Flywheel energy storage system, method and equipment based on adversarial network algorithm and medium
By combining finite element analysis and generative adversarial network algorithms to optimize the hybrid bearing system, the problems of low computational efficiency and easy getting trapped in local optima were solved, realizing the design of a high-precision, high-performance bearing system and improving static levitation force and dynamic stability.
Patent Information
- Application Number
- CN202511922356.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-18
- Publication Date
- 2026-05-19
AI Technical Summary
Existing hybrid bearing systems suffer from low computational efficiency and are prone to getting trapped in local optima when dealing with complex systems involving multiple parameters and multiple physical fields, making it difficult to achieve global optimal performance and limiting the application of high-precision, high-performance rotating equipment.
By combining finite element analysis with generative adversarial network algorithm, the structure and key parameters of the hybrid bearing system are optimized. Through the collaborative design of HTS module and AMB module, static levitation force and dynamic stability are achieved. The superconducting magnetic flux pinning effect and electromagnetic force are combined with PID controller for real-time compensation.
It improves the overall performance of the hybrid bearing system, ensures high precision and high reliability under different operating conditions, enhances dynamic response and system adaptability, and achieves the global optimal performance of the bearing.
Smart Images

Figure CN122068685A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of magnetic levitation bearing technology, specifically to a flywheel energy storage system, method, device, and medium based on adversarial network algorithms. Background Technology
[0002] In the field of existing bearing technology, high-temperature superconducting magnetic levitation bearings, due to their unique superconducting properties, exhibit high stability in a static state. However, their dynamic response capability is relatively weak, making it difficult to quickly and accurately cope with dynamic disturbances under complex working conditions. In contrast, while electromagnetic bearings possess good dynamic controllability and can adjust electromagnetic force in real time according to actual working conditions, they suffer from significant energy consumption due to their reliance on continuous electrical energy input to maintain operation.
[0003] While research on combining HTS and AMB modules into hybrid bearing systems is gaining traction, the design of the coupled structure of these modules presents challenges. Due to the complex parameter coupling between them, traditional optimization methods often face numerous difficulties. On one hand, these methods suffer from low computational efficiency when dealing with complex systems involving multiple parameters and multi-physics interactions, leading to excessively long optimization processes. On the other hand, they are prone to getting trapped in local optima, making it difficult to find the parameter combination that truly enables the entire hybrid bearing system to achieve globally optimal performance. This limits further improvements in the performance of hybrid bearing systems, necessitating innovative technologies to overcome these limitations. Summary of the Invention
[0004] In view of the above-mentioned problems, the present invention is proposed.
[0005] Therefore, the technical problem solved by this invention is to achieve synergistic optimization of the structure and key parameters of the hybrid bearing system by integrating the finite element analysis (FEA) method with the generative adversarial network (GAN) algorithm, in order to improve the overall performance of the bearing system under different working conditions and meet the application requirements of various high-precision and high-performance rotating equipment.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a flywheel energy storage system based on adversarial network algorithms, which includes the following steps: The HTS module, including a superconducting unit, a cooling unit, and a mechanical calculation unit, is used to generate static levitation force through the superconducting flux pinning effect, to provide contactless axial support and zero-friction levitation for the flywheel rotor, to maintain axial position stability by utilizing the self-stabilizing properties of superconductors, and to provide a low-temperature working environment through a liquid nitrogen cooling system. The AMB module, including an electromagnetic structure unit, a magnetic field calculation unit, and a control algorithm unit, is used to generate controllable electromagnetic force through an 8-pole electromagnetic coil array to perform active radial control of the rotor, suppress radial disturbances during high-speed rotation, and use a PID controller to compensate for displacement deviations in real time to cope with sudden load changes or external impacts. The silicon steel coated rotor serves as the magnetic circuit medium and is coupled with the permanent magnet of the HTS to transmit axial-radial forces. The axial-radial decoupled integrated module includes a mechanical coupling unit, a sensor unit, and a communication and synchronization unit. It is used to divide the force domains of HTS and AMB through a set of dynamic equilibrium equations, and to form a full-degree-of-freedom displacement sensing network using a laser rangefinder and a Hall array.
