain the coarse-grained optimal solution; Using the coarse-grained optimal solution as the initial point, fine-grained local optimization is performed through a sequential quadratic programming algorithm to obtain the fine-grained optimal solution.
[0016] Preferably, electromagnetic coupling deviation is eliminated by introducing coil mutual inductance coupling correction, specifically as follows: Using the preset current parameters corresponding to the fine-grained optimal solution as initial values, the current compensation vector of each coil is iteratively calculated based on the deviation between the target magnetic field and the actual magnetic field under the preset current parameters. The expression is as follows:
[0017] In the formula, and The first Second and third The current compensation vector for the next iteration. As a relaxation factor, The mutual inductance matrix between the coils, This is the deviation vector between the target magnetic field distribution and the actual magnetic field distribution. The current parameters of each coil are corrected using the current compensation vector so that the corrected current parameters meet the preset magnetic field distribution requirements. The magnetic field uniformity of the fine-grained optimal solution is recalculated using the corrected current parameters, and the fine-grained optimal solution with the highest magnetic field uniformity is selected as the optimal geometric parameters.
[0018] Preferably, current optimization is performed for each timing state of FRC discharge, specifically as follows: The current optimization in the injection state aims to reduce the end magnetic field strength, and its objective function is:
[0019] In the formula, Let the objective function be the injected state. It is a current vector. The root mean square relative deviation of the magnetic field. The first preset penalty coefficient, For the injected state magnetic mirror ratio, The upper limit threshold for the injected state magnetic mirror ratio; The current optimization in the constrained state aims to increase the end magnetic field strength, and its objective function is:
[0020] In the formula, Let the objective function be the constrained state. The second preset penalty coefficient, The lower limit threshold for the confined magnetic mirror ratio. For the confined state magnetic mirror ratio; The current optimization in the diffused state aims to increase the central magnetic field strength, and its objective function is:
[0021] In the formula, Let the objective function be the diffuse state. The third preset penalty coefficient, The central magnetic field strength, The average magnetic field strength, The preset target value for the ratio of the central magnetic field strength to the average magnetic field strength; Solve each objective function separately to obtain the current vector for each time series.
[0022] Preferably, the coil configuration scheme for each timing state is verified, specifically as follows: Calculate the magnetic field uniformity, end magnetic mirror ratio, and end magnetic field ratio under each time state, and confirm that all three indicators meet the corresponding preset thresholds. Calculate the steady-state temperature rise under each time state and confirm that the steady-state temperature rise of each coil does not exceed the maximum allowable temperature rise; Calculate the electromagnetic force between adjacent coils under each time state and confirm that the electromagnetic force between adjacent coils does not exceed the maximum bearing capacity of the coil structure. Calculate the current change between each timing state and confirm that the current change during timing state switching is within the allowable range of the power system response time.
[0023] Preferably, if constraint conflicts occur during the optimization process, they are handled in layers according to a preset priority: Constraints are divided into mandatory constraints and relaxable constraints; The mandatory constraints are strictly satisfied during the optimization process, and the relaxable constraints are relaxed step by step at preset steps when there is a conflict, until a feasible solution that satisfies the mandatory constraints is obtained.
[0024] This application has at least the following advantages or beneficial effects: (1) This invention constructs a full geometric parameter design space for FRC axisymmetric compression coils, provides a multi-type engineering adaptation constraint and hierarchical constraint conflict handling mechanism, and combines multi-dimensional verification of magnetic field, thermal field, force field and time sequence switching to achieve stable solution of high uniformity coil layout in the central fusion area.
[0025] (2) This invention uses an adaptive sequence response surface method to construct a Kriging surrogate model through Latin hypercube sampling to replace time-consuming high-precision magnetic field calculation. Combined with the expected improvement criterion, it achieves a balance between efficient global search and local fine optimization, thus solving the computational feasibility bottleneck of full parameter optimization of FRC coils.
