A permanent magnet synchronous motor predictive control method and device based on an adaptive digital twin model

Abstract

The application discloses a permanent magnet synchronous motor predictive control method and equipment based on an adaptive digital twin model, relates to the technical field of motor servo control and industrial informatization, and comprises the following steps: constructing a predictive control architecture of a permanent magnet synchronous motor; the predictive control architecture comprises a digital twin unit and a predictive control unit; the digital twin unit comprises a dynamics twin module and a twin model adaptive module; the dynamics twin module is constructed based on a mechanism model of the permanent magnet synchronous motor; the twin model adaptive module is embedded with an extended state observer; a cost function is constructed based on the predictive control architecture; and the permanent magnet synchronous motor is subjected to predictive control by using a closed-loop rolling mechanism based on the predictive control architecture and the cost function. By constructing the dynamics twin module and the twin model adaptive module, the application can adapt to time-varying system parameters and random load disturbances, and improve control performance.

Description

Technical Field

[0001] This application relates to the fields of motor servo control and industrial information technology, and in particular to a predictive control method and device for permanent magnet synchronous motors based on an adaptive digital twin model. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs), with their high power density, high efficiency, and excellent speed regulation performance, have become core power components for high-end industrial equipment and new energy vehicles. However, PMSMs face two major challenges in actual operation: first, the internal parameters of the motor (such as resistance, inductance, and flux linkage) can slowly change over time due to factors such as temperature rise and magnetic saturation; second, the external load torque often exhibits randomness and abrupt changes. These uncertainties make it difficult to establish an accurate mathematical model for the motor, posing a fundamental challenge to high-performance control.

[0003] Predictive control, as an advanced control strategy, predicts the system behavior over a future period based on a system model at the current moment, and solves for the optimal control sequence with the goal of optimizing a certain performance index. Only the first control input is executed, and this process is repeated in the next sampling period. This method can explicitly handle constraints, has good tracking performance and robustness, and has been applied in PMSM control. However, the performance of traditional predictive control is highly dependent on the accuracy of the predictive model. When the model parameters do not match the actual physical system, the prediction accuracy decreases, leading to deterioration of control performance or even instability. To address the model mismatch problem, existing technologies mainly employ two types of methods: one is adaptive predictive control based on online identification, which approximates the real system by updating model parameters in real time; the other is robust predictive control based on disturbance observation, such as active disturbance rejection control, which estimates and compensates for model errors as total disturbances. However, the former has a contradiction between its convergence speed under rapid time-varying disturbances and the real-time control requirements; the latter, although fast in response, relies excessively on high-gain observers, which may amplify measurement noise, and lacks in-depth understanding and visualization capabilities of the system's internal state. Digital twin (DT), as a key enabling technology for realizing the convergence of cyber-physical systems, provides a novel platform for the monitoring, prediction, and optimization of complex systems by constructing high-fidelity virtual models of physical entities and achieving real-time data interaction and synchronization between the two. In recent years, DT technology has begun to be applied to the condition monitoring, parameter estimation, and fault diagnosis of PMSMs (Programmable Module Smart Machines). For example, some studies have used DT models of PMSMs to monitor motor conditions in real time or diagnose early faults. However, most existing research treats DT as a monitoring and diagnostic tool, failing to fully leverage its potential in real-time closed-loop control, particularly in combination with model-driven predictive control. How to deeply integrate DT's dynamic high-fidelity modeling capabilities with the rolling optimization mechanism of predictive control to construct a novel predictive control architecture that can adapt to model changes and penetrate the "black box" of the physical system remains an unsolved technical challenge in this field. Summary of the Invention

[0004] The purpose of this application is to provide a predictive control method and device for permanent magnet synchronous motors based on an adaptive digital twin model, which can adapt to time-varying system parameters and random load disturbances, thereby improving control performance.

