An Adaptive Control Method and System for Permanent Magnet Synchronous Motors

By combining a twin feedback mechanism and an extended state observer, the core physical parameters of the permanent magnet synchronous motor are dynamically corrected, solving the problems of parameter drift and external load confusion, and improving the dynamic tracking accuracy and robustness of the motor.

CN123122348APending Publication Date: 2026-11-17DALIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611273864.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional control strategies are difficult to effectively cope with parameter drift and sudden changes in external load in permanent magnet synchronous motors, resulting in dynamic tracking errors and current harmonics, which affect the robustness and efficiency of the system.

Method used

The core physical parameters are dynamically and iteratively corrected using a twin feedback mechanism. Internal parameter drift is removed by an extended state observer, a high-fidelity state prediction model is constructed, and the optimal control command is solved.

Benefits of technology

It improves the dynamic tracking accuracy and system robustness of permanent magnet synchronous motors under parameter mismatch and sudden load, reduces the speed drop and dynamic adjustment time, and realizes forward-looking control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122348A_ABST
    Figure CN123122348A_ABST
Patent Text Reader

Abstract

This application discloses an adaptive control method and system for a permanent magnet synchronous motor (PMSM), relating to the field of motor servo control technology. The method includes: obtaining the tracking error between the actual output of the physical entity unit of the PMSM and the twin output of a digital twin model; iteratively updating the core physical parameters in the digital twin model based on a twin feedback mechanism and the tracking error; obtaining the lumped disturbance acting on the physical entity unit after parameter drift characteristics have been removed by an extended state observer, and treating the lumped disturbance as a pure external mechanical load; constructing a state prediction model using the updated core physical parameters and the pure external mechanical load; solving the state prediction model to obtain control commands for driving an inverter; the inverter is used to control the PMSM. This application can improve the dynamic tracking accuracy and system robustness of the PMSM under parameter mismatch and sudden load conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of motor servo control technology, and in particular to an adaptive control method and system for a permanent magnet synchronous motor. Background Technology

[0002] With the continuous evolution of high-end precision manufacturing and industrial automation technologies, the requirements for the dynamic response and steady-state accuracy of servo systems in Computer Numerical Control (CNC) equipment are becoming increasingly stringent. As the core power unit of CNC systems, permanent magnet synchronous motors (PMSMs) dominate due to their high power density and wide speed range. However, PMSMs are inherently complex, multivariable systems with strong coupling and nonlinearity. During high-speed, heavy-load operation, they are subject to the dual pressures of internal parameter perturbations such as permanent magnet flux decay due to temperature rise and inductance shift caused by magnetic circuit saturation, as well as external random cutting load mutations. Traditional control strategies struggle to guarantee the high-performance and robust operation of the system.

[0003] Model predictive control (MPC), as an advanced time-domain optimization control strategy, has become a strong candidate to replace traditional proportional-integral (PI) control due to its intuitive ability to handle multivariable hard constraints and its extremely fast dynamic feedforward response. However, the performance of MPC is highly dependent on the accuracy of its internal predictive model. When the internal mathematical model of the controller deviates from the parameters of the actual physical motor (i.e., model mismatch), the control variables calculated by MPC will deviate significantly from the global optimum, resulting in huge steady-state tracking errors, dynamic overshoot, and severe current harmonics.

[0004] To compensate for model mismatch, academia and industry commonly employ techniques such as Extended State Observers (ESOs) to construct robust MPCs. However, existing observer-based compensation strategies suffer from a fundamental flaw: traditional ESOs physically conflate the motor's "internal parameter drift" (such as changes in system input gain due to flux decay) with "external mechanical loads" (such as additive resistance from sudden external forces) into a single "black-box lumped disturbance." This approach, attempting to solve "multiplicative gain errors" with "additive error compensation," directly prevents the predictive model from revealing the motor's true physical state. During dynamic acceleration and deceleration, this undecoupled compensation not only fails to provide an accurate state evolution benchmark but also forces the controller to rely on simply and crudely increasing the current amplitude to resist disturbances, significantly reducing the control efficiency and command smoothness of the servo system. Summary of the Invention