[0026] (2) Based on the fixed magnet geometry, this invention optimizes the timing current for the pulse discharge operation characteristics of the FRC device. It solves the optimal current configuration for the three timing states of injection, constraint and diffusion, so that the same group of coils presents three different magnetic field forms at different stages of discharge timing: end flat, end peak and center high field. Attached Figure Description
[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0028] Figure 1 This is a flowchart of the full-parameter timing optimization method for the FRC center compression coil based on adaptive response surface provided by the present invention.
[0029] Figure 2 This is a comparison diagram of the position and structure of the single-sided coil on the ZR plane before and after optimization in an embodiment of the present invention.
[0030] Figure 3 This is a schematic diagram of the axial magnetic field distribution, current configuration, and timing switching process for three different timing states in an embodiment of the present invention. Detailed Implementation
[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0032] Please refer to Figure 1 , Figure 1This is a flowchart of a full-parameter timing optimization method for the FRC center compression coil based on adaptive response surface proposed in an embodiment of this application, including the following steps: Step S1: Determine the number of single-sided optimized coils and construct a design space for the geometric parameters of each single-sided optimized coil.
[0033] Specifically, the number N of single-sided optimization coils is first determined. In this embodiment, the number of single-sided optimization coils is determined to be N=12. For each optimization coil, four independent geometric parameters are selected as optimization variables, namely: axial position, inner diameter, outer diameter and winding width, corresponding to a 4N-dimensional design space.
[0034] The following adaptation constraints are imposed on the design space, including: Axial position constraints are used to limit the axial arrangement range of the coils; adjacent spacing constraints are used to prevent physical interference between adjacent coils; inner diameter constraints are limited by the outer diameter of the vacuum chamber; outer diameter constraints ensure that the winding has a minimum mechanical thickness, while being limited by the radial installation space; width constraints are limited by the winding processing technology and axial space.
[0035] Step S2: Sample the data within the design space to generate a sample set, evaluate the magnetic field uniformity of all sample points in the sample set, and train a proxy model based on the evaluation results and the sample set.
[0036] In this embodiment, an initial sample set is generated in a 4N-dimensional design space using the Latin hypercube sampling method, with each sample point corresponding to a unique combination of geometric parameters.
[0037] Furthermore, a uniform preset reference current is configured for each sample point. The axial magnetic field distribution of the central fusion region corresponding to each sample point was calculated based on the Biot-Savart law. The specific calculation model is as follows: For those located in the axial position With inner diameter , outer diameter Winding width Current The coil, which can be equivalent to a thin current sheet, is positioned on the axis. The axial magnetic field generated at that location is:
[0038] in,
[0039] In the formula, Position on the axis The axial magnetic field generated at that location. The permeability of free space, For equivalent current, For the equivalent radius, It is the axial distance. and These are the first and second type of complete elliptic integrals, respectively.
[0040] It should be understood that the above-described thin current sheet equivalent model simplifies the inter-turn delamination and skin effect of the coil winding. For the FRC center fusion region coil targeted in this embodiment, its operating frequency is below 10kHz, the copper winding diameter is small, and the skin depth is much smaller than the coil radius. The introduced magnetic field calculation error is less than 0.1%, which is lower than the design requirements for magnetic field uniformity. Therefore, the accuracy of this simplified model fully meets the needs of surrogate model training and optimization design. The axial magnetic fields of all optimized coils and fixed coils are superimposed to obtain the total axial magnetic field distribution in the central fusion region. Sampling calculation points are selected at equal intervals within the axial range of the central fusion region to obtain the magnetic field distribution vector, and the relative deviation of the root mean square (RMS) of the magnetic field is used as the evaluation index of magnetic field uniformity.
[0041] To further develop a Kriging surrogate model, we used the geometric parameters of each sample point as input and the corresponding magnetic field uniformity as output to complete the training of the Kriging surrogate model.
[0042] It should be explained that this method uses the Kriging model as a surrogate model because of the low effective dimensionality of the FRC coil optimization problem. Although the apparent optimization dimension is 4N, due to the physical constraints of the electromagnetic coupling of the coil, there is a strong correlation between the parameters, and the actual effective dimension is much lower than the apparent dimension. The Kriging model can maintain high fitting accuracy in this scenario, and at the same time, it can provide a built-in prediction mean square error estimate, which provides a basis for subsequent adaptive sampling.
