A magnetic drive rotor twin control method based on closed loop optimization

Abstract

The application discloses a kind of magnetic drive rotor twin control methods based on closed loop optimization, comprising: step 1, collection magnetic drive rotor operating data and utilize synchronous clock signal to carry out time stamp alignment and data fusion, generate synchronous data vector;Step 2, based on synchronous data vector and preset digital twin model, through error function identification key time-varying parameter estimated value, according to key time-varying parameter estimated value update digital twin model;Step 3, prediction is carried out using updated digital twin model, and prediction information package is obtained;Step 4, according to prediction information package, the deviation of the prediction performance and the actual performance of digital twin model is calculated, dynamically adjusts control strategy, generates collaborative control instruction;Step 5, execute collaborative control instruction, drive magnetic drive rotor to operate, optimize digital twin model, return step 1 and form closed loop control cycle.

Description

A magnetically driven rotor twin control method based on closed-loop optimization Technical Field

[0001] This invention relates to a digital twin collaborative control method for a magnetically driven rotor, and more particularly to a magnetically driven rotor twin control method based on closed-loop optimization. Background Technology

[0002] Magnetic drive rotor systems (including magnetic levitation motors, magnetic levitation flywheel energy storage devices, and active magnetic bearing support systems) utilize controllable electromagnetic fields to achieve contactless levitation and drive of the rotor. They possess significant advantages such as no mechanical friction, ultra-high speed limits, and long lifespan, making them key core equipment in aerospace, high-precision manufacturing, and clean energy fields. However, these systems are inherently strongly coupled, open-loop unstable, and nonlinear dynamic systems. In actual high-speed operation, influenced by factors such as temperature rise, material aging, and rotor centrifugal deformation, key physical parameters of the system (such as rotor residual unbalance, current and displacement stiffness of the electromagnetic bearings, and equivalent damping) exhibit significant nonlinear time-varying characteristics. This poses a severe challenge to high-precision modeling and the implementation of high-performance control strategies.

[0003] Twin control technology, as an advanced application of digital twins in the control field, aims to provide feedback to physical control through high-fidelity virtual models, enabling intelligent perception and precise decision-making of physical entities. However, existing twin control technologies for magnetically driven rotor systems still suffer from significant open-loop defects: on the one hand, each functional module (such as parameter identification, state prediction, and control decision-making) often operates independently, lacking a system-level closed-loop linkage mechanism, making it difficult to dynamically match the frequency of parameter updates with the cycle of control decisions; on the other hand, there is a lack of a real-time feedback optimization mechanism based on the deviation between predicted and actual performance, preventing the system from correcting the model or adjusting the strategy based on the current control effect. This separation between perception, modeling, and control makes it difficult for existing systems to form a closed-loop whole with autonomous evolution and collaborative optimization capabilities, limiting the adaptive capability and control accuracy of magnetically driven rotor systems under complex and variable operating conditions. Summary of the Invention

[0004] Purpose of the invention: The technical problem to be solved by the present invention is to provide a magnetic drive rotor twin control method based on closed-loop optimization, which addresses the shortcomings of the existing technology.

[0005] To address the aforementioned technical problems, this invention discloses a magnetically driven rotor twin control method based on closed-loop optimization, comprising the following steps:

[0006] Step 1: Collect the operating data of the magnetically driven rotor and use the synchronous clock signal to perform timestamp alignment and data fusion to generate a synchronous data vector;

[0007] Step 2: Based on the synchronized data vector and the preset digital twin model, identify the estimated values ​​of key time-varying parameters through the error function, and update the digital twin model according to the estimated values ​​of key time-varying parameters;

[0008] Step 3: Use the updated digital twin model to make predictions and obtain the prediction information package;

[0009] Step 4: Calculate the deviation between the predicted performance and the actual performance of the digital twin model based on the predicted information package, dynamically adjust the control strategy, and generate coordinated control commands.

[0010] Step 5: Execute the coordinated control command to drive the magnetic drive rotor, optimize the digital twin model, and return to Step 1 to form a closed-loop control cycle.