[0005] The purpose of this application is to provide an adaptive control method and system for permanent magnet synchronous motors, which can improve the dynamic tracking accuracy and system robustness of permanent magnet synchronous motors under parameter mismatch and sudden load.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides an adaptive control method for a permanent magnet synchronous motor, comprising: To obtain the tracking error between the actual output of the physical unit of the permanent magnet synchronous motor and the twin output of the digital twin model; The core physical parameters in the digital twin model are updated based on the twin feedback mechanism and tracking error iterative update to obtain the updated core physical parameters; the core physical parameters are preset parameters. After the extended state observer removes the parameter drift characteristics, the lumped disturbance acting on the physical entity unit is obtained, and the lumped disturbance is treated as a pure external mechanical load. A state prediction model is constructed using the updated core physical parameters and the pure external mechanical load. Solving the state prediction model yields control commands for driving the inverter; the inverter is used to control the permanent magnet synchronous motor.

[0007] Optionally, the core physical parameter is the permanent magnet flux linkage; The core physical parameters in the digital twin model are updated based on the twin feedback mechanism and tracking error iteration to obtain the updated core physical parameters, which specifically include: Based on the tracking error, the gradient descent method is used to iteratively update the core physical parameters in the digital twin model.

[0008] Optionally, based on the tracking error, the core physical parameters in the digital twin model are iteratively updated using the gradient descent rule, including the following formula: ; in, for The estimated value of the permanent magnet flux linkage at time t. for The estimated value of the permanent magnet flux linkage at time t. This is the integral adjustment factor. for The quadrature-axis control current applied at all times, for Tracking error at any given moment.

[0009] Optionally, the integral adjustment coefficient satisfies the following boundary conditions: ; in, This represents the moment of inertia of the permanent magnet synchronous motor. This represents the number of pole pairs in a permanent magnet synchronous motor. For discrete control periods, This refers to the rated quadrature-axis current of the permanent magnet synchronous motor. This is the positive bias constant.

[0010] Optionally, the extended state observer obtains the state equation after removing parameter drift features as follows: ; in, The observed value of angular velocity. The first derivative of the observed angular velocity. Based on the updated core physics parameters, For quadrature axis current, The nominal damping coefficient of the system. This is the actual mechanical angular velocity. The observed values ​​are for external mechanical loads. For the first-order observer gain, The first derivative of the observed external mechanical load. This is the gain for the second-order observer.

[0011] Optionally, the state prediction model is expressed as: ; in, for The mechanical angular velocity state quantity predicted at any given time. It is a discrete state matrix and satisfies , The nominal damping coefficient of the system. For discrete control periods, for The control matrix includes parameter corrections at all times and satisfies , This represents the moment of inertia of the permanent magnet synchronous motor. This represents the number of pole pairs in a permanent magnet synchronous motor. for Estimates of core physical parameters at time 10:00. for The quadrature-axis control current applied at all times, for The aggregate disturbance at any given moment.

[0012] Optionally, the state prediction model is solved to obtain control commands for driving the inverter, specifically including: Construct a quadratic cost function that includes a tracking error penalty term in the prediction time domain and a control energy limit term; The state prediction model is substituted into the quadratic cost function, and the quadratic cost function is solved with the goal of minimizing it to obtain the control command.

[0013] Optionally, the quadratic cost function is expressed as: ; in, The value of the quadratic cost function. The set number of prediction time-domain steps, For the prediction step index, For the predicted first The tracking error of the step, Here is the error weight matrix. For predicting control sequences for the future, To control the energy weight matrix, For terminal prediction tracking error, This is the terminal state weight matrix.

