Permanent magnet synchronous motor active disturbance rejection gain progressive correction method, device and system

Abstract

This invention belongs to the field of permanent magnet synchronous motor control technology, and discloses a method, device, and system for progressive correction of the active disturbance rejection gain of a permanent magnet synchronous motor, applicable to the condition of gain mismatch in the input channel of the speed loop active disturbance rejection control caused by flux linkage changes. The device includes a model reference adaptive flux linkage identification module, a gate value generation module, a gate fusion and gain mapping module, a safety constraint projection module, and an active disturbance rejection gain synchronous update module. By acquiring current, voltage, speed, and control input signals, the permanent magnet flux linkage is identified and a desired gate value is generated. A feasible interval for the gate variable is constructed based on a preset input channel gain safety range. The gate variable for the next control cycle is obtained through closed projection, thereby updating the fused flux linkage and input channel gain. This invention can reduce the risk of gain abrupt changes caused by transient fluctuations and improve the smoothness of current regulation and the robustness of speed response.

Description

Technical Field

[0001] This invention belongs to the field of control technology for permanent magnet synchronous motors (PMSMs), specifically relating to a method, device, and system for progressive correction of the self-disturbance rejection gain of a permanent magnet synchronous motor. Background Technology

[0002] PMSMs (Power Mitigation Controllers) possess advantages such as high power density, high efficiency, large torque-to-inertia ratio, and good speed regulation performance, and have been widely used in new energy vehicles, electric servo systems, industrial automation equipment, and high-performance drive scenarios. The control performance of a PMSM depends not only on the design of the current loop and speed loop controllers, but is also affected by the permanent magnet flux linkage, inductance, stator resistance, load disturbances, sampling noise, and inverter non-ideal factors. In actual operation, motor temperature rise, magnetic saturation, changes in the current operating point, and sudden load changes can all cause deviations in key electromagnetic parameters, leading to model mismatch in controllers designed based on nominal parameters.

[0003] Active Disturbance Rejection Control (ADRC) can estimate the system state and total disturbance online through an Extended State Observer (ESO) and compensate for it in the control law, thus exhibiting good disturbance rejection capability in PMSM speed loop control. For first-order mechanical motion objects like the speed loop, ADRC typically reconstructs the controlled object into a series integrator form, allowing the control input channel gain and the total disturbance to jointly determine the speed dynamic response. If the input channel gain is accurately selected, the ESO mainly handles the estimation of load disturbance, friction damping, and unmodeled dynamics; however, if the input channel gain does not match the actual object, the gain mismatch term will be included in the total disturbance and passively compensated by the ESO, thereby increasing the observer's burden and weakening the speed loop's dynamic response and steady-state robustness.

[0004] In PMSM (Permanent Magnet Controller), the input channel gain is closely related to the permanent magnet flux linkage. Under control conditions where the d-axis current is approximately zero, the electromagnetic torque is mainly determined by the q-axis current and the permanent magnet flux linkage. When the flux linkage changes due to temperature or magnetic state, the fixed nominal input channel gain can no longer accurately reflect the actual motor. If the fixed nominal input channel gain is still used, a deviation in the proportional relationship between the speed loop controller output and the actual electromagnetic torque will occur, manifesting as overshoot during startup, slow recovery after sudden load changes, long duration of q-axis current oscillation, and increased phase lag under sinusoidal speed commands.

[0005] To improve the controller's adaptability to parameter changes, existing research typically employs online parameter identification methods to obtain key motor parameters, and then uses the identification results to correct the controller parameters. Model Reference Adaptive Systems (MRAS) are a class of online parameter identification methods with a clear structure and easy digital implementation. For flux linkage identification, a reference model based on the stator voltage model and an adjustable model based on the algebraic relationship between flux linkage and current can be constructed. The output error of both models is used to drive adaptive parameter updates, allowing the output of the adjustable model to gradually approach the output of the reference model.

[0006] However, online identification results cannot be directly injected into the control channel as controller gain. Online identification often exhibits short-term fluctuations during startup, speed change, and load surge phases, and may even lead to local oscillations due to sampling noise, inverter nonlinearity, or insufficient state excitation. If the identification flux of transient fluctuations is directly mapped to the ADRC input channel gain, the gain will abruptly change within adjacent control cycles, resulting in q-axis current command jumps, excessively rapid changes in the pulse width modulation (PWM) duty cycle, and torque response shocks. In severe cases, this may trigger overcurrent protection or cause mechanical vibration.

