An Adaptive Model Predictive Control Method for Magnetic Driven Rotors for Twin Control

Abstract

This invention discloses an adaptive model predictive control method for magnetically driven rotors oriented towards twin control, comprising: Step 1, establishing a predictive model of the parameterized discrete state space of the magnetically driven rotor; Step 2, starting the magnetically driven rotor system and stabilizing it using a conventional controller; Step 3, collecting measured data of rotor displacement and coil current of the magnetically driven rotor system; Step 4, continuously identifying the key parameter set of the predictive model based on the measured data to obtain real-time parameter estimates; Step 5, updating the real-time parameter estimates to the current predictive model and estimating the full state of the magnetically driven rotor; Step 6, calculating the optimal control sequence and controlling the magnetically driven rotor accordingly; Step 7, continuously repeating steps 3 to 7. This invention significantly improves the control accuracy, adaptability, and robust stability of the magnetically driven rotor system throughout its entire lifecycle.

Description

Technical Field

[0001] This invention relates to an adaptive model predictive control method for magnetically driven rotors, and more particularly to an adaptive model predictive control method for magnetically driven rotors oriented towards twin control. 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, involving deep interaction between electromagnetics, rotor dynamics, and control algorithms. In actual operating conditions, influenced by factors such as temperature rise, material aging, and high-speed 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, posing a severe challenge to high-precision modeling and the implementation of high-performance control strategies.

[0003] Model Predictive Control (MPC), with its ability to handle multivariable constraints and online rolling optimization, is considered an effective means to solve complex control problems of magnetically driven rotors. However, the control performance of MPC is highly dependent on the fidelity of its internal predictive model. When the parameters of the internal model fail to track the time-varying characteristics of the physical entity in real time and a mismatch occurs, the predictive error of the controller will increase significantly, leading to a deterioration in control accuracy, and in severe cases, even causing rotor instability and rubbing.

[0004] Twin control technology, as an advanced application of digital twins in the control field, aims to provide feedback to physical control through a high-fidelity virtual model. However, existing technologies still face bottlenecks in realizing twin control of magnetically driven rotors: on the one hand, conventional online parameter identification algorithms are often independent of the control loop, lacking deep temporal coupling with the controller's rolling optimization process; on the other hand, the model update frequency of existing architectures is low, making it difficult to meet the real-time tracking requirements within the microsecond-level control cycle of magnetically driven rotors. This loose coupling between identification and control limits the ability of the predictive model to maintain consistency between virtual and real systems throughout its entire lifecycle, failing to fully leverage the potential of twin control in terms of adaptability and robust stability. Summary of the Invention

[0005] Purpose of the invention: The technical problem to be solved by the present invention is to provide an adaptive model predictive control method for magnetically driven rotors oriented towards twin control, which addresses the shortcomings of the existing technology.

[0006] To address the aforementioned technical problems, this invention discloses an adaptive model predictive control method for magnetically driven rotors oriented towards twin control, the method comprising the following steps:

[0007] Step 1: Establish a parameterized discrete state-space prediction model for the magnetically driven rotor. This prediction model includes a set of key parameters. ;

[0008] Step 2: Start the magnetic drive rotor system and use a traditional controller to ensure its stable operation;

[0009] Step 3: Collect measured data of rotor displacement and coil current of the magnetically driven rotor system; based on the measured data, analyze the key parameter set. Continuous identification is performed to obtain real-time parameter estimates. ;

[0010] Step 4, calculate the estimated real-time parameters. Update the current prediction model and use the updated prediction model to estimate the full state of the magnetically driven rotor;

[0011] Step 5: Based on the updated prediction model and its estimated full state of the magnetically driven rotor, calculate the optimal control sequence and control the magnetically driven rotor accordingly.

[0012] Step 6: Repeat steps 3 to 6 to form a control closed loop and complete the adaptive model predictive control of the magnetic drive rotor based on digital twin.