[0014] Secondly, this application provides an adaptive control system for a permanent magnet synchronous motor, wherein the adaptive control system applies the aforementioned adaptive control method for a permanent magnet synchronous motor, and the adaptive control system includes: The tracking error acquisition unit is used to acquire the tracking error between the actual output of the physical entity unit of the permanent magnet synchronous motor and the twin output of the digital twin model. The twin feedback decoupling unit is used to iteratively update the core physical parameters in the digital twin model based on the twin feedback mechanism and tracking error to obtain the updated core physical parameters; the core physical parameters are preset parameters. An extended state observation unit is used to acquire the lumped disturbance acting on the physical entity unit after removing parameter drift characteristics; and to treat the lumped disturbance as a pure external mechanical load. The state prediction model reconstruction unit is used to construct a state prediction model using the updated core physical parameters and the pure external mechanical load. The control command solving unit is used to solve the state prediction model to obtain control commands for driving the inverter; the inverter is used to control the permanent magnet synchronous motor.

[0015] Optionally, the permanent magnet synchronous motor adaptive control system includes a dual-CPU hardware architecture composed of a host computer central controller and a slave computer microprocessor. The host computer central controller is used to deploy the tracking error acquisition unit and the twin feedback decoupling unit; The lower-level microprocessor is used to deploy the extended state observation unit, the state prediction model reconstruction unit, and the control command solving unit. The lower-level microprocessor uses a TMS320F28379D dual-core control chip; The host computer central controller and the slave computer microprocessor exchange data via a high-speed communication bus.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides an adaptive control method and system for permanent magnet synchronous motors. By introducing a twin feedback mechanism, the core physical parameters of the input are dynamically and iteratively corrected based on the tracking error between the actual output and the twin output. From a physical mechanism perspective, the internal parameter perturbation is separated from the observation signal of the extended state observer, allowing the extended state observer to focus on the observation of external additive loads. This can improve the dynamic tracking accuracy and system robustness of permanent magnet synchronous motors under parameter mismatch and sudden loads. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an adaptive control method for a permanent magnet synchronous motor according to an embodiment of this application.

[0019] Figure 2 This is a detailed flowchart illustrating an adaptive control method for a permanent magnet synchronous motor provided in one embodiment of this application.

[0020] Figure 3 This document presents a schematic diagram of the principle architecture and dual-CPU hardware collaborative topology of an adaptive control method for a permanent magnet synchronous motor, provided as an embodiment of this application.

[0021] Figure 4 This is a schematic diagram of the internal logic structure of the digital twin feedback correction function and parameter decoupling module provided in an embodiment of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] In recent years, digital twin (DT) technology has shown great potential in motor fault diagnosis and condition monitoring. However, how to truly extend the dynamic high-fidelity modeling capability of digital twins to millisecond-level closed-loop control, actively decouple internal parameters from external loads from a physical mechanism using twin feedback, and how to resolve the contradiction between massive twin computing and the limited computing power of the underlying hardware are key technical challenges that urgently need to be solved in the field of high-performance motor servo systems.

[0025] To address the problem that traditional robust model predictive control (MPC) confuses parameter drift with external loads as a single "lumped disturbance," leading to model prediction distortion and inefficient compensation, specifically the issue that the coupling of flux demagnetization and external load fluctuations during the operation of the physical motor causes mismatch in the z-state prediction model, thereby weakening the control effect, this application proposes an adaptive control method for permanent magnet synchronous motors (PMSMs). Specifically, it is an adaptive predictive control method for PMSMs based on twin feedback decoupling, introducing an adaptive digital twin closed-loop architecture to strictly separate the internal parameter mismatch (affecting control gain) from the external additive load in terms of mechanism.

[0026] In one exemplary embodiment, such as Figure 1 and Figure 2 As shown, this application provides an adaptive control method for a permanent magnet synchronous motor, including steps 101-105.