[0007] Traditional processing methods often employ low-pass filtering, rate limiting, or empirical weight fusion. Low-pass filtering can reduce high-frequency fluctuations in the identification value, but it introduces hysteresis, preventing the controller from promptly utilizing actual flux linkage changes. Rate limiting can suppress abrupt changes, but it cannot theoretically guarantee that the gain will always remain within the preset safe range of the input channel gain. Empirical weight fusion relies on manual adjustment, making it difficult to simultaneously achieve speed, smoothness, and safety under different operating conditions. For motor drive systems that need to execute in real-time with fixed control cycles in embedded controllers, the above methods lack a unified design for convergence, safety, and real-time feasibility.

[0008] The Control Lyapunov Function (CLF) can be used to constrain state variables to converge toward the desired value, the Control Barrier Function (CBF) can be used to constrain variables to always be within a safe set, and Quadratic Programming (QP) can unify convergence objectives and safety constraints into a single optimization framework. Introducing these theories into the PMSM input channel gain update process can provide a theoretical basis for online gain correction. However, simply relying on general mathematical constraints is insufficient to demonstrate their correspondence with the actual physical boundaries, sampling period, and PWM execution link of the motor drive system.

[0009] In summary, there is a need for a method, device, and system that can improve the accuracy of ADRC input channel gain by utilizing MRAS online flux linkage identification, and can also perform credibility judgment, gating fusion, safety boundary constraints, and closed projection solution on the identification results in each control cycle. This would enable the adaptive gain update process to gradually approach the parameters of the real object without causing current surges and torque abrupt changes due to transient fluctuations in identification. Summary of the Invention

[0010] The purpose of this invention is to provide a method, apparatus, and system for progressive correction of active disturbance rejection gain in permanent magnet synchronous motors, which solves the problems of input channel mismatch in fixed nominal input channel gain type active disturbance rejection control under varying permanent magnet synchronous motor parameters, and the problems of sudden changes in input channel gain, current surges, and torque fluctuations caused by the direct entry of online flux linkage identification results into the control channel.

[0011] To address the aforementioned issues, this invention introduces model reference adaptive flux linkage identification results into the speed loop active disturbance rejection control structure and establishes a gating fusion stage between the identified flux linkage and the nominal flux linkage. The controller generates a desired gating value based on the MRAS error amplitude index obtained through flux linkage identification error calculation and smoothing. It determines the safe interval of the gating variable based on a preset input channel gain safety range and updates the gating variable for the next control cycle through safety constraints, convergence constraints, and closed projection, enabling the identified flux linkage to participate in input channel gain calculation in a gradual, smooth, and bounded manner. The updated input channel gain synchronously acts on the speed loop active disturbance rejection control law and the extended state observer input channel, thereby reducing the impact of input channel gain mismatch and identification transient fluctuations on speed response, current regulation, and torque output.

[0012] This invention provides the following technical solution:

[0013] In a first aspect, the present invention provides a method for progressive correction of PMSM (Motor-Induced Disturbance Rejection Model) gain. This method includes: determining the number of motor pole pairs, equivalent system moment of inertia, nominal flux linkage, preset input channel gain safety range, and ADRC controller parameters during the offline design phase; and during the online operation phase, the controller acquires current, voltage, speed, and control input signals in each sampling cycle, and performs model reference adaptive flux linkage identification, identification error evaluation, gating fusion, safety boundary mapping, joint feasible interval construction, one-dimensional closed-loop projection solution, and ADRC input channel gain update. The control input signals include PWM duty cycle or inverter switching state, used to reconstruct the stator voltage signal required for the MRAS (Motor-Induced Disturbance Rejection Model) reference model in conjunction with the DC bus voltage.

[0014] In one implementation, the gated variable is used to adjust the weights of the identified flux linkage and the nominal flux linkage in the fused flux linkage. When the model reference adaptive flux linkage identification error is large, the weight of the identified flux linkage in the fused flux linkage is reduced, making the input channel gain more dependent on the nominal flux linkage; when the model reference adaptive flux linkage identification error decreases, the weight of the identified flux linkage in the fused flux linkage is gradually increased, making the input channel gain gradually approach the gain value determined by the online identified flux linkage.

[0015] Specifically, the controller determines the fused flux linkage based on the identified flux linkage, nominal flux linkage, and gating variables, and maps the fused flux linkage to the active disturbance rejection control input channel gain according to the input channel relationship of the permanent magnet synchronous motor speed loop. Therefore, the identified flux linkage does not directly replace the nominal flux linkage in the control channel, but rather undergoes gating fusion before participating in the input channel gain calculation.

[0016] Furthermore, the controller determines the fusion flux safety range based on the preset input channel gain safety range, and combines the deviation between the identified flux and the nominal flux, as well as the gating fusion relationship, to obtain the gating variable safety range through inverse solving. The gating variable safety range is used to limit the value range of the gating variable, preventing the input channel gain from exceeding the preset safety range due to transient fluctuations in the identified flux.