[0011] Furthermore, the generation of the synchronization data vector mentioned in step 1, that is, the acquisition of magnetic drive rotor operation data to form the original data vector. Using a synchronous clock signal Perform timestamp alignment and data fusion to generate a synchronized data vector. ;in,

[0012] Data fusion employs a weighted average or Kalman filter algorithm, while timestamp alignment uses the IEEE 1588 precision clock protocol.

[0013] Furthermore, the step 2, which involves updating the digital twin model based on the estimated values ​​of key time-varying parameters, includes:

[0014] Step 2-1: Preset the digital twin model as a parameterized state-space model. Parameter identification is performed on it;

[0015] Step 2-2: Minimize the error function between the digital twin model output and the measured data. Obtain estimates of key time-varying parameters in a digital twin model. ;

[0016] Steps 2-3 involve estimating the key time-varying parameters. Substituting into the digital twin model, we obtain the updated digital twin model. ;

[0017] Among them, key time-varying parameters It includes one or more of the following: magnetic bearing displacement stiffness, magnetic bearing current stiffness, rotor imbalance, and electromagnetic bearing equivalent damping.

[0018] Furthermore, the parameter identification described in step 2-1 employs either recursive least squares or extended Kalman filtering.

[0019] Furthermore, the prediction using the updated digital twin model described in step 3 includes:

[0020] Step 3-1, based on the updated digital twin model In the prediction time domain Built-in optimization objective function ;

[0021] Step 3-2, Solve the objective function The optimal predictive control sequence is obtained. and predicted state sequence Generate prediction information package ;

[0022] Among them, the optimization objective function mentioned in step 3-1 , means as follows:

[0023]

[0024] in, for Step-by-step prediction output, For reference trajectory, To control the increment, and This is the weight matrix.

[0025] Furthermore, the collaborative control instructions mentioned in step 4 include control instructions and optimization instructions.

[0026] Furthermore, the generation of cooperative control instructions in step 4 includes:

[0027] Step 4-1, control performance evaluation, calculate the prediction performance based on the digital twin model. Compared with actual performance based on measured data deviation ;

[0028] Step 4-2, based on the preset threshold With deviation Based on the comparison results, collaborative optimization decisions are made, control instructions are generated, and optimization instructions are generated as needed;

[0029] Step 4-3, send control commands The actuator sent to the magnetically driven rotor, if an optimization instruction is generated, will then execute the optimization instruction. Feedback is sent to the digital twin model.

[0030] Furthermore, step 4-1 involves calculating the prediction performance based on the digital twin model. Compared with actual performance based on measured data deviation ,include:

[0031] Step 4-1-1, set the length as The evaluation window is used to calculate the prediction performance. Compared with actual performance The assessment window covers the period from the current moment. The most recent Each control cycle, i.e., a time interval ;

[0032] Step 4-1-2, based on the synchronization data vector The measured output data extracted from and control objectives The weighted quadratic performance index method is used to calculate the actual performance. , means as follows:

[0033]

[0034] in, It is the magnetically driven rotor at any time The actual output data, At any moment The reference input vector is the control target. At any moment Implemented control increment vector, and It is a weight matrix corresponding to the actual performance evaluation;

[0035] Step 4-1-3, The predictive performance The calculation method adopts the actual performance The same function is used for calculation, but the measured data is replaced with historical prediction data, as shown below:

[0036]

[0037] in, It is a digital twin model in Always The predicted value output at time step. and It is the weight matrix corresponding to the predictive performance evaluation;

[0038] Step 4-1-4, Calculate actual performance and prediction performance The deviation is obtained. The details are as follows:

[0039]

[0040] Furthermore, the collaborative optimization decision-making described in step 4-2 includes:

[0041] Using deviation value , indicating deviation The size, and a preset deviation threshold. ;

[0042] like Then, the first-level decision, namely normal fine-tuning, is made, which specifically includes:

[0043] Directly adopt the optimal control sequence The first control variable in This is determined as the control instruction to be distributed in the current control cycle. And no optimization instructions ;

[0044] If the condition is met for a continuous period exceeding the preset number of control cycles Then, a second-level decision-making process is performed, namely, model-controller collaborative optimization, which specifically includes:

[0045] Generate and output the first optimization instruction Adjust the parameters of the digital twin model;

[0046] Monitor the deviation value within a preset number of control cycles thereafter. The change in deviation value If the average improvement rate is lower than the preset threshold or continues to increase, a second optimization instruction will be generated and output. For the digital twin model, perform parameter identification in step 2-1, correct the key time-varying parameters that have the greatest impact on the deviation calculation, and update the digital twin model;

[0047] If the deviation value Failed to fall back to the threshold within the preset safe time window Below, or Continuous The single-cycle increase exceeds the deviation threshold. Given a predetermined ratio, the optimal predictive control sequence obtained from the current digital twin model is rejected. A control strategy is selected from a pre-set set of backup control laws, and the control quantity generated by this strategy is determined as the current control command. At the same time, a status alarm is reported.