[0043] Step S3: Iteratively search for new sample points on the surrogate model and update the sample set. Reconstruct the surrogate model based on the updated sample set until convergence, and obtain the initial solution of the geometric parameters.
[0044] In this embodiment, a global search is performed using the expected improvement criterion, with maximizing the expected improvement amount serving as the selection criterion for new sample points. The expression is as follows:
[0045] In the formula, for Expected improvement at the point For any point within the design space, This represents the minimum objective function value in the current sample set, i.e., the optimal magnetic field uniformity among all currently evaluated coil configurations. For the proxy model in The predicted value of the objective function at the point output. The cumulative distribution function of the standard normal distribution. To predict the mean square error, The probability density function is the standard normal distribution. Using the maximization of the expected improvement as the selection criterion, new sample points are obtained, and these new sample points and their corresponding magnetic field uniformity are added to the sample set.
[0046] Based on the updated sample set, the hyperparameters of the Kriging model are re-estimated, and the surrogate model is reconstructed until the dual convergence criteria are met: In five consecutive reconstruction iterations, the minimum change in the magnetic field uniformity of the newly added sample points is less than the first preset threshold; and the maximum prediction mean square error within the design space of the geometric parameters is less than the second preset threshold. If the convergence condition is not met even after the preset maximum number of iterations, the iteration is stopped and the current optimal solution is output.
[0047] The purpose of using a dual convergence criterion is to simultaneously ensure the accuracy of the optimal solution and its global optimality: the first criterion ensures that the currently found optimal solution has become stable and the objective function no longer decreases significantly; the second criterion ensures that the model uncertainty in the entire parameter space is low enough and there are no unexplored significantly better regions. Convergence can be confirmed only if both conditions are met.
[0048] Step S4: Perform multi-scale optimization on the initial solution and introduce coil mutual inductance coupling correction to eliminate electromagnetic coupling deviation, thereby obtaining the optimal geometric parameters.
[0049] As a preferred implementation, the initial solution is optimized using multi-scale methods, specifically: Using the initial solution of the geometric parameters as the initial point, a coarse-grained global search of the differential evolution algorithm is performed on the Kriging surrogate model to obtain the coarse-grained optimal solution; Using the coarse-grained optimal solution as the initial point, fine-grained local optimization is performed through a sequential quadratic programming algorithm to obtain the fine-grained optimal solution.
[0050] In this embodiment, the population size of the differential evolution algorithm is 10×4N, the maximum number of generations is 100, and the algorithm performs global exploration within a large parameter range through mutation, crossover, and selection operations to output a coarse-grained optimal solution.
[0051] As a preferred implementation, coil mutual inductance coupling correction is introduced to eliminate electromagnetic coupling deviation, specifically as follows: Using the preset current parameters corresponding to the fine-grained optimal solution as initial values, the current compensation vector of each coil is iteratively calculated based on the deviation between the target magnetic field and the actual magnetic field under the preset current parameters. The expression is as follows:
[0052] In the formula, and The first Second and third The current compensation vector for the next iteration. As a relaxation factor, The mutual inductance matrix between the coils, This is the deviation vector between the target magnetic field distribution and the actual magnetic field distribution. The current parameters of each coil are corrected using the current compensation vector so that the corrected current parameters meet the preset magnetic field distribution requirements. The magnetic field uniformity of the fine-grained optimal solution is recalculated using the corrected current parameters, and the fine-grained optimal solution with the highest magnetic field uniformity is selected as the optimal geometric parameters.
[0053] It should be explained that the magnetic field calculations in steps S1 to S3 assume that the coils operate independently, without considering the mutual inductance effect between adjacent coils. However, during actual pulse discharge, changes in coil current will generate an induced electromotive force through mutual inductance, causing a deviation between the actual current and the preset value, affecting the accuracy of the magnetic field distribution. Therefore, mutual inductance coupling correction is introduced to eliminate electromagnetic coupling deviation.