[0017] Furthermore, the controller constructs a joint feasible interval for the gated variables based on the safe feasible interval corresponding to the safety constraints of the control barrier function and the convergent feasible interval corresponding to the convergence constraints of the control Lyapunov function. The safe feasible interval is used to ensure that the updated gated variables do not exceed the preset safety boundary, and the convergent feasible interval is used to constrain the gated variables to converge asymptotically towards the desired gate value.

[0018] Furthermore, the controller constructs an unconstrained minimum point based on the desired gating value and the gating variable of the previous control cycle, and projects the unconstrained minimum point in a closed loop onto the joint feasible interval of the gating variable to obtain the gating variable for the next control cycle. Thus, the gating variable update process does not require an online iterative solver; it can be completed simply through comparison operations, arithmetic operations, and interval projection, making it suitable for real-time execution within a fixed control cycle.

[0019] Secondly, this invention provides a PMSM (Motor Motor Controller) active disturbance rejection gain asymptotic correction device. The device includes a model reference adaptive flux linkage identification module, a gating value generation module, a gating fusion and gain mapping module, a safety constraint projection module, and an active disturbance rejection gain synchronization update module. Each module can be implemented by a program module in a digital controller, or by interrupt tasks, parameter update tasks, and PWM refresh tasks in the motor controller firmware.

[0020] Specifically, the model reference adaptive system flux linkage identification module is used to obtain the identification flux linkage and MRAS flux linkage identification error based on the current, voltage, speed, and control input signal of the permanent magnet synchronous motor; the gating value generation module is used to generate a desired gating value based on the MRAS error amplitude index obtained by amplitude calculation and smoothing of the MRAS flux linkage identification error; the gating fusion and gain mapping module is used to perform gating fusion on the identification flux linkage and nominal flux linkage based on the gating variable to obtain the fused flux linkage, and map the fused flux linkage to the active disturbance rejection control input channel gain; the safety constraint projection module is used to convert the safety range of the active disturbance rejection control input channel gain. To merge the flux linkage safety interval, and based on the merged flux linkage safety interval and the gated fusion relationship between the identified flux linkage and the nominal flux linkage, the gated variable safety interval is obtained by inverse solution. It is also used to generate a joint feasible interval for the gated variables based on the safety constraints of the control barrier function and the convergence constraints of the control Lyapunov function, and to project the unconstrained minimum point within the joint feasible interval of the gated variables in a closed loop to obtain the gated variables for the next control cycle. The active disturbance rejection gain synchronization update module is used to recalculate the merged flux linkage based on the gated variables for the next control cycle, and to synchronously apply the input channel gain determined by the merged flux linkage to the velocity loop active disturbance rejection control law and the extended state observer input channel.

[0021] Thirdly, the present invention provides a PMSM (Progressive Disturbance Rejection Scale) gain correction system. The control system includes a PMSM, a three-phase inverter, a voltage sampling unit, a current sampling unit, a position and speed feedback unit, a digital controller, a PWM drive and modulation unit, and the PMSM gain correction device. In one embodiment, the control system further includes a DC power supply, a load device, and a host computer.

[0022] The voltage sampling unit is used to acquire the DC bus voltage sampling signal; the current sampling unit is used to acquire the motor phase current or the dq axis current feedback signal obtained by coordinate transformation; the position and speed feedback unit is used to acquire the rotor position and speed feedback signal; the digital controller is used to operate the PMSM active disturbance rejection gain progressive correction device.

[0023] The model reference adaptive flux linkage identification module in the gain asymptotic correction device is used to obtain the identification flux linkage and MRAS flux linkage identification error; the gate value generation module is used to generate the desired gate value based on the MRAS error amplitude index obtained by amplitude calculation and smoothing of the MRAS flux linkage identification error; the gate fusion and gain mapping module is used to generate the fused flux linkage and map the fused flux linkage to the input channel gain; the safety constraint projection module is used to generate the joint feasible interval of the gate variables and obtain the gate variables for the next control cycle through closed projection; the active disturbance rejection gain synchronization update module is used to synchronously apply the input channel gain to the speed loop ADRC control law and the ESO input channel.