[0007] As a preferred embodiment of the flywheel energy storage system based on adversarial network algorithm described in this invention, the HTS module includes a toroidal YBaCuO superconducting bulk material, which is cooled to 77K by liquid nitrogen to maintain the superconducting state and generates axial levitation force based on the magnetic flux pinning effect. The formula for levitation force is: in, For levitation force, The critical current density of the superconducting bulk material. The strength of the pinning magnetic field, The volume obtained to calculate the geometry of the superconducting bulk material; The diameter ratio of the toroidal YBaCuO superconducting bulk material is 3:2, and the thickness is 10-20 mm. The axial suspension stiffness is optimized by arranging it in a toroidal array. The formula for axial suspension stiffness is as follows: in, For levitation force, Let l be the axial suspension stiffness, l be the number of layers in the shaped array, and d be the spacing between the shaped arrays. For axial air gap, The critical current density of the superconducting bulk material. Pi is the mathematical constant of a circle.
[0008] As a preferred embodiment of the flywheel energy storage system based on adversarial network algorithm described in this invention, the AMB module includes an 8-pole radially symmetrical electromagnetic coil array with 100 turns per pole, a wire diameter of 1mm, and a rotor surface plated with 0.5mm silicon steel sheet to reduce magnetic resistance. The formula for electromagnetic force is: in, Electromagnetic force, Let N be the permeability of free space, N be the number of turns in the coil, and I be the current flowing through it. It is a radial air gap. Let π be the mathematical constant pi, r be the radius of the rotor, and L be the axial length of the rotor. The PID controller includes dynamically adjusting the coil current based on the radial offset feedback from the displacement sensor, as expressed by the formula: in, Where 'e' represents the radial air gap and 'e' represents the radial displacement deviation. The radial displacement deviation varies with time t, where t is time. The current output by the controller. This is the proportionality coefficient. For accurate tracking of the reference air gap, The integral coefficient is... is the differential coefficient.
[0009] As a preferred embodiment of the flywheel energy storage system based on adversarial network algorithm described in this invention, the axial-radial decoupling integration includes: clarifying the functional division of the HTS module and the AMB module; establishing a dynamic equilibrium equation set based on the functional division; axial mechanical decoupling design; radial mechanical decoupling design; and combining the axial and radial mechanical decoupling designs to obtain a hybrid bearing system. The formula for establishing the dynamic equilibrium equation system is as follows: in, For levitation force, Electromagnetic force, To provide a stable levitation force in the axial direction based on the magnetic flux pinning effect, This refers to the electromagnetic force generated when the rotor rotates at high speed in the radial direction. For the weight of the rotor, Let be the centrifugal force generated when the rotor rotates at high speed, m be the rotor mass, g be the acceleration due to gravity, and r be the rotor radius. ω is the angular velocity of the rotor.
[0010] Another objective of this invention is to provide a flywheel energy storage method based on adversarial network algorithms.
[0011] To address the aforementioned technical problems, this invention provides the following technical solution: a flywheel energy storage method based on adversarial network algorithms, comprising: A hybrid bearing system was constructed, and an electromagnetic-thermal coupling model was established based on finite element analysis to generate an initial parameter-performance dataset. Generative adversarial networks were used to optimize the parameters of the hybrid bearing system. Random noise was input into the generator to generate candidate parameter combinations, and adversarial training was conducted with a discriminator against a real dataset. The parameter combinations were then optimized by combining a physical constraint loss function. The optimized parameters were input into the finite element analysis model for simulation verification, and the levitation force and stability indices were extracted. The levitation stability and dynamic response performance were then verified.