[0005] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model. The method is applied to a physical entity unit, which includes: a permanent magnet synchronous motor, a power drive inverter, current and speed / position sensors, and an underlying controller. The method includes: A predictive control architecture for a permanent magnet synchronous motor is constructed. The predictive control architecture includes a digital twin unit and a predictive control unit. The digital twin unit includes a dynamic twin module and a twin model adaptive module. The dynamic twin module is constructed based on the mechanistic model of the permanent magnet synchronous motor. The twin model adaptive module embeds an extended state observer. Based on the aforementioned predictive control architecture, a cost function is constructed; Based on the aforementioned predictive control architecture and cost function, a closed-loop rolling mechanism is used to perform predictive control on the permanent magnet synchronous motor.

[0006] Optionally, the state variables of the dynamic twin module include: the stator current component corresponding to the d-axis. i d Stator current components corresponding to the q-axis i q and real-time mechanical angular velocity; The input variables of the dynamic twin module include: the stator voltage component corresponding to the d-axis. u d Stator voltage components corresponding to the q-axis u q and load torque ; The output variable of the dynamic twin module is the predicted mechanical angular velocity.

[0007] Optionally, the dynamic twin module is: ; ; ; In the formula, R This is the phase resistance of the motor; L d and L q These are the inductances on the d-axis and q-axis, respectively; p n It is the extreme logarithm; ω m It is the mechanical angular velocity; ψ f For permanent magnets; J y is the moment of inertia; y is the output of the dynamic twin module. T e The electromagnetic torque of the PMSM; T m This refers to the mechanical load torque. B is the coefficient of friction.

[0008] Optionally, the extended state observer is an 2nd-order extended state observer; The state equation of the extended state observer is: ; In the formula, x 1 and x 2 are both inputs to the extended state observer, where, x The value 1 represents the system state. x The value of 2 represents the total system disturbance. z 1 represents the observed value of the system state; z 2 represents the observed value of the total system disturbance; β 1 and β Both 2 represent the extended state observer gain; b 0 represents the disturbance coefficient of the parameter term. The initial estimate, ; c 0 represents the load term disturbance factor. The initial estimate, ; e 1 represents the observation error. .

[0009] Optionally, based on the predictive control architecture and the cost function, a closed-loop rolling mechanism is used to perform predictive control of the permanent magnet synchronous motor, specifically including: Both the virtual control signal and the disturbance compensation signal are input into the dynamic twin module to obtain the state prediction sequence; the virtual control signal is the output of the prediction control unit; the disturbance compensation signal is the output of the twin model adaptive module; the state prediction sequence includes state predictions at multiple sampling times within the prediction time domain; Based on the state prediction sequence, the cost function is transformed into a quadratic programming problem; the optimization variable of the quadratic programming problem is the future control sequence. The quadratic programming problem is solved using the effective set method to obtain the optimal predictive control sequence; the optimal predictive control sequence includes control variables at multiple sampling times within the prediction time domain. The first element of the optimal predictive control sequence is applied to the physical entity unit; Upon reaching the next sampling time, return to the step "Input both the virtual control signal and the disturbance compensation signal into the dynamic twin module to obtain the state prediction sequence".

[0010] Optionally, the cost function is: ; In the formula, Here, represents the cost function; N represents the prediction time domain; i represents the sampling time number in the prediction time domain; and k represents the current sampling time number. e(k+i|k) For prediction iControl error after the cycle; u(k+i|k) For prediction i Control quantities after the cycle; u(k+ i|k) T For prediction i Transpose of the control quantity after the cycle; x(k+i|k) For prediction i The system state after the cycle; x(k+i|k) T For prediction i Transpose of the system state after the cycle; , and All are positive definite weight matrices.

[0011] Optionally, the quadratic programming problem is: ; Let the coefficient matrix of the quadratic term The coefficient vector of the first-order term The quadratic programming problem can be simplified to: ; In the formula, For predicting the transpose of the control error sequence; To predict the control error sequence; The positive definite weight matrix The conjugate; For the optimal control sequence U k Transpose of; The positive definite weight matrix The conjugate; For reference trajectory; C , M and L average The matrix is ​​constructed from the system model and the prediction time domain; W Perturbation prediction sequences provided for the extended state observer; This is the system's current return value; For matrix The transpose of .

[0012] In a second aspect, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model as described above.