[0024] The present invention has the following beneficial effects:

[0025] 1. This invention incorporates a gating fusion stage between the identified flux linkage and the nominal flux linkage, and constructs an identification confidence level based on the model reference adaptive flux linkage identification error. This ensures that the online identification result does not directly replace the nominal flux linkage, but rather gradually enters the ADRC input channel after the confidence level meets the requirements. Therefore, during periods of significant identification error, such as startup, speed changes, or sudden load changes, the controller can maintain an appropriate dependence on the nominal flux linkage, preventing transient fluctuations in the identified flux linkage from being directly amplified into input channel gain jumps, thereby suppressing sudden changes in q-axis current commands, current surges, and electromagnetic torque surges.

[0026] 2. This invention maps the preset input channel gain safety range to a fused flux linkage safety interval, and further converts it into a gated variable safety interval. Then, it combines CBF safety constraints and CLF convergence constraints to construct a joint feasible interval for the gated variables. Therefore, the update of the gated variables no longer relies on a single empirical limiting or low-pass filtering, but rather converges asymptotically towards the desired gate value while satisfying the gain safety boundary, ensuring that the input channel gain update process balances safety, convergence, and smoothness.

[0027] 3. This invention transforms the gated variable update problem into a strictly convex quadratic programming problem on a single-variable closed interval, and obtains the gated variable for the next control cycle through closed projection of the unconstrained minimum point to the joint feasible interval. This solution process involves only a small number of arithmetic operations, comparison operations, and interval projection operations, eliminating the need for online iterative optimization of the solver. It can be embedded into motor control interrupt flows and is suitable for real-time execution in microcontroller units (MCUs) or ARM architecture controllers, following the speed loop sampling cycle or PWM refresh cycle.

[0028] 4. This invention utilizes the model reference adaptive flux linkage identification results to correct the ADRC input channel gain online, enabling the controller's nominal input channel to progressively calibrate with changes in the fused flux linkage, thereby reducing the mismatch between the fixed nominal input channel gain and the actual motor object. Consequently, the ESO no longer needs to passively compensate for the equivalent modeling error caused by flux linkage parameter mismatch over a long period, and can be used more extensively to estimate load disturbances, friction damping, and unmodeled dynamics, which is beneficial for improving the dynamic response and disturbance rejection performance of the PMSM speed loop.

[0029] 5. This invention can improve the adaptability of the input channel gain to changes in flux linkage while keeping the ADRC structure basically unchanged, and improve the current smoothness and speed tracking quality of the motor drive system. Attached Figure Description

[0030] Figure 1This is a schematic diagram of the PMSM active disturbance rejection gain asymptotic correction control structure.

[0031] Figure 2 This is a flowchart of the method of the present invention.

[0032] Figure 3 A schematic diagram for identifying the data source for MRAS magnet links.

[0033] Figure 4 The waveform diagram shows the gated fusion and input channel gain update under steady-state conditions. Figure 4 (a) shows the evolution waveforms of rotational speed, MRAS error amplitude index, and gated variable. Figure 4 (b) Waveform diagram for identifying flux linkage, fusing flux linkage and input channel gain update.

[0034] Figure 5 The waveform diagrams are of key variables for gating fusion and input channel gain update under sinusoidal variable speed operation. Figure 5 (a) is a waveform diagram of the MRAS error amplitude index. Figure 5 (b) Waveform diagrams for identifying flux linkage and fused flux linkage. Figure 5 (c) is the input channel gain waveform. Figure 5 (d) is a waveform diagram of the gated variable.

[0035] Figure 6 This is a comparison chart of simulation results under constant speed and variable load conditions, in which... Figure 6 (a) is a comparison of the simulation response to the q-axis current command. Figure 6 (b) is a comparison of the speed response simulation.

[0036] Figure 7 This is a schematic diagram of the hardware connection of the PMSM control system.

[0037] Figure 8 This is a comparison chart of simulation results under constant torque variable speed conditions, in which... Figure 8 (a) is a comparison chart of simulated rotational speed response. Figure 8 (b) is a full-time domain comparison diagram of q-axis current commands. Figure 8 (c) is a magnified comparison of the q-axis current command.

[0038] Figure 9 This is a comparison chart of simulation results under sinusoidal variable speed conditions, in which... Figure 9 (a) is a comparison chart of speed tracking simulations. Figure 9 (b) is a comparison chart of the q-axis current command across the entire time domain. Figure 9 (c) is a magnified comparison of the q-axis current command.

[0039] Figure 10 The waveform diagram is shown in the constant torque speed change experiment under a fixed nominal input channel gain. Figure 10(a) is the experimental waveform of the dq-axis current. Figure 10 (b) is a waveform diagram of speed command and speed feedback.

[0040] Figure 11 The waveform diagram is shown in the constant torque speed change experiment using the method of the present invention. Figure 11 (a) is the experimental waveform of the dq-axis current. Figure 11 (b) shows the speed command and speed feedback waveforms. Figure 11 (c) Waveforms for identifying flux linkage, merging flux linkage, and input channel gain.