[0012] As a preferred embodiment of the flywheel energy storage method based on adversarial network algorithm described in this invention, the establishment of the electromagnetic-thermal coupling model includes: establishing a three-dimensional electromagnetic-thermal coupling model in COMSOL Multiphysics, defining the Bean critical state model of YBaCuO superconductor, the nonlinear BH curve of copper coil and the magnetic properties of NdFeB permanent magnet, setting liquid nitrogen cooling boundary conditions and rotor motion constraints, and calculating levitation force and electromagnetic force through a transient solver; The optimization of parameters of the hybrid bearing system includes: the generator receiving a combination of parameters generated by 100-dimensional Gaussian noise, and the discriminator inputting the actual parameters and performance scores. The adversarial loss function is expressed as: in, To counteract the loss function, For mathematical expectation operators, For the discriminator function, For the true parameter combination generated through finite element analysis, Let z be the candidate parameter combination generated by the generator based on random noise z, where z is a random noise vector following a Gaussian distribution. For physical constraint functions, The finite element simulation performance results are for the parameter combination θ. Here, A represents the parameter combination, and A represents the target performance. For the total loss function, The weighting coefficients are used to balance the adversarial loss and the physical constraint loss; the hyperparameter λ of the total loss function is optimized through grid search.
[0013] As a preferred embodiment of the flywheel energy storage method based on adversarial network algorithm described in this invention, the generator includes a multi-layer neural network, and the output parameters include the number of turns of the current coil, the air gap, and the PID gain parameters. The discriminator includes a convolutional neural network, which takes a combination of input parameters and a performance score as input and outputs a probability of authenticity. The optimization of parameter combinations using the physical constraint loss function includes calculating the Euclidean distance between the finite element simulation performance and the target performance, expressed by the formula: in, Let be the finite element analysis simulation performance value of the i-th item, and be the preset performance value of the i-th item, where i is the variable index and n is the total number of performance indicators. For Euclidean distance.
[0014] As a preferred embodiment of the flywheel energy storage method based on adversarial network algorithm described in this invention, the extraction of levitation force and stability index includes: comparing the magnitude of levitation force obtained by the optimized simulation with the data before optimization, calculating the percentage change of levitation force, comparing the distribution cloud map of levitation force before and after optimization, and observing whether the uniformity and rationality of the levitation force distribution have been improved. The formula for calculating the percentage change in levitation force is expressed as: Where P represents the percentage change in levitation force. The levitation force before optimization For optimized levitation force; Stability metrics include standard deviation, extreme value difference, and coefficient of variation. The formula for calculating standard deviation is: Where v is the data sample size. Standard deviation Let j be the levitation force value of sample j, where j is the variable index. The mean levitation force is given by ; the extreme value difference is the difference between the maximum and minimum levitation force. The formula for the coefficient of variation is expressed as: in, The coefficient of variation is 1. Standard deviation This represents the average levitation force.
[0015] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the flywheel energy storage method based on the adversarial network algorithm.
[0016] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the flywheel energy storage method based on an adversarial network algorithm.
[0017] The beneficial effects of this invention are as follows: Through a meticulously designed hybrid bearing system structure, including the design of the annular YBaCuO superconducting bulk material of the HTS module, the design of the cooling system, and the optimization of static levitation force, a stable static levitation force based on magnetic flux pinning characteristics is achieved. This ensures uniform magnetic force distribution and load-bearing stability of the bearing in the axial direction, while maintaining the performance of the superconducting material, extending its service life, and enhancing the stability and reliability of the levitation system. The radially symmetrical electromagnetic coil array design, electromagnet and magnetic circuit topology design, and PID controller design of the AMB module ensure the dynamic stability of the bearing under high-speed rotation and complex operating conditions, improving the dynamic performance of the bearing. System adaptability; the coupled structure design realizes the functional division and decoupling of the HTS module and AMB module, avoiding mutual interference between axial and radial forces, and improving the overall system performance and operational reliability; through the GAN-FEA joint optimization method, a high-fidelity parameter-performance dataset was constructed, realizing automatic optimization of design parameters and improving system performance; by comparing and evaluating the optimization effect, the improvement of levitation force and the improvement of system stability were clarified, providing an efficient and reliable solution for bearing applications under high-speed rotation and complex working conditions, and significantly improving the performance of hybrid bearing systems in terms of static levitation force, dynamic stability and control accuracy. Attached Figure Description
[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a system block diagram of a flywheel energy storage system based on an adversarial network algorithm, provided as an embodiment of the present invention.