[0048] Furthermore, the method is executed within a collaborative control architecture comprised of a physical entity subsystem and a digital twin subsystem, wherein:

[0049] The physical entity subsystem includes a magnetically driven rotor body, a sensing system, and a physical control unit;

[0050] The digital twin subsystem includes a data acquisition and fusion module, a parameter identification and update module, an adaptive predictive control module, and a collaborative optimization arbitrator module.

[0051] The digital twin subsystem interacts with the physical entity subsystem via a communication network.

[0052] Beneficial effects:

[0053] 1. This invention effectively solves the problem of limited overall system performance caused by the disconnect between perception, modeling, and control processes by introducing a closed-loop optimization architecture with a collaborative optimization arbiter at its core.

[0054] 2. This invention dynamically adjusts the model identification strategy and controller parameters based on real-time feedback, ensuring that the system maintains high control accuracy throughout its entire lifecycle.

[0055] 3. This invention enhances the adaptability to time-varying parameters and external disturbances by continuously comparing the deviation between predicted and measured performance.

[0056] 4. The present invention adopts an integrated collaborative design to optimize internal information flow, reduce decision delay, and ensure the feasibility and efficiency of the method in deployment at the industrial edge. Attached Figure Description

[0057] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0058] Figure 1 is a schematic diagram of the overall system architecture and closed-loop collaboration of the present invention.

[0059] Figure 2 is a flowchart of the collaborative control method based on closed-loop optimization of the present invention.

[0060] Figure 3 is a flowchart of the internal logic decision-making process of the collaborative optimization arbitrator of the present invention.

[0061] Figure 4 is a schematic diagram of the evolution of control performance deviation and decision-making stage in one embodiment.

[0062] Figure 5 is a schematic diagram of the control commands output by the coordination and optimization arbitrator in one embodiment.

[0063] Figure 6 is a schematic diagram comparing the effects of the method of the present invention with other methods in one embodiment. Detailed Implementation

[0064] This invention proposes a digital twin control method for magnetically driven rotors based on closed-loop optimization. The overall approach is as follows: by designing a collaborative control architecture driven by closed-loop optimization, the information and control flows between various technical modules are optimized and integrated, achieving real-time dynamic collaboration and closed-loop optimization. This constructs an intelligent control system capable of autonomous learning, self-evolution, and collaborative optimization based on real-time feedback. The ultimate goal is to fully explore the potential of digital twin technology in improving the overall performance, operating efficiency, and long-term reliability of magnetically driven rotor systems, promoting their development towards a higher level of intelligence.

[0065] This invention provides a closed-loop optimization-based magnetic drive rotor twin control method. By constructing a system architecture centered on a collaborative optimization arbiter, it aims to address the problem of limited overall control performance in current digital twin systems caused by the independent and uncoordinated nature of sensing, modeling, and control components. Specifically, this method dynamically adjusts the model identification strategy and control parameters using real-time feedback information, enabling the system to maintain high control accuracy throughout its entire lifecycle. By continuously comparing the performance deviation between predicted outputs and measured results, it improves the adaptability to uncertainties such as time-varying system parameters and external disturbances. Furthermore, it achieves efficient integration and collaborative optimization of information flows from multiple modules within the system, reducing decision latency and thus enhancing the feasibility and operational efficiency of this method in industrial edge computing environments.

[0066] The specific technical solution is as follows:

[0067] A magnetic drive rotor twin control method based on closed-loop optimization has an overall system architecture as shown in Figure 1. The magnetic drive rotor digital twin cooperative control method based on closed-loop optimization is implemented in a cooperative control architecture consisting of a physical entity subsystem 100 and a digital twin subsystem 200.