[0041] Figure 12 The waveform diagram for the sinusoidal speed command experiment under a fixed nominal input channel gain is shown below. Figure 12 (a) is the experimental waveform of the dq-axis current. Figure 12 (b) is a waveform diagram of speed command and speed feedback.

[0042] Figure 13 This is an experimental waveform diagram of the sinusoidal velocity command under the method of the present invention, wherein... Figure 13 (a) is the experimental waveform of the dq-axis current. Figure 13 (b) shows the speed command and speed feedback waveforms. Figure 13 (c) Waveforms for identifying flux linkage, merging flux linkage, and input channel gain.

[0043] Figure 14 The waveform diagram is shown in the constant speed variable load experiment under a fixed nominal input channel gain. Figure 14 (a) is the experimental waveform of the dq-axis current. Figure 14 (b) is a waveform diagram of speed command and speed feedback.

[0044] Figure 15 The waveform diagram is shown in the constant speed variable load test under the method of the present invention. Figure 15 (a) is the experimental waveform of the dq-axis current. Figure 15 (b) shows the speed command and speed feedback waveforms. Figure 15 (c) Waveforms for identifying flux linkage, merging flux linkage, and input channel gain. Detailed Implementation

[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments. The described embodiments are used to explain the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Without departing from the core idea of ​​the present invention, the relevant controller, sampling interface, PWM modulation method and control cycle can be adaptively adjusted according to the specific motor drive platform.

[0046] In one implementation, the system composition and signal links are as follows:

[0047] like Figure 7As shown, the PMSM control system in this embodiment includes a PMSM, a three-phase inverter, a DC power supply, a load device, a voltage sampling unit, a current sampling unit, a position and speed feedback unit, a digital controller, a PWM drive and modulation unit, and a host computer. Figure 1 This represents the speed loop ADRC control link and the input channel gain asymptotic correction link within the digital controller. The digital controller can be an MCU or ARM architecture controller, internally running the speed loop ADRC, MRAS flux linkage identification, and the input channel gain asymptotic correction algorithm of this invention. A DC power supply provides DC bus voltage to the three-phase inverter, which outputs three-phase drive voltage to the PMSM according to the PWM drive signal. The PMSM drives the load device through a mechanical connection. The voltage sampling unit, current sampling unit, and position and speed feedback unit provide voltage sampling signals, current sampling signals, and position and speed feedback signals to the digital controller, respectively. Figure 7 The gated fusion and gain asymptotic correction unit in the middle is Figure 1 and Figure 2 The diagram illustrates the integrated representation of gate value generation, gate fusion and gain mapping, safety constraint projection, and active disturbance rejection gain synchronization update functions within the digital controller. The MRAS flux linkage identification unit outputs identified flux linkage and error information to the gate fusion and gain asymptotic correction unit; the gate fusion and gain asymptotic correction unit outputs the asymptotically corrected input channel gain to the ADRC speed loop control unit; the ADRC speed loop control unit includes an observer state and disturbance estimation link, and a voltage sampling unit for acquiring the DC bus voltage; the digital controller reconstructs the stator voltage signal required for the MRAS reference model based on the DC bus voltage and the PWM duty cycle or switching state.

[0048] like Figure 1 As shown, the PMSM speed loop control main link is: speed setpoint With actual speed feedback value The comparison generates a speed error signal, which is then processed by the proportional feedback gain in the speed loop active disturbance rejection law. After the action, a velocity loop control quantity is generated, which, combined with the total disturbance estimate output by the observer and the input channel gain, generates a q-axis current command. The q-axis current command is input into the current loop, which drives the PMSM to operate based on the current closed-loop control result. The PMSM outputs the actual rotational speed. It is then fed back to the velocity loop and the observer state and disturbance estimation link. Figure 1 The right side consists of an integrator and an observer gain. and and state estimator and This constitutes an observer state and disturbance estimation link, used to estimate the speed state and total disturbance based on the actual rotational speed feedback value, the q-axis current feedback value, and the input channel gain. This represents the velocity state estimate. This represents the estimated total disturbance. and This represents the observer gain. The observer state corresponds to the extended state observer function in the speed loop ADRC, and the output total disturbance estimate is used for disturbance compensation in the speed loop ADRC control law.

[0049] Figure 1 The gain asymptotic correction link in the middle receives the identified flux from the MRAS flux identification unit. and MRAS error amplitude index This includes gated fusion, security boundary mapping, CBF feasible region construction, CLF feasible region construction, and QP closed projection. It also includes identifying magnetic flux linkages. With nominal magnetic flux Gated variables After fusion, a fused magnetic flux is obtained. The fused flux linkage is further mapped to the input channel gain. , Figure 1 In With the formula in this article They have the same meaning, both referring to the input channel gain used by the speed loop ADRC.