[0013] Furthermore, the prediction model for the parameterized discrete state space described in step 1 includes a system matrix, which depends on the parameter set. ;

[0014] The prediction model is expressed as follows:

[0015]

[0016] in, for The system state vector of the magnetically driven rotor at any given time. To control the input vector, The output vector of the prediction model. , , and For the parameter set The system matrix.

[0017] Furthermore, the conventional controller described in step 2 is a proportional-integral controller;

[0018] The control law of the proportional-integral controller is as follows:

[0019]

[0020] in, This is the control vector of the proportional-integral controller. This represents the displacement deviation of the magnetically driven rotor. and These are the proportional and integral control parameters, respectively.

[0021] Furthermore, the key parameter set described in step 3... Continuous identification is performed, that is, online parameter identification is performed using the recursive least squares method;

[0022] The optimization objective of the recursive least squares method is to minimize the sum of squared errors between the predicted model output and the measured output, as expressed below:

[0023]

[0024] in, for The measured output of the magnetically driven rotor at any given time.

[0025] Furthermore, in step 4, the updated prediction model is used to estimate the full state of the magnetically driven rotor, that is, the full state variables of the magnetically driven rotor are estimated in real time through the state observer.

[0026] Furthermore, the state observer mentioned in step 4 uses a Kalman filter, that is, the Kalman filter is used to estimate the full state variables of the magnetically driven rotor in real time.

[0027] Furthermore, in step 4, the real-time estimation of the full-state variables of the magnetically driven rotor using a Kalman filter is performed, wherein the Kalman filter in each sampling period... It contains two stages: a prediction step and an update step, used for time updates and measurement updates respectively. By recursively executing the prediction and update steps, the Kalman filter outputs the optimal full-state estimate. .

[0028] Furthermore, the prediction step and update step are specifically as follows:

[0029] Prediction Step:

[0030] Based on the posterior state estimate from the previous time step and its estimated error covariance matrix Using the updated prediction model, predict the current Prior state estimate at time 1 and the prior estimation error covariance matrix Specifically, it is expressed as follows:

[0031]

[0032]

[0033] in, The process noise covariance matrix; Represents the matrix transpose operation; Indicates prior estimation; This represents the posterior estimate;

[0034] Update steps:

[0035] In obtaining Actual displacement measurement value at time 10:00 Then, the Kalman filter calculates the Kalman gain matrix. And use this gain matrix to predict the prior state values By fusing the data with the current actual displacement measurement information, the posterior state estimate at the current time can be obtained. And update the posterior estimation error covariance matrix. Specifically, it is expressed as follows:

[0036]

[0037]

[0038]

[0039] in, To measure the noise covariance matrix; It is the identity matrix; This represents the matrix inversion operation.

[0040] Furthermore, in step 5, the calculation of the optimal control sequence and the control of the magnetic drive rotor accordingly involves using a model predictive controller to solve the optimization problem in the finite time domain in each sampling period to obtain the optimal control sequence, and then applying the first control quantity of the optimal control sequence to the magnetic drive rotor to control it.

[0041] Furthermore, step 5, which involves calculating the optimal control sequence and controlling the magnetically driven rotor accordingly, specifically includes the following steps:

[0042] Step 5-1: Estimate the current state value and the updated system matrix , , , Passed to the model prediction controller;

[0043] Step 5-2, Construct the optimization problem , means as follows:

[0044]

[0045] in, To predict the time domain, For time steps, To control the time domain, Based on the current Time-prediction model and state pair The prediction output at any given time. for Reference trajectory at any moment To control the increment, and For the square of the weighted Euclidean norm and This is the weight matrix;

[0046] Step 5-3, the model predictive controller solves the optimization problem. To obtain the future The optimal control sequence for the step is represented as follows:

[0047]

[0048] in, Based on the current Time prediction model and state solution about The optimal control quantity at any given time;

[0049] Step 5-4, set the first control variable of the optimal control sequence. It is applied to the magnetically driven rotor.