[0068] The physical entity subsystem 100 specifically includes a magnetically driven rotor body 101, a sensing system 102 deployed thereon, and a physical control unit 103; the sensing system includes displacement sensors, current sensors, and acceleration sensors, etc., for collecting information such as rotor center displacement, control current, and triaxial acceleration, which together constitute the system output raw data vector. The physical control unit 103 is responsible for receiving control commands generated by the digital twin subsystem 200. And drive the magnetically driven rotor body 101;

[0069] The digital twin subsystem 200 is deployed on a high-performance edge computing platform, and its core functions include the following modules: data acquisition and fusion module 201, parameter identification and update module 202, adaptive predictive control module 203, and collaborative optimization arbitrator module 204, which is the core of this invention.

[0070] The system data acquisition and fusion module 201 is based on the system output. Perform time-series synchronization and fusion operations to form a synchronized data vector. The parameter identification and update module 202 identifies and updates parameters based on the synchronization data vector. Perform parameter identification to obtain the optimal parameter estimates for the current time step, and update the digital twin model. The adaptive predictive control module 203 uses a digital twin model. Based on the system performance evolution analysis, a model predictive control strategy is used to generate the optimal prediction sequence and the prediction state sequence, which together form the prediction information package. The collaborative optimization arbiter module 204 performs key closed-loop optimization functions, specifically evaluating the current system control performance, dynamically adjusting the optimization strategy, and generating the final output control command. and / or optimization instructions sent to other modules ;

[0071] The digital twin subsystem 200 interacts with the physical entity subsystem 100 via a local area network or serial communication bus.

[0072] The specific execution flow of the method of the present invention is shown in Figure 2, and includes the following steps:

[0073] Step S1: Multi-source heterogeneous data acquisition and time-series synchronization fusion;

[0074] The sensing system 102, deployed in the physical entity subsystem 100, collects multi-source heterogeneous operating data such as rotor displacement, current, and triaxial acceleration in real time, forming a raw data vector:

[0075]

[0076] To achieve precise synchronization, a statistical clock protocol is required to provide a time base. This embodiment preferably uses the IEEE 1588 (PTP) precision clock protocol to assign a unified timestamp to all sensor data. After receiving this data, the data acquisition and fusion module 201 performs timestamp alignment and redundant data removal operations, and uses algorithms such as weighted average or Kalman filtering as the data fusion function. Generate high-confidence synchronization data vectors:

[0077]

[0078] The vector is distributed in parallel via a real-time data bus to the physical control unit 103 in the physical entity subsystem 100 and subsequent modules in the digital twin subsystem 200.

[0079] Step S2: Identification of key parameters and model update of the online system;

[0080] The parameter identification and update module 202 is based on the received synchronization data vector. and the preset parameterized state-space model of the magnetically driven rotor Real-time identification is performed. , which is the key time-varying parameter vector to be identified. In a preferred embodiment, the key time-varying parameter vector... This mainly includes magnetic bearing displacement stiffness, magnetic bearing current stiffness, rotor imbalance, and electromagnetic bearing equivalent damping.

[0081] The parameter identification and update module 202 outputs the minimized digital twin model. Compared with physical entity measured output The error function between them, to achieve the... Accurate identification. The error function is defined as:

[0082]

[0083] The optimal parameters are estimated in real time using a computationally efficient recursive least squares algorithm. Subsequently, the digital twin model is based on the identification of optimal parameters. Dynamically updated to This ensures that the virtual model can accurately track the operating status of the physical entity and bridge the dynamic characteristic changes caused by temperature rise, aging, etc.

[0084] Step S3: Collaborative prediction and decision-making oriented towards optimization objectives;

[0085] Adaptive predictive control module 203 is based on an updated high-precision digital twin model Set the prediction time domain and control time domain An optimization objective function is constructed based on one or more comprehensive performance indicators of the control system. In this embodiment, minimizing vibration is preferred as the control objective. The optimization objective function is constructed as follows:

[0086]

[0087] in, Based on the current The first time information was made Step-by-step prediction output, For the reference trajectory at the corresponding time, To control the increment, and This is the weight matrix, which can be initially set to empirical values.