[0050] During the gating variable update process, the input channel gain safety boundary and The fusion flux safety boundary is obtained through equivalent mapping. and And further construct the safe and feasible range of CBF. The MRAS error magnitude index is used to construct the desired gate value. Furthermore, the convergence feasible interval of CLF is constructed. The final gated feasible interval is obtained by intersecting the safe feasible interval of CBF and the convergent feasible interval of CLF. The closed projection stage in the safety constraint projection module calculates the gate variable for the next control cycle based on the final gated feasible interval and the gate variable of the previous control cycle. Figure 1 In This represents a control cycle delay, used to store the gating variables from the previous control cycle. This indicates the integration process.

[0051] thus, Figure 1 The control structure shown causes the identified flux output by the MRAS flux identification unit to not directly enter the ADRC input channel, but rather to first... and The system integrates gated value generation, gated fusion, safety boundary constraints, CBF / CLF joint constraints, and closed projection processing. The input channel gain obtained by the fused flux linkage mapping is then synchronously applied to the velocity loop ADRC control law and the observer input channel, thereby avoiding the direct abrupt change in input channel gain caused by identifying transient flux linkage fluctuations.

[0052] like Figure 2 As shown, within each control cycle, the controller first acquires current, voltage, speed, and control input signals, and sends these signals to the MRAS flux linkage identification unit to obtain the identified flux linkage and the flux linkage identification error between the MRAS reference model and the adjustable model. Subsequently, the MRAS error magnitude index is calculated from the MRAS flux linkage identification error, and the desired gating value is generated. Simultaneously, a safe interval for the gating variable is constructed based on the deviation between the identified flux linkage and the nominal flux linkage. The controller further constructs a joint feasible interval for CBF / CLF and obtains the gating variable for the next control cycle through closed projection. Then, this gating variable is used to perform gating fusion of the identified flux linkage and the nominal flux linkage, mapping the fused flux linkage to the input channel gain, and synchronously updating the ADRC control law and the ESO input channel.

[0053] The MRAS error amplitude index is not directly used to change the motor output, but is advanced into the gate value generation module. This unit generates the desired gate value according to equation (13), and converts the control logic of "the smaller the identification error, the more reliable the identification value" into a continuous gate quantity. Subsequently, the gate fusion link in the gate fusion and gain mapping module generates the fusion flux according to equation (1).

[0054] The fused flux linkage is mapped to the input channel gain via equation (7), and the nominal input channel gain used in the speed loop control law and ESO is synchronously replaced by the active disturbance rejection gain synchronization update module. Since this gain update is performed under the constraint of the closed projection loop in the safety constraint projection module, the risk of significant jumps in input channel gain caused by transient fluctuations can be reduced. Finally, the q-axis current command output by the speed loop enters the current loop, and a PWM signal is generated through coordinate transformation and space vector pulse width modulation (SVPWM) to drive the three-phase inverter to act on the PMSM.

[0055] Furthermore, the implementation method for MRAS magnetic flux linkage identification is as follows:

[0056] like Figure 3As shown, the MRAS flux linkage identification unit includes an MRAS reference model, an MRAS adjustable model, and a parameter adaptive update stage. The MRAS reference model obtains the reference model flux linkage based on voltage sampling signals, current sampling signals, and position and speed feedback signals; the MRAS adjustable model obtains the adjustable model flux linkage based on current sampling signals, position and speed feedback signals, and adjustable parameters.

[0057] The difference between the reference model flux linkage and the adjustable model flux linkage forms the MRAS flux linkage identification error. This MRAS flux linkage identification error serves two purposes: firstly, it is used for adaptive parameter updates, feeding the updated adjustable parameters back to the MRAS adjustable model, allowing the adjustable model output to gradually approximate the reference model output, thereby obtaining the identified flux linkage; secondly, through amplitude calculation and smoothing, it yields the MRAS error amplitude index, which characterizes the reliability of the current identification result.

[0058] The identified flux linkage serves as the input to the gating fusion and gain mapping module, used for gating fusion with the nominal flux linkage; the MRAS error amplitude index serves as the input to the gating value generation module, used to generate the desired gating value. Therefore, the MRAS flux linkage identification unit does not directly change the ADRC input channel gain, but rather provides data for subsequent gating value generation, gating fusion and gain mapping, and safety constraint projection stages.