[0050] Beneficial effects:

[0051] 1. The present invention provides an adaptive model predictive control method for magnetically driven rotors for twin control, which can effectively solve the problem of prediction model mismatch caused by time-varying model parameters, and avoid the performance degradation and system instability risk of traditional model predictive control.

[0052] 2. By deeply integrating online parameter identification, real-time model updates, and rolling optimization of model predictive control within a single hardware platform, this invention achieves high-frequency self-correction of the predictive model for the time-varying characteristics of the system, significantly improving control accuracy throughout the entire lifecycle.

[0053] 3. The deep adaptive mechanism proposed in this invention enhances the system's robustness against parameter drift and external disturbances. At the same time, the integrated solution reduces communication latency and ensures the real-time performance and engineering applicability of the algorithm when deployed at the industrial edge, thereby providing continuously optimal and highly reliable control performance for magnetically driven rotors. Attached Figure Description

[0054] 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.

[0055] Figure 1 This is a schematic diagram of the system architecture of the present invention.

[0056] Figure 2 This is a flowchart of the adaptive model predictive control method of the present invention.

[0057] Figure 3 This is a diagram showing the online identification and tracking effect of the key parameter of displacement stiffness coefficient in an embodiment of the present invention.

[0058] Figure 4 This is a diagram showing the online identification and tracking effect of the key parameter of current stiffness coefficient in an embodiment of the present invention.

[0059] Figure 5 This is a comparison chart of the control performance under pulse disturbance in an embodiment of the present invention. Detailed Implementation

[0060] The overall concept of this invention is as follows: it provides an adaptive model predictive control method for magnetically driven rotors oriented towards twin control. This method aims to achieve high-frequency adaptive updates of the predictive model based on real-time system operating data by constructing a deep closed-loop fusion mechanism between online parameter identification and the model predictive controller. Specifically, this invention addresses the problem that traditional model predictive control cannot effectively track the time-varying characteristics of the system due to fixed internal model parameters, resulting in decreased control accuracy and insufficient robustness. By ensuring that the model predictive controller always performs rolling optimization based on the most accurate current system model, it significantly improves the control performance, adaptability, and system stability of the magnetically driven rotor when facing internal parameter drift and external disturbances. The core of this invention is to deeply couple a digital twin with a model predictive controller (MPC) through online parameter identification technology, constructing a closed-loop control system capable of high-frequency adaptive updates based on real-time system dynamics.

[0061] Example 1:

[0062] This embodiment provides an adaptive model predictive control method for magnetically driven rotors oriented towards twin control, the overall architecture of which is as follows: Figure 1As shown, it mainly consists of two major components: the physical entity layer and the digital twin control layer;

[0063] The physical entity layer consists of a magnetically driven rotor system, mainly including key components such as electromagnetic bearings, rotors, current drivers, and displacement sensors.

[0064] The electromagnetic bearing, as the core actuator of the system, generates a controllable magnetic field under current excitation, thereby forming an electromagnetic force acting on the rotor to achieve active adjustment of the rotor position and dynamic suspension support. The rotor, as the controlled object, maintains a non-contact suspension state in the nonlinear electromagnetic force field provided by the electromagnetic bearing and can achieve stable rotation under different speed conditions according to operating requirements. The current drive receives the output signal from the control unit and converts it into a drive current with corresponding amplitude and dynamic characteristics to provide power input to the electromagnetic bearing. The displacement sensor is responsible for detecting the radial or axial offset of the rotor axis relative to the theoretical center of the electromagnetic bearing in real time. The collected position signal constitutes a key feedback variable in the closed-loop control system and is the basis for achieving stable rotor suspension and vibration suppression.