[0088] The adaptive predictive control module 203 solves the optimization problem in each control cycle to obtain the future... The optimal control sequence for the step:

[0089]

[0090] and its corresponding predicted state sequence And the prediction information packet:

[0091]

[0092] Send to the collaborative optimization arbiter module 204.

[0093] Step S4: Control decision-making collaborative optimization adjudication and instruction distribution;

[0094] This step is the core of the closed-loop optimization in this invention. Its internal logical decision-making process is detailed in Figure 3, and specifically includes the following sub-steps:

[0095] Sub-step S4.1: Control performance evaluation;

[0096] The collaborative optimization arbiter module 204 receives the prediction information packet from step S3. and the latest real-time synchronized data vector from step S1 . It includes the actual system output after fusion of current and recent historical moments. For example, physical quantities such as rotor displacement and current.

[0097] The core of performance evaluation is to calculate and compare short-term actual performance indicators based on measured data. Compared with short-term forecasting performance metrics based on digital twin models Both are in the same space with a length of The calculation is performed within the evaluation window, which covers the period from the current time. The most recent Each control cycle, i.e., a time interval .

[0098] Characterized in the past The actual control effect of the physical entity system within a control cycle. Its calculation depends on the... The measured output data extracted from and the system's control objectives .

[0099] The actual performance indicators The following weighted quadratic performance index is used for calculation:

[0100]

[0101] in, At any moment The actual system output direction, At any moment The reference input vector, At any moment Implemented control increment vector, and It is the weight matrix corresponding to the actual performance evaluation, usually a positive semi-definite matrix or a positive definite matrix. and The value of can be related to the weight matrix of the objective function in the adaptive predictive control module 203. , They can be configured independently depending on the different evaluation focuses.

[0102] Characterized in the same past Control cycle The control effects predicted by the digital twin model require the use of historical prediction data stored in the prediction information package.

[0103] Specifically, for each historical moment within the evaluation window The adaptive predictive control module 203 generated a predicted state sequence at that time, from which the state corresponding to time was extracted. One-step prediction output .

[0104] The predictive performance metrics The calculation uses a function with the same form as the actual performance index, but replaces the measured data with historical prediction data:

[0105]

[0106] in, It is a digital twin model in Always The predicted value output at time step. and This corresponds to the weight matrix for predictive performance evaluation, and is typically set to maintain consistency in the evaluation benchmark. , .

[0107] Calculate and Then, the collaborative optimization arbiter module 204 calculates the deviation between the two:

[0108]

[0109] This deviation This effectively quantifies the difference between the prediction accuracy and actual control performance of the digital twin model within the current evaluation window. Significant positive deviation ( A deviation much greater than 0 usually means that the model is too optimistic and the actual system performs worse than the model's predictions; a significant negative deviation (which may occur under certain performance function definitions) may mean that the model is too conservative. The magnitude of the value directly reflects the degree of matching between the model and the entity, as well as the effectiveness of the current control strategy, providing a key basis for subsequent collaborative optimization decisions.

[0110] Sub-step S4.2. Collaborative optimization decision-making

[0111] The collaborative optimization arbiter module 204 will calculate the absolute value of the performance deviation. Deviation threshold from preset The comparison is performed, and based on the comparison results and the changing trend of ∆J(t), the following hierarchical decision-making logic is executed:

[0112] Level 1 Decision (Normal Fine-tuning): If This indicates that the digital twin model and the physical entity currently match well, and the predictive control sequence is reliable. At this point, the collaborative optimization arbitrator module 204 directly adopts the optimal control sequence from the adaptive predictive control module 203. The first control quantity in This is determined as the control instruction to be distributed in the current control cycle. Under this decision, optimized instructions typically are not sent to other modules. .

[0113] Second-level decision (model and controller co-optimization): If Furthermore, this state has lasted for more than the preset number of decision delay cycles. (Preferred in this invention) This indicates a significant and persistent deviation between the model predictions and actual performance, potentially stemming from model parameter mismatch or changes in external disturbance characteristics. In this case, the collaborative optimization arbitrator module 204 initiates the optimization procedure according to a preset step-by-step optimization strategy:

[0114] Level 1, Online Adjustment of Controller Parameters: The arbiter first generates optimization instructions. The adaptive predictive control module 203 is then used. The optimization strategy can be selected based on the deviation characteristics. If enhanced system robustness is required, the control increment weight matrix can be increased. If it is necessary to improve the dynamic response, the prediction time domain or control time domain parameters can be adjusted by command.