[0020] Figure 2 The above is a flowchart of a flywheel energy storage method based on an adversarial network algorithm provided in one embodiment of the present invention. Detailed Implementation
[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0022] Example 1, referring to Figure 1This is the first embodiment of the present invention, which provides a flywheel energy storage system based on an adversarial network algorithm, comprising: The HTS module, including a superconducting unit, a cooling unit, and a mechanical calculation unit, is used to generate static levitation force through the superconducting flux pinning effect, to provide contactless axial support and zero-friction levitation for the flywheel rotor, to maintain axial position stability by utilizing the self-stabilizing properties of superconductors, and to provide a low-temperature working environment through a liquid nitrogen cooling system. The HTS module also includes a ring-shaped YBaCuO superconducting bulk material, which is cooled to 77K by liquid nitrogen to maintain the superconducting state and generates axial levitation force based on the magnetic flux pinning effect. The superconducting unit comprises a ring-shaped YBaCuO superconducting bulk material with a diameter ratio of 3:2 and a thickness of 10-20 mm. A ring array arrangement optimizes the axial suspension stiffness. The suspension force formula is: in, For levitation force, The critical current density of the superconducting bulk material. The strength of the pinning magnetic field, The volume obtained to calculate the geometry of the superconducting bulk material.
[0023] The cooling unit includes liquid nitrogen cooling channels designed around the superconducting bulk material. The channel layout is designed to ensure that liquid nitrogen can flow evenly across the surface of the superconducting bulk material, carrying away heat and keeping the superconducting bulk material at a temperature of 77K.
[0024] When the mechanical calculation unit includes a ring array of superconducting bulk materials, it is necessary to consider the influence of two key parameters on the axial suspension stiffness: the spacing and number of layers of the ring array. Maximizing these parameters by changing the spacing and number of layers improves the load-bearing stability of the HTS module in the axial direction. The axial suspension stiffness is obtained by taking the partial derivative of the axial air gap, expressed by the formula: in, For levitation force, Let l be the axial suspension stiffness, l be the number of layers in the shaped array, and d be the spacing between the shaped arrays. For axial air gap, The critical current density of the superconducting bulk material. Pi; The AMB module, including an electromagnetic structure unit, a magnetic field calculation unit, and a control algorithm unit, is used to generate controllable electromagnetic force through an 8-pole electromagnetic coil array to perform active radial control of the rotor, suppress radial disturbances during high-speed rotation, and use a PID controller to compensate for displacement deviations in real time to cope with sudden load changes or external impacts. The silicon steel coated rotor serves as the magnetic circuit medium and is coupled with the permanent magnet of the HTS to transmit axial-radial forces. The AMB module also includes an 8-pole radially symmetrical electromagnetic coil array with 100 turns per pole, a wire diameter of 1mm, and a rotor surface plated with 0.5mm silicon steel sheets to reduce magnetic resistance. The formula for electromagnetic force is: in, Electromagnetic force, Let N be the permeability of free space, N be the number of turns in the coil, and I be the current flowing through it. It is a radial air gap. Let π be the circumference of a circle, r be the radius of the rotor, and L be the axial length of the rotor.
[0025] The PID controller dynamically adjusts the coil current based on the radial offset feedback from the displacement sensor, as expressed by the formula: in, Where 'e' represents the radial air gap and 'e' represents the radial displacement deviation. The radial displacement deviation varies with time t, where t is time. The current output by the controller. This is the proportionality coefficient. For accurate tracking of the reference air gap, The integral coefficient is... These are the differential coefficients. Amperes per millimeter, Amperes per millimeter per second The ampere-second per millimeter allows the system to quickly and stably adjust the electromagnetic force to maintain the stability of the radial air gap when faced with radial disturbances under different operating conditions.