[0065] The digital twin layer is used to execute the core digital twin model and model predictive control algorithm of this invention, based on the system output generated by the physical entity layer. Perform solution analysis and algorithm iteration, and output control signals. Achieving high-stability control of the magnetically driven rotor. Specifically, the digital twin layer internally includes an online parameter identification module, an adaptive prediction model, and a model prediction controller;

[0066] The online parameter identification module is based on system output. and control signals Execute the parameter identification algorithm to obtain the estimated values ​​of the internal parameters of the current system. The digital twin prediction model is based on internal parameter estimates. Simulate the system state and output the current state estimate. The model predictive controller executes a key predictive control algorithm based on the state estimate. Solve the relevant optimization problem in the finite prediction time domain to obtain the optimal control sequence in the target control time domain, and then extract the first element of the optimal control sequence. Output, as a new control signal It is applied to the magnetically driven rotor system to achieve stable control of the system.

[0067] Based on the above system architecture, this embodiment proposes an adaptive model predictive control method for magnetically driven rotors oriented towards twin control. Its core lies in deeply coupling the digital twin with the model predictive controller (MPC) through online parameter identification technology, constructing a closed-loop control system capable of high-frequency adaptive control based on real-time system dynamics. The process of the method is as follows: Figure 2 As shown, the specific steps include:

[0068] Step S1: Establishment of the prediction model for the magnetically driven rotor;

[0069] This step forms the basis for constructing a digital twin. Based on the linearized dynamics and electromagnetic equations of the magnetically driven rotor near the equilibrium point, a predictive model in discrete state space form is established. This model is parameterized, and its system matrix depends on a set of key physical parameters. The mathematical representation of the model is as follows:

[0070]

[0071] in, for time A 3D system state vector typically includes rotor displacement and speed information, etc. for The 3D control input vector typically includes information about the magnetic bearing control current. for The 3D output vector typically includes the measurable displacement of the rotor. , , , Let be the system matrix, with dimensions respectively. , , , These are used to characterize the system's internal dynamic characteristics, the effect of input on state, the effect of state on output, and the effect of input on output, respectively; internal parameters. The time-varying characteristics are reflected in the changes of these matrix elements, and further affect the system characteristics. Parameters Includes time-varying parameters that have a significant impact on system dynamics, such as the current stiffness coefficient. and displacement stiffness coefficient In practice, Depending on the system complexity and identification requirements, other parameters can be added, such as the equivalent damping coefficient and rotor imbalance.

[0072] Step S2: System initialization and basic stability control;

[0073] After the magnetically driven rotor system starts up, an initialization phase is required before activating the high-performance adaptive MPC to bring the system to a stable operating point. During this phase, only rotor levitation control is applied; speed control is not. In this phase, a conventional proportional-integral (PI) controller, initially tuned, is used for closed-loop control. Its control law is as follows:

[0074]

[0075] in, For displacement deviation, This is a given reference equilibrium position, typically zero. and These are proportional and integral control parameters, which can be pre-tuned offline.

[0076] The goal of the initialization phase is to quickly suppress the initial rotor sway and stabilize it near the reference position, while continuously acquiring the measured displacement of the rotor. With coil current This provides a stable data foundation for subsequent high-precision parameter identification.

[0077] Step S3: Online continuous identification of key parameters;

[0078] When the system reaches stability via the PI controller, that is, when the displacement deviation... Once the fluctuation falls below a preset threshold, the online parameter identification module within the digital twin is activated. The core task of this module is to continuously estimate and track a set of key parameters using real-time operational data. The changes.

[0079] This invention preferably employs Recursive Least Squares (RLS) for online parameter identification due to its advantages of high computational efficiency and ease of online implementation. The optimization objective of the identification algorithm is to minimize the sum of squared errors between the model output and the actual measured output:

[0080]

[0081] The RLS algorithm is based on the newly acquired data in each sampling period. Recursively update parameter estimates This process is executed once in each control cycle, enabling high-frequency, continuous tracking of parameters and sensitively capturing parameter drift caused by temperature rise and changes in operating conditions.

[0082] Step S4: Adaptive update and state estimation of the prediction model;

[0083] The latest parameter set identified in step S3 This is immediately used to update the prediction model established in step S1. Specifically, it will... Substituting into the system matrix, we obtain the time-varying system matrix that best reflects the dynamics of the actual system at the current moment: , , , .