[0115] Level 2, Model Parameter Re-identification Triggered: After the controller adjustment is executed, the arbitrator continues to monitor subsequent consecutive... One control cycle (preferred in this invention) ).like The average improvement rate is lower than the preset threshold, or If the value continues to increase, it is determined that the model needs to be updated. The adjudicator then generates instructions. The parameter identification and update module 202 is then triggered to urgently re-identify the key parameters.

[0116] Level 3, Security Assurance and Policy Switching: If the above optimizations are performed, During the security control cycle Inside (preferred in this invention) The water level has not fallen back to the safe threshold. Below, or Continuous The single-cycle increase exceeds the deviation threshold. of In such cases, the arbitrator will switch to a preset backup robust control strategy (such as distributed PID control with fixed parameters) and report an alarm.

[0117] Sub-step S4.3. Closed-loop instruction dispatch:

[0118] Based on the decision result of sub-step S4.2, the collaborative optimization arbitrator module 204 executes the following instruction distribution operation:

[0119] Inevitable action: The currently determined control command will be executed. The information is sent to the physical control unit 103 in the physical entity subsystem 100 via the communication interface. Regardless of the decision-making process, a definite outcome is required for this cycle. It is issued to ensure the continuity of control.

[0120] Conditional action: If an optimization instruction is generated during the decision-making process of sub-step S4.2. Corresponding to the above-mentioned optimization operations such as controller tuning and model re-identification, these instructions are synchronously distributed to the corresponding functional modules inside the digital twin subsystem 200, namely the adaptive predictive control module 203 or the parameter identification and update module 202.

[0121] Step S5: Control command execution and closed-loop cycle;

[0122] The physical control unit 103 in the physical entity subsystem 100 receives and executes control commands. This drives the magnetically driven rotor body 101 to operate, completing one control cycle. Simultaneously, related modules within the digital twin subsystem 200 (if receiving optimization instructions)... Then, corresponding parameter adjustments and model re-identification are performed simultaneously. The system then returns to step S1, starting a new cycle of data acquisition, fusion, identification, prediction, decision-making, and execution based on the new system state. Through this cyclical, closed-loop operation that includes decision feedback, the entire system achieves deep collaboration between perception, modeling, prediction, and control, enabling it to continuously adapt to dynamic changes in the system and constantly optimize itself.

[0123] Example:

[0124] This embodiment provides a specific simulation implementation plan. Referring to the architecture shown in Figure 1, a simulation model of a magnetic drive rotor digital twin cooperative control system is constructed in the simulation environment.

[0125] 1. Simulation parameter settings:

[0126] Controlled object: Establish a five-degree-of-freedom magnetically driven rotor model, whose key parameters (such as the displacement stiffness of the radial magnetic bearing) are... and current stiffness The value is set to have slow time-varying characteristics to simulate the temperature rise effect. For example, 5 seconds after the simulation starts, let... From initial value linearly reduced to .

[0127] Comparison object: A set of traditional PID controllers with proven performance were used as the comparison benchmark, and their parameters remained constant throughout the simulation.

[0128] System parameters for this embodiment:

[0129] The initial parameters of the digital twin model are consistent with the initial values ​​of the physical model, but the subsequent time-varying patterns are unknown. The bias threshold of the co-optimized arbitrator is then determined. Control time domain of model predictive control Predicting the time domain Evaluation window length .

[0130] 2. Simulation process and result analysis:

[0131] During simulation, the system operates according to the process shown in Figure 2. Figures 4 and 5 illustrate the dynamic decision-making process of the collaborative optimization arbitrator module 204 and its output control command sequence in this simulation.

[0132] Phase 1 (0-4 seconds, stable operation period): The system's initial state is stable, the model is accurate, as shown in Figure 4, performance deviation. Always below the threshold As shown in Figure 5, the arbiter remains in the "Level 1 Decision (Normal Fine-tuning)" state, directly outputting the optimal control command calculated by Model Predictive Control (MPC). At this point, the instruction curve is smooth, and the system is operating in its optimal performance range.