[0059] In one implementation, the voltage sampling unit acquires the DC bus voltage, and the digital controller combines this with the PWM duty cycle or switching state to obtain the voltage information required for MRAS reference model calculation. The current sampling unit acquires the inverter output phase current, and the position and speed feedback unit provides rotor position and speed feedback signals. These signals are collectively used in the MRAS flux linkage identification process.

[0060] To reduce the impact of short-term fluctuations in identification errors during startup, speed change, and load surge phases on subsequent control processes, the MRAS flux linkage identification error is smoothed before generating the MRAS error amplitude index. This smoothing can be achieved using low-pass filtering, moving average, or exponential smoothing. Specific smoothing parameters can be set according to the sampling period and motor operating conditions.

[0061] In this invention, the MRAS flux linkage identification unit is mainly used to provide identification flux linkage and MRAS error amplitude indicators. To avoid the MRAS output transient error directly entering the ADRC controller, this invention does not directly replace the nominal flux linkage with the identification flux linkage. Instead, it determines the degree to which the identification flux linkage enters the input channel gain through gating variables, safety boundary constraints, the CBF / CLF joint feasible interval, and closed projection.

[0062] Furthermore, the gating fusion and gain mapping implementation methods are as follows:

[0063] like Figures 1 to 5 As shown, this implementation uses gated variables as the core, connecting the MRAS identification flux, nominal flux, preset input channel gain safety range, CLF convergence constraint, and CBF safety constraint into a closed-loop derivation chain. This derivation chain is used to explain the source of the gated feasible interval in equation (20) and to give the closed-loop projection optimization problem shown in equation (21) a clear precondition basis.

[0064] First, the gating fusion stage in the gating fusion and gain mapping module calculates the fused flux linkage actually used in the control channel based on the identification flux linkage value output online by MRAS and the pre-given nominal flux linkage value. Its expression is:

[0065] (1);

[0066] In the formula, for The flux linkage value after time fusion, superscript Indicates the amount of fusion. for Gated variables at time t, for The magnetic flux linkage value at any given time. This is the nominal flux linkage value. When... When = 1, the fused flux linkage is exactly equal to the identified flux linkage, and the system fully trusts the identification result; when When the flux linkage is 0, the fused flux linkage is exactly equal to the nominal flux linkage, and the system adopts a conservative nominal gain.

[0067] Based on the electromagnetic torque model of the motor, the ADRC input channel gain is determined by the fused flux linkage value, and the expression is as follows:

[0068] (2);

[0069] In the formula, ADRC input channel gain; The mapping function from flux linkage to input channel gain, determined based on the motor electromagnetic torque model, is predetermined during the controller design phase.

[0070] When the d-axis current is approximately zero, the electromagnetic torque is mainly determined by the q-axis current and the permanent magnet flux linkage, and the corresponding expression is:

[0071] (3);

[0072] In the formula, For electromagnetic torque, This represents the number of pole pairs of the motor. This refers to the flux linkage amplitude of the permanent magnet. The q-axis current feedback value is obtained through coordinate transformation. The coefficient 3 / 2 originates from the power conservation constraint of the coordinate transformation from a three-phase stationary coordinate system to a two-phase rotating coordinate system (Clarke-Park transformation). This equation shows that under a control strategy where the d-axis current is approximately zero, the electromagnetic torque and the q-axis current are linearly related, and the proportionality coefficient is determined by the number of pole pairs and the amplitude of the permanent magnet flux linkage. Substituting this equation into the mechanical dynamics equation, we can obtain the speed dynamic equation:

[0073] (4);

[0074] In the formula, The equivalent rotational inertia of the system, For load torque, The viscous damping coefficient is... The rotor's mechanical angular velocity, for The derivative with respect to time. The ADRC velocity loop writes the object as a cascaded integrator:

[0075] (5);

[0076] In the formula, The total disturbance term in the speed loop includes the equivalent disturbances referred to the speed loop, such as load torque, viscous damping, parameter perturbations, and unmodeled dynamics. From equations (3) to (5), the physical mapping relationship between the input channel gain and the permanent magnet flux linkage can be obtained:

[0077] (6);

[0078] Substituting the fused flux linkage in equation (1) into equation (6), we obtain the online updated input channel gain:

[0079] (7);

[0080] When using non-zero d-axis current control, field weakening control, or other current distribution strategies, the mapping function from flux linkage to input channel gain can be redefined based on the corresponding PMSM electromagnetic torque model; gating fusion, safety boundary constraints, CBF / CLF joint feasible interval, and closed projection update process remain unchanged.