[0084] In practical systems, the system state vector The included variables, such as all displacement and velocity components of the rotor, cannot all be directly measured by displacement sensors. Therefore, the digital twin needs to utilize the updated high-precision prediction model in step S4 to integrate a state observer to perform real-time, optimal estimation of all system state variables. This invention preferably uses a Kalman filter as the state observer because it can provide a minimum mean square error estimate of the system state even in the presence of process noise and measurement noise.

[0085] The execution process of the Kalman filter consists of two core stages within each sampling period k: a prediction step (time update) and an update step (measurement update). The specific algorithm is described below:

[0086] S4.1, Prediction Step;

[0087] Based on the posterior state estimate from the previous time step and its estimated error covariance matrix Using the updated system model, predict the prior state estimate at the current time (time k). and the prior estimation error covariance matrix Its core formula is:

[0088] State prediction:

[0089] Covariance prediction:

[0090] in, This is the process noise covariance matrix, used to characterize the degree of uncertainty in the model; superscript Represents the matrix transpose operation; superscript symbol Indicates a priori estimate, i.e., a prediction incorporating the current measurement; superscript symbol This represents the posterior estimate, which is the optimal estimate after incorporating the current measurement value.

[0091] S4.2, Update Step:

[0092] Obtain the actual displacement measurement value of the sensor at time k. Then, the Kalman filter calculates the Kalman gain matrix. And use this gain to predict the prior state value By fusing the data with current actual measurement information, a more accurate posterior state estimate at the current time can be obtained. And update the posterior estimation error covariance matrix. Its core formula is:

[0093] Kalman gain calculation:

[0094] Status Update:

[0095] Covariance update:

[0096] in, The noise covariance matrix is ​​used to characterize the uncertainty of sensor measurements. It is the identity matrix; superscript This represents the matrix inversion operation. The formula contains... The term is often referred to as innovation, which reflects the difference between the actual measured value and the model prediction.

[0097] By recursively executing the above steps, the Kalman filter outputs the optimal full-state estimate. This provides accurate state information for the model predictive controller.

[0098] Step S5: Solving and outputting the adaptive model predictive control law;

[0099] When the parameter estimate After convergence, that is, when its rate of change is less than a certain set tolerance. The system seamlessly switches from PI control to adaptive model predictive control mode. During each sampling period... Perform the following sub-steps:

[0100] Sub-step S5.1, State and Model Injection;

[0101] The digital twin estimates the current state. and the updated system matrix , , , The data is then passed to the MPC model predictive controller. This data transfer process is completed within the same hardware platform, ensuring extremely low data transmission latency. This embodiment of the invention is preferably implemented in an edge computing device based on FPGA and ARM architecture to ensure real-time control.

[0102] Sub-step S5.2, Scrolling optimization;

[0103] The MPC controller solves an optimal control problem in the finite-time domain based on the injected latest model and state. Its objective function is designed as follows:

[0104]

[0105] in, To predict the time domain, To control the time domain , Based on the current model and state, the future... Prediction of step output, It is a reference trajectory for the future. It controls the increment. , The weights are semi-positive definite matrices, used to balance the tracking error and the rate of change of the control variable, respectively. By solving this quadratic programming problem, the future... The optimal control sequence is obtained by taking the optimal control sequence step by step. .

[0106] This step continues, generating a corresponding optimal control sequence for each input time step, ensuring that the time steps of the optimal control sequence advance sequentially, thus achieving the effect of rolling optimization.

[0107] Sub-step S5.3, Control Implementation;

[0108] For the rolling optimal control sequence generated in sub-step S5.3, for a specific time step, only the first element of its optimal control sequence is selected. The power amplifier device, which is the output of the model predictive controller, is applied to the magnetically driven rotor.

[0109] Step S6: Closed-loop iteration.