[0054] In this embodiment, a comparison diagram of the positional structure of the single-sided coil on the ZR plane before and after optimization is shown below. Figure 2 As shown in the figure. The optimal geometric parameters of the 12 coils on one side after optimization are shown in Table 1: Table 1 Optimal Geometric Parameters
[0055] It can be seen that the optimized coils exhibit a non-uniform layout with varying sizes and spacing. The radial dimensions and width of the central coils are larger, while the dimensions of the coils at both ends are gradually adjusted to match the magnetic field homogenization requirements of the central fusion zone.
[0056] Step S5: Fix the optimal geometric parameters, and use only the current as the optimization variable to optimize the current for each time state of FRC discharge, so as to obtain the optimal current vector corresponding to each time state.
[0057] It should be explained that this method does not perform two optimization solutions for the current. The preset reference current (I_0) is a uniform fixed value, which is only used as the magnetic field evaluation benchmark for different geometric schemes. The current parameter itself does not participate in the optimization in this stage. All optimization variables are only geometric parameters such as the axial position, inner diameter, outer diameter and winding width of the coil. The goal is to search for the feasible solution of the geometric layout with the optimal magnetic field uniformity.
[0058] Specifically, current optimization is performed for the three time-series states of FRC discharge: injection state, confined state, and diffusion state, to obtain the optimal current vector for each time-series state.
[0059] The current optimization in the injection state aims to reduce the end magnetic field strength, and its objective function is:
[0060] In the formula, Let the objective function be the injected state. It is a current vector. The root mean square relative deviation of the magnetic field. The first preset penalty coefficient, For the injected state magnetic mirror ratio, The upper limit threshold for the injected state magnetic mirror ratio; The current optimization in the constrained state aims to increase the end magnetic field strength, and its objective function is:
[0061] In the formula, Let the objective function be the constrained state. The second preset penalty coefficient, The lower limit threshold for the confined magnetic mirror ratio. For the confined state magnetic mirror ratio; The current optimization in the diffused state aims to increase the central magnetic field strength, and its objective function is:
[0062] In the formula, Let the objective function be the diffuse state. The third preset penalty coefficient, The central magnetic field strength, The average magnetic field strength, This is a preset target value for the ratio of the central magnetic field strength to the average magnetic field strength.
[0063] Among them, the axial magnetic field distribution, current configuration, and timing switching process of the three different timing states are as follows: Figure 3As shown above, the three timing states correspond to the complete physical process of pulsed discharge in the FRC device: the injection state has a lower end magnetic field, which is conducive to plasma cluster injection; the confined state has a higher end magnetic field, which forms a strong magnetic mirror effect to prevent axial loss of particles; and the diffusion state has a higher central magnetic field, which promotes plasma energy release.
[0064] As a preferred embodiment, this embodiment uses a normalized calibration method based on the particle loss threshold of magnetohydrodynamics to determine each penalty coefficient, and equates the magnetic mirror ratio deviation to magnetic field inhomogeneity, eliminating the influence of artificial empirical parameters and ensuring the uniqueness and reproducibility of the optimization results.
[0065] Among them, the injection state has a lower end magnetic field, which is conducive to plasma cluster injection; the confined state has a higher end magnetic field, forming a strong magnetic mirror effect to prevent axial particle loss; and the diffuse state has a higher central magnetic field, which promotes plasma energy release. The three field types are achieved by timing current switching, and the switching time is determined by the response speed of the power supply system.
[0066] As a preferred implementation, the coil configuration scheme for each timing state is verified, specifically as follows: Calculate the magnetic field uniformity, end magnetic mirror ratio, and end magnetic field ratio under each time state, and confirm that all three indicators meet the corresponding preset thresholds. Calculate the steady-state temperature rise under each time state and confirm that the steady-state temperature rise of each coil does not exceed the maximum allowable temperature rise; Calculate the electromagnetic force between adjacent coils under each time state and confirm that the electromagnetic force between adjacent coils does not exceed the maximum bearing capacity of the coil structure. Calculate the current change between each timing state and confirm that the current change during timing state switching is within the allowable range of the power system response time.