[0013] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the predictive control method for permanent magnet synchronous motor based on an adaptive digital twin model as described above.

[0014] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model as described above.

[0015] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a predictive control method and device for permanent magnet synchronous motors (PMSMs) based on an adaptive digital twin model. The method is applied to a physical entity unit and includes: constructing a predictive control architecture for the PMSM; and performing predictive control of the PMSM using a closed-loop rolling mechanism based on the predictive control architecture and a cost function. Addressing the technical bottleneck of traditional predictive control methods for PMSMs, where the predictive model is fixed and difficult to adapt to time-varying system parameters and random load disturbances, leading to decreased control performance, this invention proposes a novel adaptive predictive control architecture. The digital twin unit constructs an adaptive twin model including a dynamic twin module and an adaptive twin model module. An extended state observer is used to observe and compensate for the total system disturbance in real time, ensuring the high fidelity of the virtual model. The predictive control unit performs state sequence prediction based on this high-fidelity model and generates the optimal control signal by solving a quadratic programming problem. This invention achieves online adaptive correction of the predictive model through digital twin technology, fundamentally solving the model mismatch problem caused by parameter perturbations and load changes, and significantly improving the control accuracy, anti-interference capability, and system robustness of the PMSM under complex operating conditions. Establish a self-calibrating high-fidelity predictive model: By constructing a digital twin of the permanent magnet synchronous motor control system and integrating an extended state observer, the "overall disturbance" constituted by the drift of internal motor parameters (such as changes in resistance, inductance, and flux linkage) and external uncertain loads is observed in real time. The estimated value of this disturbance is dynamically fed forward to compensate the twin model, thereby achieving online adaptation and real-time fidelity of the predictive model, fundamentally solving the model mismatch problem. Achieve high-performance predictive control based on the high-fidelity model: Utilize the accurate state prediction sequence provided by the above-mentioned adaptive twin model to drive the predictive controller to perform rolling optimization. By transforming model uncertainty into known and compensable disturbances, the high accuracy of the future dynamic predictions on which the optimization calculation is based is ensured. Thus, even under large-scale parameter perturbations and strong random disturbances, the theoretically optimal control signal can still be output, significantly improving the system's tracking accuracy, dynamic response speed, and anti-interference robustness. Construct a transparent and interactive new paradigm for control systems: The entire physical control system (including the motor body, driver, and controller) is completely mapped into the digital space, forming a "digital accompaniment." This not only makes all internal states, control logic, and predicted trajectories completely transparent to the operator, greatly enhancing the interpretability and debuggability of the system, but also provides a unified digital foundation and rich data interfaces for advanced intelligent services such as remote monitoring, online parameter tuning, control strategy simulation verification, and predictive maintenance. Attached Figure Description

[0016] 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.

[0017] Figure 1 This is a flowchart of a predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model, according to one embodiment of this application. Figure 2 This is a schematic diagram of a predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model, according to one embodiment of this application. Figure 3 This is a schematic diagram of a predictive control architecture in one embodiment of this application; Figure 4 This is a diagram of a digital twin unit in one embodiment of this application; Figure 5 This is a schematic diagram of the predictive control principle of a permanent magnet synchronous motor based on an adaptive digital twin model in one embodiment of this application. Detailed Implementation

[0018] 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.

[0019] 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.

[0020] In one exemplary embodiment, a predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model is provided, the method being applied to a physical entity unit. For example... Figure 3 The physical units include: permanent magnet synchronous motor, power drive inverter, current and speed / position sensors, and underlying controller.

[0021] like Figures 1-2 , Figures 4-5 The predictive control method for permanent magnet synchronous motors based on adaptive digital twin models includes: Step 101: Construct the predictive control architecture for the permanent magnet synchronous motor. The predictive control architecture includes a digital twin unit and a predictive control unit. The digital twin unit includes a dynamic twin module and a twin model adaptive module. The dynamic twin module is constructed based on the mechanistic model of the permanent magnet synchronous motor. The twin model adaptive module embeds an extended state observer.