[0110] The system enters a continuous closed-loop iterative phase. Steps S3 to S5 are repeated in the next sampling cycle, forming a complete adaptive closed loop of "sensing-identification-update-prediction-control". This mechanism ensures that the control law can self-optimize online based on the system's real-time dynamic changes. Theoretically, the system can maintain high-precision, high-robust, and stable operation even in the face of internal parameter drift and external disturbances.

[0111] Example 2:

[0112] This embodiment uses a simulation platform to numerically verify the adaptive model predictive control method and compares it with the traditional fixed-parameter MPC control strategy. To more realistically simulate the time-varying characteristics of magnetically driven rotor parameters in actual engineering, this embodiment uses a composite function containing slow drift and periodic fluctuations to describe the changes in key parameters.

[0113] Simulation parameter settings:

[0114] Magnetic drive rotor model parameters (rated operating conditions): rotor mass Displacement stiffness coefficient Current stiffness coefficient Sampling period ;

[0115] To simulate the complex parameter drift caused by the combined effects of temperature rise and periodic thermal effects during actual operation, the following settings are made: and The parameters undergo a slow, nonlinear drift in the initial stage, and gradually superimposed with a progressive periodic fluctuation, which is closer to the real physical process.

[0116] Controller parameters:

[0117] Adaptive MPC: Prediction in the Time Domain Control time domain Weight matrix , ;

[0118] Traditional MPC: uses fixed parameters .

[0119] Simulation results and analysis:

[0120] (1) Parameter tracking performance verification:

[0121] The output of the drawing parameter identification module and The actual parameters described by the aforementioned complex time-varying function are the displacement stiffness coefficients. Current stiffness coefficient The comparison curves are as follows: Figure 3 and Figure 4 As shown. Simulation results show that:

[0122] The online recursive least squares (RLS) identification module of adaptive MPC can effectively track the nonlinear slow-varying trend and high-frequency fluctuation components of parameters. In the initial stage of system startup (within about 20 seconds), the identified parameters quickly converge to near the true parameters; throughout the simulation process, the identified values ​​fluctuate closely around the true values, demonstrating the algorithm's ability to accurately track complex dynamic characteristics, with a maximum steady-state relative error of <4%.

[0123] This result strongly demonstrates that the online identification mechanism can provide high-precision real-time parameters for the prediction model, which is key to the success of adaptive control.

[0124] (2) Comparison of control performance:

[0125] An instantaneous pulse disturbance with an amplitude of 0.15 mm is applied at t=100 seconds. The recovery curves of rotor displacement deviation under the two control strategies are compared, as follows: Figure 5 As shown:

[0126] Traditional fixed-parameter MPC: Due to a severe mismatch between the internal model and the real system, the controller generates incorrect predictions. This leads to large, low-frequency continuous oscillations in the rotor displacement (with a maximum overshoot of 0.12 mm), decreased system stability, a settling time exceeding 5 seconds, and the inability to completely eliminate steady-state errors.

[0127] Adaptive MPC: Thanks to the predictive model updating based on high-precision parameters identified in real time, the controller can accurately predict system dynamics. Displacement deviations are suppressed rapidly and smoothly, with a settling time of less than 1 second, no significant overshoot, and zero steady-state error, demonstrating excellent anti-interference capability and robustness.

[0128] This embodiment, through simulation using a more realistic model of complex parameter variations, powerfully demonstrates the significant advantages of the adaptive method in handling nonlinear and coupled time-varying characteristics. This method ensures that the prediction model remains highly synchronized with the physical entity, thereby achieving control accuracy and system stability far exceeding traditional methods across a wide range of operating conditions.

[0129] 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 an adaptive model predictive control method for magnetically driven rotors oriented towards twin control, 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.

[0130] 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.