[0027] Step 101: Obtain the tracking error between the actual output of the physical unit of the permanent magnet synchronous motor and the twin output of the digital twin model.

[0028] Step 102: Iteratively update the core physical parameters in the digital twin model based on the twin feedback mechanism and tracking error to obtain the updated core physical parameters; the core physical parameters are preset parameters.

[0029] Step 103: Obtain the lumped disturbance acting on the physical entity unit after the parameter drift feature has been removed by the extended state observer, and treat the lumped disturbance as a pure external mechanical load.

[0030] Step 104: Construct a state prediction model using the updated core physical parameters and the pure external mechanical load.

[0031] Step 105: Solve the state prediction model to obtain control commands for driving the inverter; the inverter is used to control the permanent magnet synchronous motor.

[0032] This application proposes an adaptive digital twin closed-loop feedback decoupling architecture. By introducing a twin feedback mechanism, the input gain parameter is dynamically and iteratively corrected based on the tracking error between the actual output and the twin output. From a physical mechanism perspective, the internal parameter perturbation is orthogonally separated from the observation signal of the extended state observer, enabling the extended state observer to focus on the observation of external additive loads, thus reconstructing the full-band high-fidelity prediction model from the source.

[0033] In an exemplary embodiment, prior to step 101, the adaptive control method for a permanent magnet synchronous motor of this application further includes: collecting the actual operating state of the permanent magnet synchronous motor and constructing a closed-loop control architecture that includes physical entity units and adaptive digital twin units (digital twin models).

[0034] The tracking error is defined to satisfy: ; in, For discrete time step index, Let k be the tracking error at time k. For the first The actual mechanical angular velocity of the physical entity unit at each sampling time, i.e., the actual output. For the first The predicted mechanical angular velocity output by the adaptive digital twin unit at each sampling time, i.e., the twin output.

[0035] In one exemplary embodiment, the core physical parameter is the permanent magnet flux linkage.

[0036] Step 102 involves iteratively updating the core physical parameters in the digital twin model online based on the twin feedback mechanism and tracking error, obtaining the updated core physical parameters (input gain term). Specifically, this includes: using the gradient descent rule to iteratively update the core physical parameters in the digital twin model with the goal of minimizing the tracking error, and estimating the permanent magnet flux linkage. The formula is as follows: ; in, For the next moment ( The estimated value of the permanent magnet flux linkage at time ( ). for The estimated value of the permanent magnet flux linkage at time t. This is the integral adjustment factor. For the current time ( The quadrature-axis control current applied at (time) for Tracking error at any given moment.

[0037] More specifically, such as Figure 2 As shown, determine Is the absolute value greater than the set steady-state error dead zone threshold? If yes, then dynamic feedback correction is performed; otherwise, the estimated value of the permanent magnet flux linkage is kept unchanged to optimize the external load disturbance estimation.

[0038] In an exemplary embodiment, to ensure the global asymptotic stability of the parameter update process, the integral adjustment coefficients satisfy the following boundary conditions: ; in, This represents the moment of inertia of the permanent magnet synchronous motor. This represents the number of pole pairs in a permanent magnet synchronous motor. For discrete control periods, This refers to the rated quadrature-axis current of the permanent magnet synchronous motor. To prevent the smallest positive bias constant from being divided by zero.

[0039] In one exemplary embodiment, the lumped disturbance after parameter drift characteristics have been removed is obtained by an extended state observer and is used as a pure external mechanical load for feedforward compensation.

[0040] The extended state observer obtains the state equation after removing parameter drift features as follows: ; in, The observed value of angular velocity. The first derivative of the observed angular velocity. Based on the updated core physics parameters, For quadrature axis current, The nominal damping coefficient of the system. This is the actual mechanical angular velocity. The observed values ​​represent purely external mechanical loads. For the first-order observer gain, The first derivative of the observed external mechanical load. This is the gain for the second-order observer.