[0022] The state variables of the dynamic twin module include: the stator current component corresponding to the d-axis. i d Stator current components corresponding to the q-axis i q And real-time mechanical angular velocity. The input variables of the dynamic twin module include: the stator voltage component corresponding to the d-axis. u d Stator voltage components corresponding to the q-axis u q and load torque The output variable of the dynamic twin module is the predicted mechanical angular velocity.

[0023] The architecture involves tight coupling with physical entity units at both the physical and informational levels.

[0024] The physical entities, as controlled objects, include permanent magnet synchronous motors, power drive inverters, current and speed / position sensors, and low-level controllers (usually PI controllers) responsible for fast current tracking.

[0025] Digital twin unit: Serving as the system's "virtual brain" and "digital mirror," it synchronizes data with the physical entity unit via a high-speed, real-time communication link. Its core is the construction and operation of an adaptive twin model, which consists of the following two functional modules: Dynamic Twin Module: Based on the mechanistic model of the permanent magnet synchronous motor (e.g., the state-space model composed of voltage and motion equations in the dq rotating coordinate system), this module accurately reproduces the dynamic characteristics of the motor in digital space. Its core function is to quickly simulate and calculate the predicted sequence of motor states over multiple future sampling periods, given a set of control input sequences. .

[0026] ; ; ; In the formula, R This is the phase resistance of the motor; L d and L q These are the inductances on the d-axis and q-axis, respectively; p nIt is the extreme logarithm; ω m It is the mechanical angular velocity; ψ f For permanent magnets; J y is the moment of inertia; y is the output of the dynamic twin module. T e The electromagnetic torque of the PMSM; T m This refers to the mechanical load torque. B is the coefficient of friction.

[0027] The twin model adaptive module is key to achieving model self-calibration. Its core embeds an extended state observer (ESO). This ESO will handle the total system disturbance (including parameter perturbations). and load disturbance The system is treated as an extended state variable for real-time online observation. Its key technical points are: the ESO design is based on reducing or reconstructing the original system model (e.g., a simplified model for the velocity loop, and pole placement using Lyapunov stability theory to ensure sufficient bandwidth and extremely fast convergence speed). This module continuously receives output feedback from the physical entity and outputs a real-time estimate of the total perturbation. This estimate is then fed forward to compensate the input or state equation of the dynamic twin module, thereby dynamically correcting model errors and ensuring that the virtual model remains synchronized with the physical entity at all times with high fidelity.

[0028] Predictive Control Unit: Acting as the decision-making center, it is deeply integrated with the digital twin unit. In each control cycle k, this unit performs the following operations: Obtain a high-fidelity state prediction sequence for the next N steps (prediction time domain) from the digital twin unit, based on the current state and an adaptive model. .

[0029] Construct a standard quadratic cost function J that explicitly weighs the cumulative penalty of future tracking errors against the cost of changes in control parameters.

[0030] The predictive control problem is transformed into a problem based on future control sequences. This is a quadratic programming problem for optimizing variables. Because the model is accurate and the cost function is convex, this problem can be solved efficiently.

[0031] Solve this optimization problem to obtain the optimal control sequence. It outputs only its first element u(k | k) as the actual control quantity at the current moment (such as the q-axis current setpoint) to the physical entity unit.

[0032] Further optimizations can be made by organizing the digital twin unit according to the five-dimensional model architecture of digital twins. The adaptive twin model is the "virtual entity". On this basis, a "service" layer is developed that includes functions such as state visualization, historical data tracing, interactive tuning of control parameters, and simulation of multi-scenario control strategies, thereby forming a fully functional predictive control digital twin platform.

[0033] Step 102: Construct the cost function based on the predictive control architecture.

[0034] Step 103: Based on the predictive control architecture and cost function, predictive control of the permanent magnet synchronous motor is performed using a closed-loop rolling mechanism, including: inputting both the virtual control signal and the disturbance compensation signal into the dynamic twin module to obtain the state prediction sequence. The virtual control signal is the output of the predictive control unit. The disturbance compensation signal is the output of the twin model adaptive module. The state prediction sequence includes the state prediction quantities at multiple sampling times within the prediction time domain. Based on the state prediction sequence, the cost function is transformed into a quadratic programming problem. The optimization variable of the quadratic programming problem is the future control sequence. The effective set method is used to solve the quadratic programming problem to obtain the optimal predictive control sequence. The optimal predictive control sequence includes the control quantities at multiple sampling times within the prediction time domain. The first element of the optimal predictive control sequence is applied to the physical entity unit. When the next sampling time is reached, return to step "input both the virtual control signal and the disturbance compensation signal into the dynamic twin module to obtain the state prediction sequence".