[0131] This invention provides an idea and method for adaptive model predictive control of magnetically driven rotors oriented towards twin control. 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. An adaptive model predictive control method for magnetically driven rotors oriented towards twin control, characterized in that, The method includes the following steps: Step 1: Establish a parameterized discrete state-space prediction model for the magnetically driven rotor. This prediction model includes a set of key parameters. ; Step 2: Start the magnetic drive rotor system and use a traditional controller to ensure its stable operation; Step 3: Collect measured data of rotor displacement and coil current of the magnetically driven rotor system; based on the measured data, analyze the key parameter set. Continuous identification is performed to obtain real-time parameter estimates. ; Step 4, calculate the estimated real-time parameters. Update the current prediction model and use the updated prediction model to estimate the full state of the magnetically driven rotor; Step 5: Based on the updated prediction model and its estimated full state of the magnetically driven rotor, calculate the optimal control sequence and control the magnetically driven rotor accordingly. Step 6: Repeat steps 3 to 6 to form a control closed loop and complete the adaptive model predictive control of the magnetic drive rotor based on digital twin. The prediction model for the parameterized discrete state space described in step 1 includes a system matrix, which depends on the parameter set. ; The prediction model is expressed as follows: ； in, for The system state vector of the magnetically driven rotor at any given time. To control the input vector, The output vector of the prediction model. , , and For the parameter set The system matrix.

2. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 1, characterized in that, The conventional controller mentioned in step 2 is a proportional-integral controller.

3. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 2, characterized in that, The key parameter set described in step 3 Continuous identification is performed, that is, online parameter identification is performed using the recursive least squares method.

4. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 3, characterized in that, The step 4 involves estimating the full state of the magnetically driven rotor using the updated prediction model, which means estimating the full state variables of the magnetically driven rotor in real time using a state observer.

5. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 4, characterized in that, The state observer mentioned in step 4 uses a Kalman filter, which is used to estimate the full state variables of the magnetically driven rotor in real time.

6. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 5, characterized in that, Step 4 describes the real-time estimation of the full-state variables of the magnetically driven rotor using a Kalman filter. The Kalman filter is used in each sampling period... It contains two stages: a prediction step and an update step, used for time updates and measurement updates respectively. By recursively executing the prediction and update steps, the Kalman filter outputs the optimal full-state estimate. .

7. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 6, characterized in that, The prediction step and update step are as follows: Prediction Step: Based on the posterior state estimate from the previous time step and its estimated error covariance matrix Using the updated prediction model, predict the current Prior state estimate at time 1 and the prior estimation error covariance matrix ; Update steps: In obtaining Actual displacement measurement value at time 10:00 Then, the Kalman filter calculates the Kalman gain matrix. And use this gain matrix to predict the prior state values By fusing the data with the current actual displacement measurement information, the posterior state estimate at the current time can be obtained. And update the posterior estimation error covariance matrix. .

8. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 7, characterized in that, The step 5 involves calculating the optimal control sequence and controlling the magnetic drive rotor accordingly. This involves using a model predictive controller to solve the optimization problem in the finite time domain in each sampling period to obtain the optimal control sequence. The first control quantity of the optimal control sequence is then applied to the magnetic drive rotor to control it.

9. The adaptive model predictive control method for magnetically driven rotors oriented towards twin control according to claim 8, characterized in that, Step 5, which involves calculating the optimal control sequence and controlling the magnetically driven rotor accordingly, specifically includes the following steps: Step 5-1: Estimate the current state value and the updated system matrix , , , Passed to the model prediction controller; Step 5-2, Construct the optimization problem , means as follows: ； in, To predict the time domain, For time steps, To control the time domain, Based on the current Time-prediction model and state pair The prediction output at any given time. for Reference trajectory at any moment To control the increment, and For the square of the weighted Euclidean norm and This is the weight matrix; Step 5-3, the model predictive controller solves the optimization problem. To obtain the future The optimal control sequence for the step is represented as follows: ； in, Based on the current Time prediction model and state solution about The optimal control quantity at any given time; Step 5-4, set the first control variable of the optimal control sequence. It is applied to the magnetically driven rotor.

Images