[0133] Phase 2 (5-7 seconds, parameter drift response period): Starting from the 5th second, as shown in Figure 4, the displacement stiffness of the physical model... It begins to drift slowly. This causes the model to gradually mismatch, and the actual performance begins to deviate from the predicted performance. Continuously exceeding the threshold The arbiter then enters the "second-level decision-making (model and controller co-optimization)" state. As shown in Figure 5, the arbiter first issues optimization instructions to fine-tune the controller weights, which is manifested as instructions. The high-frequency components are moderately suppressed, the command curve becomes smoother, and an attempt is made to overcome model errors with greater robustness.

[0134] Phase 3 (Model Learning and Recovery Period after 7 seconds): Due to persistent parameter drift, controller adjustments alone cannot completely eliminate the bias. Around the 7th second, the arbitrator initiates emergency parameter re-identification. The digital twin model quickly tracks and learns the changes in physical parameters, and model accuracy recovers rapidly. As can be seen from Figures 4 and 5, thereafter... The temperature drops rapidly and eventually falls back below the threshold. The arbiter then returns to the "normal fine-tuning" state, and the control commands also return to the fine-tuning mode.

[0135] To highlight the comprehensive performance advantages of the method of the present invention, as shown in Figure 6, the rotor center offset of the system using the method of the present invention and the system using a traditional fixed-parameter PID control system (taking one radial degree of freedom as an example) are compared under the same parameter drift disturbance.

[0136] The method of this invention (as shown by the solid line) states that after parameter drift occurs (at the 5th second), due to the presence of the collaborative optimization arbitrator, the system rapidly adapts to the changes in the controlled object by adjusting the controller and updating the model online. The rotor offset only fluctuates slightly during the transition phase and is quickly suppressed to within the allowable accuracy range, demonstrating extremely strong adaptability and robustness.

[0137] Traditional PID control (as shown by the dashed line): Because its controller parameters are fixed, it cannot detect changes in the characteristics of the object. When parameter drift occurs, the system performance continues to deteriorate, the rotor offset increases significantly and fluctuates considerably, the control accuracy decreases severely, and there is even a risk of instability.

[0138] This simulation example demonstrates that the closed-loop optimization-based digital twin cooperative control method for magnetically driven rotors provided by this invention can effectively address the challenges posed by time-varying internal system parameters. Through the intelligent decision-making of the cooperative optimization arbiter, online cooperative optimization of controller parameters and model parameters is achieved, significantly improving the system's control accuracy, adaptability, and robustness, with results far superior to traditional control methods.

[0139] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention's content regarding a closed-loop optimized magnetic drive rotor twin control method, as well as some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0140] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0141] This invention provides a concept and method for a magnetically driven rotor twin control method based on closed-loop optimization. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A magnetically driven rotor twin control method based on closed-loop optimization, characterized in that, Includes the following steps: Step 1: Collect the operating data of the magnetically driven rotor and perform timestamp alignment and data fusion using a synchronous clock signal to generate a synchronous data vector; Step 2: Based on the synchronous data vector and a preset digital twin model, identify the estimated values ​​of key time-varying parameters through an error function, and update the digital twin model according to the estimated values ​​of key time-varying parameters; Step 3: Use the updated digital twin model for prediction to obtain a prediction information packet; Step 4: Calculate the deviation between the predicted performance and actual performance of the digital twin model based on the prediction information packet, dynamically adjust the control strategy, and generate cooperative control commands; Step 5: Execute the cooperative control commands to drive the magnetically driven rotor, optimize the digital twin model, and return to Step 1 to form a closed-loop control cycle; wherein, the generation of cooperative control commands in Step 4 includes: Step 4-1, control performance evaluation, calculating the predicted performance based on the digital twin model prediction. Compared with actual performance based on measured data deviation Step 4-2, based on the preset threshold With deviation Based on the comparison results, collaborative optimization decisions are made, control commands are generated, and optimization commands are generated as needed; step 4-3, the control commands are... The actuator sent to the magnetically driven rotor, if an optimization instruction is generated, will then execute the optimization instruction. Feedback is sent to the digital twin model; the collaborative optimization decision-making described in step 4-2 includes: using the deviation value , indicating deviation The size, and a preset deviation threshold. ;like Then, the first-level decision, namely normal fine-tuning, is performed, which specifically includes: directly adopting the optimal control sequence. The first control variable in This is determined as the control instruction to be distributed in the current control cycle. And no optimization instructions If the preset number of control cycles is met continuously... Then, the second-level decision-making process, namely model-controller collaborative optimization, is performed, which specifically includes: generating and outputting the first optimization instruction. The parameters of the digital twin model are adjusted; the deviation value is monitored within a preset number of control cycles thereafter. The change in deviation value If the average improvement rate is lower than the preset threshold or continues to increase, a second optimization instruction will be generated and output. Upon reaching the digital twin model, perform parameter identification in step 2-1, correct the key time-varying parameters that have the greatest impact on the deviation calculation, and update the digital twin model; if the deviation value The value did not fall back to the threshold within the preset safe time window. Below, or Continuous The single-cycle increase exceeds the deviation threshold. Given a predetermined ratio, the optimal predictive control sequence obtained from the current digital twin model is rejected. A control strategy is selected from a pre-set set of backup control laws, and the control quantity generated by this strategy is determined as the current control command. At the same time, a status alarm is reported.