[0035] This method is applied to the above architecture and is characterized by including the following closed-loop execution steps: S1: System modeling and digital twin initialization.

[0036] A detailed mechanistic model of the permanent magnet synchronous motor is established as the foundation. In the digital twin unit, an adaptive twin model is initialized, which includes a dynamic twin module and an adaptive twin model module integrating ESO, and bidirectional communication with the physical entity unit is established.

[0037] S2: Real-time disturbance observation and online model compensation (adaptive process).

[0038] During system operation, the ESO continues to run. It receives the actual output measurement value y(k) of the physical entity and, based on the built-in observer model, estimates the total disturbance currently acting on the system in real time. This estimate is immediately used to correct the prediction equations of the dynamic twin module, for example, by using it as a known additional input or by performing online corrections to the model parameters, thereby completing the instantaneous adaptive update of the twin model.

[0039] S3: Multi-step state prediction based on adaptive models.

[0040] Using the highly fidelity twin model after compensation in step S2, forward simulation is performed based on the current true state x(k) of the system and a set of future control input trial sequences U(k) to be optimized, to obtain the state prediction sequence for the next N sampling points. This forecast already includes the future impacts of the observed disturbances.

[0041] S4: Rolling optimization to find the optimal control sequence.

[0042] The high-precision prediction sequence obtained in step S3 Compared with the expected reference trajectory By comparison, a quadratic programming optimization problem as described in Scheme 1 is constructed. This problem is then solved online using an efficient numerical optimization algorithm to obtain the cost function. The smallest optimal control sequence Uk.

[0043] S5: Optimal control implementation and cycle rolling.

[0044] The first control variable u(k | k) of the optimal sequence is applied to the physical entity unit. Waiting for the next sampling time, the system state is updated, and then immediately returning to step S2, starting a new round of the "observation-compensation-prediction-optimization" cycle with the latest state and the latest disturbance estimate. This closed-loop rolling mechanism ensures the control system's continuous suppression of time-varying disturbances and model uncertainties.

[0045] Compared with existing permanent magnet synchronous motor control schemes, the technical solution proposed in this application brings the following significant and synergistic benefits: This invention fundamentally overcomes the model mismatch problem: the performance bottleneck of traditional predictive control lies in model inaccuracy. It creatively introduces an ESO (Earthquake Detection Unit) as a "model error detection unit" to quantify and compensate for total disturbances in real time, making the digital twin model a dynamically adaptive "accompaniment model." This ensures the accuracy of predictive information from the source, allowing subsequent optimization calculations to be based on reliability and completely reversing the passive situation where control performance deteriorates due to model errors.

[0046] This achieves a unified leap in dynamic performance and robustness: on the one hand, the extremely fast convergence characteristic of ESO (millisecond-level) ensures the real-time adaptability of the model, enabling the system to respond quickly to sudden disturbances; on the other hand, predictive control based on a high-fidelity model has look-ahead optimization capabilities, effectively handling dynamic constraints and smoothing control actions. The combination of these two aspects allows the system to exhibit extremely fast recovery speed, minimal tracking error, and near-overshoot-free stability when dealing with slowly time-varying parameters and strong load disturbances, resulting in a qualitative improvement in both dynamic performance and steady-state robustness.

[0047] This invention opens a new path to transparency and intelligence in control systems: it is not only an improvement in control algorithms but also a revolution in system architecture. The introduction of a digital twin transforms the traditional "signal black box" into a fully transparent, digital mirror image of the entire process and state. Engineers can intuitively observe the real-time changes of each variable in the virtual model, the deviation between predicted trajectories and actual values, and the magnitude and source analysis of disturbances, greatly reducing the difficulty of debugging and maintaining complex systems. Simultaneously, this architecture naturally integrates massive amounts of operational data and simulation capabilities, providing an ideal platform and abundant possibilities for importing machine learning algorithms for parameter self-tuning and realizing advanced intelligent applications such as model-based fault prediction and health management (PHM).