[0067] As a preferred implementation, if constraint conflicts occur during the optimization process, they are handled in a hierarchical manner according to a preset priority: Constraints are divided into mandatory constraints and relaxable constraints; The mandatory constraints are strictly satisfied during the optimization process, and the relaxable constraints are relaxed step by step at preset steps when there is a conflict, until a feasible solution that satisfies the mandatory constraints is obtained.
[0068] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.
[0069] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.
[0070] The above provides a detailed description of the FRC center compression coil full parameter timing optimization method based on adaptive response surface provided in this application. Specific examples have been used to illustrate the principle and implementation of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A full parameter timing optimization method for FRC central compression coils based on adaptive response surface, characterized in that, Includes the following steps: Step S1: Determine the number of single-sided optimized coils and construct a design space for the geometric parameters of each single-sided optimized coil; Step S2: Sample the sample within the design space to generate a sample set, evaluate the magnetic field uniformity of all sample points in the sample set, and train the surrogate model based on the evaluation results and the sample set. Step S3: Iteratively search for new sample points on the surrogate model and update the sample set. Reconstruct the surrogate model based on the updated sample set until convergence, and obtain the initial solution of the geometric parameters. Step S4: Perform multi-scale optimization on the initial solution and introduce coil mutual inductance coupling correction to eliminate electromagnetic coupling deviation, thereby obtaining the optimal geometric parameters; Step S5: Fix the optimal geometric parameters, use only current as the optimization variable, and perform current optimization for each time state of FRC discharge to obtain the optimal current vector corresponding to each time state. Step S6: Combine the optimal geometric parameters with the optimal current vector to form coil configuration schemes for each timing state. Verify the coil configuration schemes for each timing state. If any timing state does not meet the preset index, then optimize again.
2. The FRC central compression coil full-parameter timing optimization method based on adaptive response surface according to claim 1, characterized in that, For each of the single-sided optimized coils, a design space for geometric parameters is constructed as follows: The axial position, inner diameter, outer diameter, and winding width of the single-sided optimization coil are used as optimization variables of geometric parameters. The range of values for each optimization variable under the adaptation constraint is determined, and the design space of geometric parameters is obtained. The adaptation constraints include axial position constraints, adjacent spacing constraints, inner diameter constraints, outer diameter constraints, and width constraints.
3. The full parameter timing optimization method of FRC central compression coils based on adaptive response surface according to claim 1, characterized in that, Step S2, specifically: A sample set is formed by extracting several samples within the design space, and each sample point corresponds to a set of geometric parameters. Set preset current parameters for each sample point, calculate the axial magnetic field distribution of the central fusion region corresponding to each sample point, and calculate the root mean square relative deviation of the magnetic field based on the axial magnetic field distribution, which is used as the magnetic field uniformity corresponding to the sample point. A Kriging surrogate model is constructed, with the geometric parameters of each sample point as input and the corresponding magnetic field uniformity as output, to complete the training of the Kriging surrogate model.
4. The full parameter timing optimization method of FRC central compression coils based on adaptive response surface according to claim 1, wherein, Iteratively search for new sample points on the surrogate model and update the sample set, specifically as follows: A global search is performed using the expected improvement criterion, expressed as: In the formula, for Expected improvement at the point For any point within the design space, This represents the minimum objective function value in the current sample set, i.e., the optimal magnetic field uniformity among all currently evaluated coil configurations. For the proxy model in The predicted value of the objective function at the point output. The cumulative distribution function of the standard normal distribution. To predict the mean square error, The probability density function is the standard normal distribution. Using the maximization of the expected improvement as the selection criterion, new sample points are obtained, and these new sample points and their corresponding magnetic field uniformity are added to the sample set.