2. The magnetic drive rotor twin control method based on closed-loop optimization according to claim 1, characterized in that, The generation of the synchronization data vector in step 1 refers to the acquisition of magnetically driven rotor operating data to form the original data vector. Using a synchronous clock signal Perform timestamp alignment and data fusion to generate a synchronized data vector. 。 3. The magnetic drive rotor twin control method based on closed-loop optimization according to claim 2, characterized in that, The step 2, which involves updating the digital twin model based on the estimated values ​​of key time-varying parameters, includes: Step 2-1, where the digital twin model is preset to be a parameterized state-space model. Step 2-2 involves parameter identification of the digital twin model; step 2-2 involves minimizing the error function between the digital twin model output and the measured data. Obtain estimates of key time-varying parameters in a digital twin model. Steps 2-3 involve estimating the key time-varying parameters. Substituting into the digital twin model, we obtain the updated digital twin model. 。 4. The magnetic drive rotor twin control method based on closed-loop optimization according to claim 3, characterized in that, The parameter identification described in step 2-1 uses either the recursive least squares method or the extended Kalman filter method.

5. The magnetic drive rotor twin control method based on closed-loop optimization according to claim 4, characterized in that, The prediction using the updated digital twin model described in step 3 includes: Step 3-1, based on the updated digital twin model In the prediction time domain Built-in optimization objective function Step 3-2: Solve the objective function. The optimal predictive control sequence is obtained. and predicted state sequence Generate prediction information package 。 6. The magnetic drive rotor twin control method based on closed-loop optimization according to claim 5, characterized in that, The collaborative control instructions mentioned in step 4 include control instructions and optimization instructions.

7. The magnetic drive rotor twin control method based on closed-loop optimization according to claim 6, characterized in that, Step 4-1 describes calculating the prediction performance based on the digital twin model. Compared with actual performance based on measured data deviation This includes: Step 4-1-1, setting the length as... The evaluation window is used to calculate the prediction performance. Compared with actual performance The assessment window covers the period from the current moment. The most recent Each control cycle, i.e., a time interval Step 4-1-2, based on the synchronization data vector The measured output data extracted from and control objectives The weighted quadratic performance index method is used to calculate the actual performance. Step 4-1-3, The predictive performance The calculation method adopts the actual performance The same function is used for calculation, but the measured data is replaced with historical prediction data; step 4-1-4, calculate the actual performance. and prediction performance The deviation is obtained. 。 8. A magnetic drive rotor twin control method based on closed-loop optimization according to any one of claims 1-7, characterized in that, The method is executed in a collaborative control architecture consisting of a physical entity subsystem (100) and a digital twin subsystem (200), wherein: the physical entity subsystem (100) includes a magnetically driven rotor body (101), a sensing system (102), and a physical control unit (103); the digital twin subsystem (200) includes a data acquisition and fusion module (201), a parameter identification and update module (202), an adaptive predictive control module (203), and a collaborative optimization arbitrator module (204); the digital twin subsystem (200) interacts with the physical entity subsystem (100) through a communication network.

Images