[0026] The axial-radial decoupled integrated module includes a mechanical coupling unit, a sensor unit, and a communication and synchronization unit. It is used to divide the force domains of HTS and AMB through a set of dynamic equilibrium equations and to form a full-degree-of-freedom displacement sensing network using a laser rangefinder and a Hall array. The axial-radial decoupling integration includes clarifying the functional division of the HTS module and the AMB module, establishing a dynamic equilibrium equation set based on the functional division, axial mechanical decoupling design, radial mechanical decoupling design, and combining the axial and radial mechanical decoupling designs to obtain a hybrid bearing system. The formula for establishing the dynamic equilibrium equation system is as follows: in, For levitation force, Electromagnetic force, To provide a stable levitation force in the axial direction based on the magnetic flux pinning effect, This refers to the electromagnetic force generated when the rotor rotates at high speed in the radial direction. For the weight of the rotor, Let be the centrifugal force generated when the rotor rotates at high speed, m be the rotor mass, g be the acceleration due to gravity, and r be the rotor radius. ω is the angular velocity of the rotor.
[0027] By continuously monitoring and adjusting relevant parameters, these two balance equations are always maintained, ensuring the normal operation of the system. In the axial direction, the HTS module bears the axial load, making full use of the static suspension force advantage. In the radial direction, the AMB module actively controls the system. The non-contact displacement sensor provides real-time and accurate feedback of air gap data. The PID controller calculates and adjusts the electromagnetic force of the AMB module to achieve precise control of the radial displacement, ensuring that the rotor is always in a stable suspension state. Furthermore, the axial and radial supports work together to form a complete high-performance hybrid bearing system.
[0028] The HTS module only bears the axial load, i.e. the weight of the rotor. The stable levitation force generated by the HTS module in the axial direction based on the magnetic flux pinning effect is equal to the weight of the rotor, ensuring static stability in the axial direction. Axial displacement detection uses a laser rangefinder to measure the axial air gap, which can accurately obtain the distance change between the rotor and the corresponding components in the axial direction.
[0029] The AMB module compensates for radial disturbances. When the rotor rotates at high speed, the centrifugal force generated is equal to the electromagnetic force. By monitoring the radial air gap change in real time, the electromagnetic force is adjusted by the PID controller to counteract radial disturbances such as centrifugal force, ensuring dynamic stability in the radial direction. Radial displacement detection uses a Hall sensor array to measure the radial air gap. By arranging the Hall sensor array around the rotor, the changes in the radial air gap can be monitored at multiple points in real time, quickly capturing the dynamic changes in the radial air gap during high-speed rotation, and providing timely and accurate data for subsequent control.
[0030] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
[0031] Example 2, refer to Figure 2 This is a second embodiment of the present invention, which provides a flywheel energy storage method based on an adversarial network algorithm, including: S1: Construct a hybrid bearing system, establish an electromagnetic-thermal coupling model based on finite element analysis, and generate an initial parameter-performance dataset.
[0032] The establishment of the electromagnetic-thermal coupling model includes: establishing a three-dimensional electromagnetic-thermal coupling model in COMSOL Multiphysics; defining the Bean critical state model of YBaCuO superconductor, the nonlinear BH curve of copper coil and the magnetic properties of NdFeB permanent magnet; setting liquid nitrogen cooling boundary conditions and rotor motion constraints; and calculating levitation force and electromagnetic force through a transient solver. The optimization of parameters of the hybrid bearing system includes: the generator receiving a combination of parameters generated by 100-dimensional Gaussian noise, and the discriminator inputting the actual parameters and performance scores. The adversarial loss function is expressed as: in, To counteract the loss function, For mathematical expectation operators, For the discriminator function, For the true parameter combination generated through finite element analysis, Let z be the candidate parameter combination generated by the generator based on random noise z, where z is a random noise vector following a Gaussian distribution. For physical constraint functions, The finite element simulation performance results are for the parameter combination θ. Here, A represents the parameter combination, and A represents the target performance. For the total loss function, The weighting coefficients are used to balance the adversarial loss and the physical constraint loss; the hyperparameter λ of the total loss function is optimized through grid search.