[0041] In an exemplary embodiment, the full-band high-fidelity discrete-state prediction model is reconstructed using the updated core physical parameters and the purely external mechanical load. Specifically, this includes: updating the permanent magnet flux linkage estimate... Substituting the discretized state-space equations, reconstructing the control matrix, and compensating for external mechanical loads, the state prediction model is expressed as: ; in, for The mechanical angular velocity state quantity predicted at any given time. It is a discrete state matrix and satisfies , The nominal damping coefficient of the system. For discrete control periods, for The control matrix includes parameter corrections at all times and satisfies , This represents the moment of inertia of the permanent magnet synchronous motor. This represents the number of pole pairs in a permanent magnet synchronous motor. for Estimates of core physical parameters at time 10:00. for The quadrature-axis control current applied at all times, for The aggregate disturbance at any given moment.

[0042] In an exemplary embodiment, the state prediction model is solved to obtain control commands for driving the inverter. Specifically, this involves solving a cost function within a model predictive control framework to obtain the optimal control command sequence and outputting it to the underlying inverter. This includes: constructing a quadratic cost function containing a tracking error penalty term and a control energy limitation term in the prediction time domain; substituting the state prediction model into the quadratic cost function; solving the quadratic cost function with the goal of minimizing the quadratic cost function to obtain the control commands; and outputting these control commands to the underlying inverter.

[0043] The quadratic cost function is expressed as: ; in, The value of the quadratic cost function. The set number of prediction time-domain steps, For the prediction step index, For the predicted first The tracking error of the step, Here is the error weight matrix. For predicting control sequences for the future, To control the energy weight matrix, For the final prediction tracking error, This is the state weight matrix.

[0044] In an exemplary embodiment, taking a typical high-frequency CNC servo platform as an example, this platform relies on a dual-CPU hardware architecture where a host computer CPU and a slave microprocessor work together. The host computer uses a high-performance industrial computer with embedded MATLAB software, and the slave computer uses a Texas Instruments TMS320F28379D microprocessor chip. The basic nominal parameters of the surface-mounted permanent magnet synchronous motor (SPMSM) controlled by this platform are: rated voltage... V, rated speed r / min, stator phase resistance Quadrature and direct axis synchronous inductors mH, initial nominal permanent magnet flux Wb, polar logarithm Moment of inertia Viscous friction coefficient Discrete control cycle of the underlying TMS320F28379D chip It is strictly set to 0.1 milliseconds (i.e., 10 kHz sampling rate).

[0045] The closed-loop execution logic of this adaptive predictive control method specifically includes the following steps: Step S1: Initialization of the discrete digital twin model. The main control dual-CPU system is based on the physical motor. Mechanism equations in a rotating coordinate system are used to initialize and construct the digital twin module. The forward time-domain steps for predictive control are then set. And the weight matrix in the quadratic cost function used to balance tracking accuracy and energy loss. Meanwhile, to prevent subsequent integration calculations from triggering hardware division by zero anomalies, a small positive bias constant is preset. and steady-state error dead zone threshold Assign initial flux linkage parameters to the host computer's digital twin. .

[0046] Step S2: Dynamic observation and parameter decoupling based on host computer twin feedback. After the system starts closed-loop operation, the lower-level microprocessor in each... Interruption time in milliseconds ( (Indexed by discrete time step) The actual mechanical angular velocity is acquired through sensors. and cross-axis stator current (i.e., the system's control input) This data is then uploaded to the host computer in real time via a high-speed bus. This step involves two main parallel mechanisms: First, the host computer performs twin feedback correction. For example... Figure 3 As shown, the same control signal Simultaneously applied to both the physical entity and the virtual digital twin, the tracking error between the actual physical angular velocity and the twin's predicted angular velocity is calculated using the following formula. .