[0048] This invention offers high performance and low-cost engineering feasibility: The core algorithms (ESO, quadratic programming) of the method described in this invention have clear mathematical forms and moderate computational complexity, making them easy to implement on existing industrial controllers (such as high-performance DSPs and industrial PCs). By employing digital twin technology, many advanced and complex algorithms that are difficult to implement in real time on physical controllers (such as fine simulation and data mining) can be deployed on host computers or edge servers and implemented through collaborative computing. This improves system performance while ensuring engineering feasibility and cost control.

[0049] The system architecture used in this implementation is as follows: Figure 2 As shown, the system includes a physical entity unit, a digital twin unit, and a predictive control unit. The physical entity unit uses a Texas Instruments TMS320F28379D dual-core DSP as the main controller to control a surface-mount permanent magnet synchronous motor with a rated power of 64W, a rated voltage of 24V, and a rated speed of 3000rpm. Current and position signals are acquired through a Hall effect current sensor and a 23-bit absolute encoder, respectively, with a control cycle set to 0.1 milliseconds. The digital twin unit runs on an industrial computer equipped with an Intel Core i7 processor and communicates synchronously with the physical entity unit via EtherCAT industrial Ethernet, also with a synchronization cycle of 0.1 milliseconds.

[0050] The construction of the adaptive twin model is the core of this invention. First, a precise mathematical model is established based on the physical mechanism of the permanent magnet synchronous motor. In the two-phase rotating dq coordinate system, the motor's voltage equation, electromagnetic torque equation, and mechanical motion equation together form the basis of the dynamic twin module. Specifically, the d-axis current is selected... i d q-axis current i q and mechanical angular velocity ω m As a state variable, the d-axis voltage u d q-axis voltageu q and load torque Using angular velocity as the input variable and angular velocity as the output variable, the continuous state-space equations are discretized and implemented as a real-time simulation module. This module receives virtual control signals from the predictive control unit and disturbance compensation signals from the adaptive module, and can quickly calculate the state prediction sequence for multiple future control cycles.

[0051] The implementation of the twin model adaptive module is the key innovation of this invention. For the speed control loop, the complete model is reasonably simplified to obtain a first-order system with speed as the state, and all unmodeled dynamics, parameter perturbations, and external loads are uniformly considered as the total disturbance. w Based on this model, a second-order extended state observer is designed. The state equation of this observer is designed as follows: The state equation of the extended state observer is: ; In the formula, x 1 and x 2 are both inputs to the extended state observer, where, x The value 1 represents the system state. x The value of 2 represents the total system disturbance. z 1 represents the observed value of the system state; z 2 represents the observed value of the total system disturbance; β 1 and β Both 2 represent the extended state observer gain; b 0 represents the disturbance coefficient of the parameter term. The initial estimate, ; c 0 represents the load term disturbance factor. The initial estimate, ; e 1 represents the observation error. .

[0052] To ensure the rapid convergence and stability of the observer, pole placement is performed based on Lyapunov stability theory. This is achieved by placing the eigenvalues ​​of the observer error system at sufficiently far positions on the negative real axis (e.g., ...). ω 0 = 10000 rad / s), and let β 1=2 ω 0, β 2= ω 0 2 This ensures that the observation error converges to a bounded small neighborhood within an extremely short time (less than 0.5 milliseconds). The observer receives the actual rotational speed signal of the physical entity in real time and outputs an estimate of the total disturbance. This value is then used as a feedforward compensation amount and added to the state update calculation of the dynamic twin module in real time, thereby dynamically correcting the model and achieving adaptability.