[0033] It should be noted that, The smaller the value, the closer the system performance of the generated parameter combination is to the target performance, and the higher the degree of compliance between physical feasibility and performance requirements; conversely, the larger the deviation, the more the parameter combination needs to be optimized. The parameters of the generator and discriminator are updated by backpropagation based on the total loss function. Through multiple iterations of training, the generator can generate a globally optimal parameter combination that meets physical constraints and has excellent performance.
[0034] S2: Generative adversarial networks are used to optimize the parameters of the hybrid bearing system. Random noise is input into the generator to generate candidate parameter combinations, and the discriminator is used to conduct adversarial training with the real dataset. The parameter combinations are then optimized by combining the physical constraint loss function.
[0035] The generator includes a multi-layer neural network, and its output parameters include the number of turns of the current coil, the air gap, and the PID gain parameters. The discriminator includes a convolutional neural network, which takes a combination of input parameters and a performance score as input and outputs a probability of authenticity. The optimization of parameter combinations using the physical constraint loss function includes calculating the Euclidean distance between the finite element simulation performance and the target performance, expressed by the formula: in, Let be the finite element analysis simulation performance value of the i-th item, and be the preset performance value of the i-th item, where i is the variable index and n is the total number of performance indicators. For Euclidean distance.
[0036] S3: Input the optimized parameters into the finite element analysis model for simulation verification, extract the suspension force and stability index, and verify the suspension stability and dynamic response performance.
[0037] The extraction of suspension force and stability indicators includes comparing the magnitude of the suspension force obtained from the optimized simulation with the data before optimization, calculating the percentage change in suspension force, comparing the distribution cloud map of suspension force before and after optimization, and observing whether the uniformity and rationality of the suspension force distribution have been improved. The formula for calculating the percentage change in levitation force is expressed as: Where P represents the percentage change in levitation force. The levitation force before optimization For optimized levitation force; Stability metrics include standard deviation, extreme value difference, and coefficient of variation. The formula for calculating standard deviation is: Where v is the data sample size. Standard deviation Let j be the levitation force value of sample j, where j is the variable index. The mean levitation force is given by ; the extreme value difference is the difference between the maximum and minimum levitation force. The formula for the coefficient of variation is expressed as: in, The coefficient of variation is 1. Standard deviation This represents the average levitation force.
[0038] It should be noted that the standard deviation is the most crucial quantitative indicator. By calculating the degree of deviation between the suspension force data and the mean, it reflects the data dispersion. The smaller the standard deviation, the more concentrated the suspension force values are around the mean, and the more uniform the distribution. The extreme value difference is the difference between the maximum and minimum suspension force values. The smaller the difference, the smaller the fluctuation range of the suspension force at different locations, and the better the uniformity of the distribution. The coefficient of variation is the ratio of the standard deviation to the mean. It is used to compare the uniformity of data at different magnitudes. When the difference between the mean suspension force before and after optimization is greater than a set threshold, the coefficient of variation can eliminate the influence of magnitude and more objectively reflect the change in uniformity.
[0039] Compare the natural frequencies of the system before and after optimization, analyze the impact of the change in natural frequencies on the dynamic performance of the system, compare the damping ratio before and after optimization, and evaluate whether the vibration reduction capability of the system has been improved. If the damping ratio increases, it indicates that the system can recover stability more quickly after being disturbed.
[0040] Furthermore, the levitation force results are extracted, and the magnitude of the levitation force experienced by the rotor under different positions and operating conditions is extracted. The force components (axial force and radial force) are calculated separately. Using Comsol's post-processing function, the distribution cloud map of the levitation force on the rotor surface or a specific cross section is plotted to visually observe the distribution of the levitation force. The stability index results are extracted, and the natural frequency of the hybrid bearing system is obtained through modal analysis.