[0047] When the system is in a dynamic acceleration / deceleration process (i.e.) When the error originates, the host computer uses the gradient descent algorithm to trace the source of the error and iteratively updates the core input gain parameter (permanent magnet flux linkage) in the twin model online. ). Calculate the precise It is then sent to the TMS320F28379D via the serial bus. This mechanism dynamically adjusts... This forces the decoupling of internal parameter perturbations from the total system disturbance.

[0048] Secondly, the lower-level machine executes the extended state observer. After the parameter perturbation is stripped away by the aforementioned feedback, the lower-level machine uses the negative feedback logic of the error system to perform high-frequency observation.

[0049] Step S3: Full-band high-fidelity prediction model reconstruction. This involves reconstructing the flux linkage parameters transmitted in real-time during the previous steps. Substitute the discrete state-space matrix of the motor back in. The model control matrix is ​​now updated as follows: ; And the purely external load observed by ESO Injected as an additive perturbation term. The final updated state prediction equation is: .

[0050] This reconstruction process eliminates model mismatch at its source, giving subsequent predictive control a very high degree of physical baseline realism.

[0051] Step S4: The lower-level microprocessor executes robust model predictive control. Based on the reconstructed high-fidelity predictive model, the system anticipates and calculates the future system state trajectory. A quadratic cost function is constructed, incorporating speed tracking error penalties and current energy penalties.

[0052] This step substitutes the state model into the cost function, transforming the problem into a standard convex quadratic programming optimization process. Due to the prediction model matrix... Having been precisely corrected by twin feedback, the optimized solver no longer relies on passively and blindly increasing the control current amplitude to resist disturbances; instead, relying on a pure external load feedforward term, the solver achieves optimal "current phase lead compensation" in the time domain. The lower-level TMS320F28379D chip utilizes its floating-point unit (FPU) to instantly solve for the optimal control sequence through matrix inversion, extracting the first element of the sequence and outputting it to the underlying space vector pulse width modulation (SVPWM) inverter unit for power drive, such as... Figure 4 As shown. The 0.1 millisecond cycle ends, and it waits for the next clock interrupt.

[0053] This application establishes a closed-loop twin feedback mechanism and a dual-CPU collaborative hardware architecture to actively remove the pollution of lumped disturbances caused by physical parameter perturbations, enabling model predictive control to perform optimal current phase advance planning based on high-fidelity physical images. This fundamentally solves the problems of high parameter sensitivity and underlying computing power bottleneck in traditional predictive control.

[0054] Based on the same inventive concept, this application also provides an adaptive control system for a permanent magnet synchronous motor to implement the aforementioned adaptive control method for a permanent magnet synchronous motor. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more embodiments of the adaptive control system for a permanent magnet synchronous motor provided below can be found in the limitations of the adaptive control method for a permanent magnet synchronous motor described above, and will not be repeated here.

[0055] In one exemplary embodiment, an adaptive control system for a permanent magnet synchronous motor is provided, comprising: The tracking error acquisition unit is used to construct a digital twin model and acquire the tracking error between the actual output of the physical entity unit of the permanent magnet synchronous motor and the twin output of the digital twin model. The twin feedback decoupling unit is used to iteratively update the core physical parameters in the digital twin model online based on the twin feedback mechanism, control input, and tracking error, so as to obtain the updated core physical parameters and achieve independent separation of parameter drift features; the core physical parameters are preset parameters. An extended state observation unit is used to acquire the lumped disturbance acting on the physical entity unit after removing parameter drift characteristics; and to treat the lumped disturbance as a pure external mechanical load. The state prediction model reconstruction unit is used to construct a state prediction model using the updated core physical parameters and the pure external mechanical load. The control command solving unit is used to solve the state prediction model (based on rolling optimization of the high-fidelity state prediction model) to obtain control commands for driving the inverter; the inverter is used to control the permanent magnet synchronous motor.

[0056] In one exemplary embodiment, the permanent magnet synchronous motor adaptive control system further includes a dual-CPU hardware architecture composed of a host computer central controller and a slave computer microprocessor, and the system relies on this dual-CPU hardware architecture.