5. The method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface according to claim 1, characterized in that, The proxy model is reconstructed based on the updated sample set until convergence, specifically as follows: Based on the updated sample set, the hyperparameters of the Kriging model are re-estimated, and the surrogate model is reconstructed. Repeat the steps of iteratively searching for new sample points, updating the sample set, and reconstructing the surrogate model until the following convergence criteria are met: in five consecutive reconstruction iterations, the minimum change in the magnetic field uniformity of the new sample points is less than the first preset threshold; and the maximum prediction mean square error in the design space of the geometric parameters is less than the second preset threshold. If the convergence condition is not met even after the preset maximum number of iterations, the iteration is stopped and the current optimal solution is output.
6. The method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface according to claim 1, characterized in that, The initial solution is then optimized using a multi-scale method, specifically as follows: Using the initial solution of the geometric parameters as the initial point, a coarse-grained global search of the differential evolution algorithm is performed on the Kriging surrogate model to obtain the coarse-grained optimal solution; Using the coarse-grained optimal solution as the initial point, fine-grained local optimization is performed through a sequential quadratic programming algorithm to obtain the fine-grained optimal solution.
7. The method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface according to claim 1, characterized in that, Furthermore, coil mutual inductance coupling is introduced to correct and eliminate electromagnetic coupling deviation, specifically as follows: Using the preset current parameters corresponding to the fine-grained optimal solution as initial values, the current compensation vector of each coil is iteratively calculated based on the deviation between the target magnetic field and the actual magnetic field under the preset current parameters. The expression is as follows: In the formula, and The first Second and third The current compensation vector for the next iteration. As a relaxation factor, The mutual inductance matrix between the coils, This is the deviation vector between the target magnetic field distribution and the actual magnetic field distribution. The current parameters of each coil are corrected using the current compensation vector so that the corrected current parameters meet the preset magnetic field distribution requirements. The magnetic field uniformity of the fine-grained optimal solution is recalculated using the corrected current parameters, and the fine-grained optimal solution with the highest magnetic field uniformity is selected as the optimal geometric parameters.
8. The method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface according to claim 1, characterized in that, Current optimization is performed for each timing state of FRC discharge, specifically as follows: The current optimization in the injection state aims to reduce the end magnetic field strength, and its objective function is: In the formula, Let the objective function be the injected state. For current vectors, The root mean square relative deviation of the magnetic field. The first preset penalty coefficient, For the injected state magnetic mirror ratio, The upper limit threshold for the injected state magnetic mirror ratio; The current optimization in the constrained state aims to increase the end magnetic field strength, and its objective function is: In the formula, Let the objective function be the constrained state. The second preset penalty coefficient, The lower limit threshold for the confined magnetic mirror ratio. For the confined state magnetic mirror ratio; The current optimization in the diffused state aims to increase the central magnetic field strength, and its objective function is: In the formula, Let the objective function be the diffuse state. The third preset penalty coefficient, The central magnetic field strength, The average magnetic field strength, The preset target value for the ratio of the central magnetic field strength to the average magnetic field strength; Solve each objective function separately to obtain the current vector for each time series.
9. The method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface according to claim 1, characterized in that, The coil configuration schemes for each timing state are verified, specifically as follows: Calculate the magnetic field uniformity, end magnetic mirror ratio, and end magnetic field ratio under each time state, and confirm that all three indicators meet the corresponding preset thresholds. Calculate the steady-state temperature rise under each time state and confirm that the steady-state temperature rise of each coil does not exceed the maximum allowable temperature rise; Calculate the electromagnetic force between adjacent coils under each time state and confirm that the electromagnetic force between adjacent coils does not exceed the maximum bearing capacity of the coil structure. Calculate the current change between each timing state and confirm that the current change during timing state switching is within the allowable range of the power system response time.
10. The method for full-parameter timing optimization of the FRC center compression coil based on adaptive response surface according to claim 1, characterized in that, If constraint conflicts occur during the optimization process, they will be handled in a tiered manner according to preset priorities: Constraints are divided into mandatory constraints and relaxable constraints; The mandatory constraints are strictly satisfied during the optimization process, and the relaxable constraints are relaxed step by step at preset steps when there is a conflict, until a feasible solution that satisfies the mandatory constraints is obtained.