[0053] The predictive control unit is implemented following the rolling optimization framework of model predictive control. In each control cycle, a high-fidelity state prediction sequence based on the current state and the latest disturbance compensation is first obtained from the digital twin unit. The cost function is designed in standard quadratic form, comprehensively considering both tracking error and the rate of change of the control variable. The cost function is: ; In the formula, Here, represents the cost function; N represents the prediction time domain; i represents the sampling time number in the prediction time domain; and k represents the current sampling time number. e(k+i|k) For prediction i Control error after the cycle; u(k+i|k) For prediction i Control quantities after the cycle; u(k+ i|k) T For prediction i Transpose of the control quantity after the cycle; x(k+i|k) For prediction i The system state after the cycle; x(k+i|k) T For prediction i Transpose of the system state after the cycle; , and All are positive definite weight matrices. N For prediction in the time domain, this embodiment takes N =5. By substituting into the prediction model, the optimization problem is transformed into one with future control sequences. U k A convex quadratic programming problem with decision variables: .

[0054] Let the coefficient matrix of the quadratic term The coefficient vector of the first-order term The quadratic programming problem can be simplified to: ; In the formula, For predicting the transpose of the control error sequence; To predict the control error sequence; The positive definite weight matrix The conjugate; For the optimal control sequence U k Transpose of; The positive definite weight matrix The conjugate; For reference trajectory; C , M and L average The matrix is ​​constructed from the system model and the prediction time domain; W Perturbation prediction sequences provided for the extended state observer; This is the system's current return value; For matrix The transpose of .

[0055] Because the model is accurate and H The problem is positive definite, and a unique optimal solution exists. An online solution using the effective set method is employed, with computation time on the industrial computer controlled to within 0.05 milliseconds. The optimal control sequence is obtained from the solution. U k Then, only the first element u ( k | k The current value of the q-axis current at the current moment is sent to the current loop controller of the physical entity unit.

[0056] The control system operates in a strictly periodic closed loop. After power-on initialization, the system enters the main loop. Within each 0.1-millisecond synchronous interrupt service routine, the following steps are executed sequentially: reading sensor data to obtain the current physical state; running the extended state observer algorithm to estimate the current total disturbance; compensating the disturbance estimate to the digital twin model and performing multi-step prediction based on the latest model and the current state; the predictive control unit solving a quadratic programming problem to obtain the optimal q-axis current setpoint; and executing SVPWM modulation to drive the inverter to output the corresponding voltage vector. After the interrupt service routine completes, it waits for the next synchronous cycle interrupt, repeating this cycle to achieve continuous rolling optimization control.

[0057] The system constructed using the above-described specific implementation methods enables the extended state observer to accurately estimate the combined disturbances caused by changes in motor parameters and external loads. For example, when the moment of inertia varies within 20% to 200% of its rated value, or when the load experiences a step or sinusoidal disturbance with an amplitude of 1 N·m, the observer can achieve accurate tracking within 0.5 milliseconds, with a steady-state observation error of less than 5%. Based on this high-fidelity twin model predictive controller, in speed step response, the settling time is reduced by approximately 40% compared to traditional PI controllers and fixed-model predictive controllers, with almost no overshoot. When dealing with sudden load changes, the dynamic speed drop is reduced by more than 90% and recovers rapidly, significantly improving the system's dynamic tracking performance and anti-interference robustness. The entire system not only achieves high-performance control, but its digital twin architecture also provides a complete virtual mirror, making all state variables and control processes visible, laying the foundation for system debugging, analysis, and the development of advanced intelligent applications.

[0058] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection.

[0059] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0060] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0061] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0062] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0063] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0064] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0065] 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.

[0066] 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. A predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model, characterized in that, The method is applied to a physical entity unit; The physical entity units include: a permanent magnet synchronous motor, a power drive inverter, current and speed / position sensors, and an underlying controller; The method includes: A predictive control architecture for a permanent magnet synchronous motor is constructed. The predictive control architecture includes a digital twin unit and a predictive control unit. The digital twin unit includes a dynamic twin module and a twin model adaptive module. The dynamic twin module is constructed based on the mechanistic model of the permanent magnet synchronous motor. The twin model adaptive module embeds an extended state observer. Based on the aforementioned predictive control architecture, a cost function is constructed; Based on the aforementioned predictive control architecture and cost function, a closed-loop rolling mechanism is used to perform predictive control on the permanent magnet synchronous motor.