[0057] The host computer central controller has an embedded operating environment for deploying the tracking error acquisition unit and the twin feedback decoupling unit, and for performing highly complex twin feedback iterative calculations and high-precision status monitoring.

[0058] The lower-level microprocessor is used to deploy the extended state observation unit, the state prediction model reconstruction unit, and the control command solving unit to realize high sampling rate low-level hard real-time drive and inverter control.

[0059] The lower-level microprocessor uses a TMS320F28379D dual-core control chip.

[0060] The host computer central controller and the slave computer microprocessor interact with each other via a high-speed communication bus to achieve microsecond-level data synchronization and real-time interaction of digital twin parameters.

[0061] This application ensures microsecond-level execution of extremely complex nonlinear decoupling algorithms and ultra-high sampling rates at the underlying level through a dual-CPU hardware architecture.

[0062] Compared with the prior art, this application has the following advantages: First, this application proposes a closed-loop physical layer perturbation decoupling mechanism based on twin feedback. Unlike traditional robust MPC methods that confuse internal and external system perturbations as lumped perturbations, this application's scheme uses a gradient descent algorithm to accurately identify and correct the input gain (magnetic flux linkage parameter) online, thus "purifying" the observation signal from the extended observer at its source. This not only allows the prediction model to highly reproduce the actual physical dynamic evolution but also enables the MPC solver to perform intelligent feedforward compensation through optimal "phase lead" when dealing with sudden external loads, completely avoiding the blind and inefficient current "amplitude surge" strategy of traditional algorithms. When parameters experience significant mismatches or sudden step loads, the method in this application significantly reduces the speed drop and shortens the dynamic adjustment time, achieving a balance between forward-looking control and extreme robustness.

[0063] Secondly, this application innovatively constructs a dual-CPU heterogeneous computing platform that coordinates the upper-level CPU (PC / MATLAB environment) and the lower-level microprocessor (TMS320F28379D). By moving the computationally intensive digital twin adaptive evolution algorithm to the upper-level computer and delegating the real-time-critical MPC rolling optimization and high-frequency sampling control to the lower-level computer, the contradiction between the extremely complex nonlinear decoupling algorithm and the computing power bottleneck of the underlying chip is perfectly resolved. Coupled with microsecond-level high-speed communication between the dual CPUs, this architecture not only ensures a high sampling rate for the underlying control loop but also endows the servo drive system with unprecedented intelligent analysis and high-order algorithm processing capabilities.

[0064] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0065] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An adaptive control method for a permanent magnet synchronous motor, characterized in that, include: To obtain the tracking error between the actual output of the physical unit of the permanent magnet synchronous motor and the twin output of the digital twin model; The core physical parameters in the digital twin model are updated based on the twin feedback mechanism and tracking error iterative update to obtain the updated core physical parameters; the core physical parameters are preset parameters. After the extended state observer removes the parameter drift characteristics, it obtains the lumped disturbance acting on the physical entity unit and treats the lumped disturbance as a pure external mechanical load. A state prediction model is constructed using the updated core physical parameters and the pure external mechanical load. Solving the state prediction model yields control commands for driving the inverter; the inverter is used to control the permanent magnet synchronous motor.

2. The adaptive control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The core physical parameter is the permanent magnet flux linkage; Based on the twin feedback mechanism and tracking error iterative update of the core physical parameters in the digital twin model, the updated core physical parameters are obtained, specifically including: Based on the tracking error, the gradient descent method is used to iteratively update the core physical parameters in the digital twin model.

3. The adaptive control method for a permanent magnet synchronous motor according to claim 2, characterized in that, Based on the tracking error, the core physical parameters in the digital twin model are iteratively updated using the gradient descent method, including the following formulas: ; in, for The estimated value of the permanent magnet flux linkage at time t. for The estimated value of the permanent magnet flux linkage at time t. This is the integral adjustment factor. for The quadrature-axis control current applied at all times, for Tracking error at any given moment.