[0041] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
[0042] Example 3 is the third embodiment of the present invention, which differs from the previous two embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0043] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0044] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0045] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
Claims
1. A flywheel energy storage system based on adversarial network algorithm, characterized in that: include, The HTS module, including a superconducting unit, a cooling unit, and a mechanical calculation unit, is used to generate static levitation force through the superconducting flux pinning effect, to provide contactless axial support and zero-friction levitation for the flywheel rotor, to maintain axial position stability by utilizing the self-stabilizing properties of superconductors, and to provide a low-temperature working environment through a liquid nitrogen cooling system. The AMB module, including an electromagnetic structure unit, a magnetic field calculation unit, and a control algorithm unit, is used to generate controllable electromagnetic force through an 8-pole electromagnetic coil array to perform active radial control of the rotor, suppress radial disturbances during high-speed rotation, and use a PID controller to compensate for displacement deviations in real time to cope with sudden load changes or external impacts. The silicon steel coated rotor serves as the magnetic circuit medium and is coupled with the permanent magnet of the HTS to transmit axial-radial forces. The axial-radial decoupled integrated module includes a mechanical coupling unit, a sensor unit, and a communication and synchronization unit. It is used to divide the force domains of HTS and AMB through a set of dynamic equilibrium equations, and to form a full-degree-of-freedom displacement sensing network using a laser rangefinder and a Hall array.
2. The flywheel energy storage system based on adversarial network algorithm as described in claim 1, characterized in that: The HTS module includes a ring-shaped YBaCuO superconducting bulk material, which is cooled to 77K by liquid nitrogen to maintain the superconducting state and generates axial levitation force based on the magnetic flux pinning effect. The formula for levitation force is: in, For levitation force, The critical current density of the superconducting bulk material. The strength of the pinning magnetic field, The volume obtained to calculate the geometry of the superconducting bulk material; The diameter ratio of the toroidal YBaCuO superconducting bulk material is 3:2, and the thickness is 10-20 mm. The axial suspension stiffness is optimized by arranging it in a toroidal array. The formula for axial suspension stiffness is as follows: in, For levitation force, Let l be the axial suspension stiffness, l be the number of layers in the shaped array, and d be the spacing between the shaped arrays. For axial air gap, The critical current density of the superconducting bulk material. Pi is the mathematical constant of a circle.
3. The flywheel energy storage system based on adversarial network algorithm as described in claim 2, characterized in that: The AMB module includes an 8-pole radially symmetrical electromagnetic coil array with 100 turns per pole, a wire diameter of 1mm, and a rotor surface plated with 0.5mm silicon steel sheets to reduce magnetic resistance. The formula for electromagnetic force is: in, Electromagnetic force, Let N be the permeability of free space, N be the number of turns in the coil, and I be the current flowing through it. It is a radial air gap. Let π be the mathematical constant pi, r be the radius of the rotor, and L be the axial length of the rotor. The PID controller includes dynamically adjusting the coil current based on the radial offset feedback from the displacement sensor, as expressed by the formula: in, Where 'e' represents the radial air gap and 'e' represents the radial displacement deviation. The radial displacement deviation varies with time t, where t is time. The current output by the controller. This is the proportionality coefficient. For accurate tracking of the reference air gap, The integral coefficient is... is the differential coefficient.
4. The flywheel energy storage system based on adversarial network algorithm as described in claim 3, characterized in that: The axial-radial decoupling integration includes clarifying the functional division of the HTS module and the AMB module, establishing a dynamic equilibrium equation set based on the functional division, axial mechanical decoupling design, radial mechanical decoupling design, and combining the axial and radial mechanical decoupling designs to obtain a hybrid bearing system. The formula for establishing the dynamic equilibrium equation system is as follows: in, For levitation force, Electromagnetic force, To provide a stable levitation force in the axial direction based on the magnetic flux pinning effect, This refers to the electromagnetic force generated when the rotor rotates at high speed in the radial direction. For the weight of the rotor, Let be the centrifugal force generated when the rotor rotates at high speed, m be the rotor mass, g be the acceleration due to gravity, and r be the rotor radius. ω is the angular velocity of the rotor.