2. The predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model according to claim 1, characterized in that, The state variables of the dynamic twin module include: the stator current component corresponding to the d-axis. i d Stator current components corresponding to the q-axis i q and real-time mechanical angular velocity; The input variables of the dynamic twin module include: the stator voltage component corresponding to the d-axis. u d Stator voltage components corresponding to the q-axis u q and load torque ; The output variable of the dynamic twin module is the predicted mechanical angular velocity.

3. The predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model according to claim 2, characterized in that, The dynamic twin module is: ； ； ； In the formula, R This is the phase resistance of the motor; L d and L q These are the inductances on the d-axis and q-axis, respectively; p n It is the extreme logarithm; ω m It is the mechanical angular velocity; ψ f For permanent magnets; J y is the moment of inertia; y is the output of the dynamic twin module. T e The electromagnetic torque of the PMSM; T m This refers to the mechanical load torque. B is the coefficient of friction.

4. The predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model according to claim 3, characterized in that, The extended state observer is an ii-order extended state observer; The state equation of the extended state observer is: ； In the formula, x 1 and x 2 are both inputs to the extended state observer, where, x The value 1 represents the system state. x The value of 2 represents the total system disturbance. z 1 represents the observed value of the system state; z 2 represents the observed value of the total system disturbance; β 1 and β Both 2 represent the extended state observer gain; b 0 represents the disturbance coefficient of the parameter term. The initial estimate, ; c 0 represents the load term disturbance factor. The initial estimate, ; e 1 represents the observation error. .

5. The predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model according to claim 4, characterized in that, Based on the aforementioned predictive control architecture and cost function, predictive control of the permanent magnet synchronous motor is performed using a closed-loop rolling mechanism, specifically including: Both the virtual control signal and the disturbance compensation signal are input into the dynamic twin module to obtain the state prediction sequence; the virtual control signal is the output of the prediction control unit; the disturbance compensation signal is the output of the twin model adaptive module; the state prediction sequence includes state predictions at multiple sampling times within the prediction time domain; Based on the state prediction sequence, the cost function is transformed into a quadratic programming problem; the optimization variable of the quadratic programming problem is the future control sequence. The quadratic programming problem is solved using the effective set method to obtain the optimal predictive control sequence; the optimal predictive control sequence includes control variables at multiple sampling times within the prediction time domain. The first element of the optimal predictive control sequence is applied to the physical entity unit; Upon reaching the next sampling time, return to the step "Input both the virtual control signal and the disturbance compensation signal into the dynamic twin module to obtain the state prediction sequence".

6. The predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model according to claim 5, characterized in that, The cost function is: ； In the formula, Here, represents the cost function; N represents the prediction time domain; i represents the sampling time number in the prediction time domain; and k represents the current sampling time number. e(k+i|k) For prediction i Control error after the cycle; u(k+i|k) For prediction i Control quantities after the cycle; u(k+i|k ) T For prediction i Transpose of the control quantity after the cycle; x(k+i|k) For prediction i The system state after the cycle; x(k+i|k) T For prediction i Transpose of the system state after the cycle; , and All are positive definite weight matrices.

7. The predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model according to claim 6, characterized in that, The quadratic programming problem is: ； Let the coefficient matrix of the quadratic term The coefficient vector of the first-order term The quadratic programming problem can be simplified to: ； In the formula, For predicting the transpose of the control error sequence; To predict the control error sequence; The positive definite weight matrix Conjugate; For the optimal control sequence U k Transpose of; The positive definite weight matrix Conjugate; For reference trajectory; C , M and L all The matrix is ​​constructed from the system model and the prediction time domain; W Perturbation prediction sequences provided for the extended state observer; This is the system's current return value; For matrix The transpose of .

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the predictive control method for a permanent magnet synchronous motor based on an adaptive digital twin model as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the predictive control method for permanent magnet synchronous motors based on an adaptive digital twin model as described in any one of claims 1-7.

Images