4. The adaptive control method for a permanent magnet synchronous motor according to claim 3, characterized in that, The integral adjustment coefficient satisfies the following boundary conditions: ; in, This represents the moment of inertia of the permanent magnet synchronous motor. This represents the number of pole pairs in a permanent magnet synchronous motor. For discrete control periods, This refers to the rated quadrature-axis current of the permanent magnet synchronous motor. This is the positive bias constant.

5. The adaptive control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The extended state observer obtains the state equation after removing parameter drift features as follows: ; in, The observed value of angular velocity. The first derivative of the observed angular velocity. Based on the updated core physics parameters, For quadrature axis current, The nominal damping coefficient of the system. This is the actual mechanical angular velocity. The observed values ​​are for external mechanical loads. For the first-order observer gain, The first derivative of the observed external mechanical load. This is the gain for the second-order observer.

6. The adaptive control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The state prediction model is expressed as follows: ; in, for The mechanical angular velocity state quantity predicted at any given time. It is a discrete state matrix and satisfies , The nominal damping coefficient of the system. For discrete control periods, for The control matrix includes parameter corrections at all times and satisfies , This represents the moment of inertia of the permanent magnet synchronous motor. This represents the number of pole pairs in a permanent magnet synchronous motor. for Estimates of core physical parameters at time 10:

00. for The quadrature-axis control current applied at all times, for The aggregate disturbance at any given moment.

7. The adaptive control method for a permanent magnet synchronous motor according to claim 1, characterized in that, Solving the state prediction model yields control commands for driving the inverter, specifically including: Construct a quadratic cost function that includes a tracking error penalty term in the prediction time domain and a control energy limit term; The state prediction model is substituted into the quadratic cost function, and the quadratic cost function is solved with the goal of minimizing it to obtain the control command.

8. The adaptive control method for a permanent magnet synchronous motor according to claim 7, characterized in that, The quadratic cost function is expressed as: ; in, The value of the quadratic cost function. The set number of prediction time-domain steps, For the prediction step index, For the predicted first The tracking error of the step, Here is the error weight matrix. For predicting control sequences for the future, To control the energy weight matrix, For terminal prediction tracking error, This is the terminal state weight matrix.

9. An adaptive control system for a permanent magnet synchronous motor, characterized in that, The permanent magnet synchronous motor adaptive control system applies the permanent magnet synchronous motor adaptive control method according to any one of claims 1-8, and the permanent magnet synchronous motor adaptive control system includes: The tracking error acquisition unit is used to acquire the tracking error between the actual output of the physical entity unit of the permanent magnet synchronous motor and the twin output of the digital twin model. The twin feedback decoupling unit is used to iteratively update the core physical parameters in the digital twin model based on the twin feedback mechanism and tracking error to obtain the updated core physical parameters; the core physical parameters are preset parameters. An extended state observation unit is used to acquire the lumped disturbance acting on the physical entity unit after removing parameter drift characteristics; and to treat the lumped disturbance as a pure external mechanical load. The state prediction model reconstruction unit is used to construct a state prediction model using the updated core physical parameters and the pure external mechanical load. The control command solving unit is used to solve the state prediction model to obtain control commands for driving the inverter; the inverter is used to control the permanent magnet synchronous motor.

10. The adaptive control system for a permanent magnet synchronous motor according to claim 9, characterized in that, This includes a dual-CPU hardware architecture consisting of a host computer central controller and a slave computer microprocessor working together; The host computer central controller is used to deploy the tracking error acquisition unit and the twin feedback decoupling unit; The lower-level microprocessor is used to deploy the extended state observation unit, the state prediction model reconstruction unit, and the control command solving unit. The lower-level microprocessor uses a TMS320F28379D dual-core control chip; The host computer central controller and the slave computer microprocessor exchange data via a high-speed communication bus.