5. A flywheel energy storage method based on adversarial network algorithm, characterized in that: include, A hybrid bearing system was constructed, and an electromagnetic-thermal coupling model was established based on finite element analysis to generate an initial parameter-performance dataset. Generative adversarial networks are used to optimize the parameters of a hybrid bearing system. Random noise is input into the generator to generate candidate parameter combinations, and the discriminator is used to perform adversarial training against a real dataset. The parameter combinations are then optimized by combining a physical constraint loss function. The optimized parameters were input into the finite element analysis model for simulation verification, the suspension force and stability index were extracted, and the suspension stability and dynamic response performance were verified.
6. The flywheel energy storage method based on adversarial network algorithm as described in claim 5, characterized in that: The establishment of the electromagnetic-thermal coupling model includes: establishing a three-dimensional electromagnetic-thermal coupling model in COMSOL Multiphysics; defining the Bean critical state model of YBaCuO superconductor, the nonlinear BH curve of copper coil and the magnetic properties of NdFeB permanent magnet; setting liquid nitrogen cooling boundary conditions and rotor motion constraints; and calculating levitation force and electromagnetic force through a transient solver. The optimization of parameters of the hybrid bearing system includes: the generator receiving a combination of parameters generated by 100-dimensional Gaussian noise, and the discriminator inputting the actual parameters and performance scores. The adversarial loss function is expressed as: in, To counteract the loss function, For mathematical expectation operators, For the discriminator function, For the true parameter combination generated through finite element analysis, Let z be the candidate parameter combination generated by the generator based on random noise z, where z is a random noise vector following a Gaussian distribution. For physical constraint functions, The finite element simulation performance results are for the parameter combination θ. Here, A represents the parameter combination, and A represents the target performance. For the total loss function, The weighting coefficients are used to balance the adversarial loss and the physical constraint loss; the hyperparameter λ of the total loss function is optimized through grid search.
7. The flywheel energy storage method based on adversarial network algorithm as described in claim 6, characterized in that: The generator includes a multi-layer neural network, and its output parameters include the number of turns of the current coil, the air gap, and the PID gain parameters. The discriminator includes a convolutional neural network, which takes a combination of input parameters and a performance score as input and outputs a probability of authenticity. The optimization of parameter combinations using the physical constraint loss function includes calculating the Euclidean distance between the finite element simulation performance and the target performance, expressed by the formula: in, Let be the finite element analysis simulation performance value of the i-th item, and be the preset performance value of the i-th item, where i is the variable index and n is the total number of performance indicators. For Euclidean distance.
8. The flywheel energy storage method based on adversarial network algorithm as described in claim 7, characterized in that: The extraction of suspension force and stability indicators includes comparing the magnitude of the suspension force obtained from the optimized simulation with the data before optimization, calculating the percentage change in suspension force, comparing the distribution cloud map of suspension force before and after optimization, and observing whether the uniformity and rationality of the suspension force distribution have been improved. The formula for calculating the percentage change in levitation force is expressed as: Where P represents the percentage change in levitation force. The levitation force before optimization For optimized levitation force; Stability metrics include standard deviation, extreme value difference, and coefficient of variation. The formula for calculating standard deviation is: Where v is the data sample size. Standard deviation Let j be the levitation force value of sample j, where j is the variable index. The mean levitation force is given by ; the extreme value difference is the difference between the maximum and minimum levitation force. The formula for the coefficient of variation is expressed as: in, The coefficient of variation is 1. Standard deviation This represents the average levitation force.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the flywheel energy storage method based on the adversarial network algorithm as described in any one of claims 5 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the flywheel energy storage method based on the adversarial network algorithm as described in any one of claims 5 to 8.