[0081] To prevent the input channel gain from exceeding the preset safety range due to flux linkage fusion, a lower limit for the input channel gain is preset. and upper limit of input channel gain And convert it equivalently into a usable magnetic flux safety zone:

[0082] (8);

[0083] In the formula, and These are respectively determined by the lower limit of the input channel gain. and upper limit of input channel gain The equivalent conversion yields the lower and upper bounds of the fused flux linkage safety. The deviation between the identified flux linkage and the nominal flux linkage is:

[0084] (9);

[0085] From equations (1), (8), and (9), it can be seen that the safe interval of the gated variable comes from the inverse solution of the fused flux linkage safe interval. To simultaneously adapt... For the case where the value is positive or negative, the two boundary candidate values ​​can be written as:

[0086] (10);

[0087] In the formula, and These are the two candidate boundary values ​​for the gated variables obtained by inverse solving of the upper and lower boundaries of the fused magnetic flux safety interval. When the gating variable is at that time, the lower and upper bounds are:

[0088] (11);

[0089] In the formula, and These are the lower and upper bounds of the safe interval for the gated variable, respectively. When hour, To preset the flux linkage deviation threshold, assuming the identified flux linkage is approximately the same as the nominal flux linkage, [0,1] can be selected as the safe interval, or the safe value of the gate variable from the previous control cycle can be maintained, thus avoiding numerical amplification when the denominator is too small. Therefore, the safe and feasible interval for CBF is obtained:

[0090] (12);

[0091] In the formula, For the first The safe and feasible interval under the CBF constraint at each sampling time.

[0092] To ensure that the gated variables not only remain within the safety boundary but also converge toward the credible identification direction at the desired rate, an MRAS error magnitude index is constructed. And determine the expected gate value based on this error index. :

[0093] (13);

[0094] In the formula, The MRAS error amplitude index. , This is the gating sensitivity coefficient, used to adjust the sensitivity of the desired gating value to the identification error. When the MRAS identification error amplitude index... When the value is large, i.e. when non-convergence is detected or the operating conditions change drastically, the function value of equation (13) approaches zero. Automatically tending towards zero makes the fused flux linkage more conservative towards the nominal flux linkage value; when When approaching zero, The system automatically tends towards complete trust in the identification results, allowing for a smooth transition from fused flux linkage to the identified flux linkage value. Parameters The sensitivity of this mapping has been adjusted: The larger, The more sensitive the gating error is, the more cautious the gating behavior becomes. The smaller the value, the more aggressive the gating behavior. The selection of this function ensures... It always holds true and has a monotonically decreasing property, providing a mathematical guarantee for the rationality and feasibility of the CLF constraint convergence objective.

[0095] CBF constraints ensure that gated variables do not cross the safe zone through upper and lower boundary barrier functions, which are:

[0096] (14);

[0097] In the formula, This is the barrier function corresponding to the lower limit, whose value is equal to the difference between the current gate variable value and the lower limit value. Descending to At that time, B1=0; For the barrier function corresponding to the upper limit, its value is equal to the difference between the upper limit value and the current gate variable value. Rise to When B2 = 0, both are non-negative functions, taking positive values ​​within the safe region and zero values ​​at the boundary.

[0098] Applying an exponential barrier inequality to the barrier function:

[0099] (15);

[0100] In the formula, This is the index of the barrier function. Corresponding lower limit barrier function , Corresponding upper limit barrier function , Represents the barrier function with respect to time The derivative of the barrier function describes the rate and direction of change of the barrier function value over time; is a positive constant used to adjust the strictness of the obstacle constraint. Equation (15) means that even if the obstacle function value decreases, its rate of decrease... It shall not exceed .because As the threshold approaches the boundary, the threshold gradually decreases to zero, and the allowable decay rate also decreases to zero simultaneously. This makes the resistance stronger the gating variable gets as it gets closer to the boundary, eventually making it difficult to cross the boundary. In the discrete control implementation of this embodiment, the CBF safety constraint is specifically embodied in the safe and feasible interval of the gating variable shown in equation (12), and the closed projection shown in equation (23) restricts the gating variable from exceeding the safe and feasible interval in the next control cycle. Equations (14) and (15) are used to explain the control meaning of maintaining the safety boundary.

[0101] CLF constraints are used to ensure that the gated variable converges to the desired gated value. The tracking error and CLF are defined as follows:

[0102] (16);

[0103] In the formula, for Gating tracking error at any given moment; The CLF is constructed based on the tracking error. Its value is equal to half the square of the tracking error. It has positive definiteness and takes the value of zero if and only if the gate variable is exactly equal to the expected value. for The expected gating value at any given time.

[0104] The continuous domain exponential decay constraint is:

[0105] (17);

[0106] In the formula, Let be the exponential decay rate constant of the continuous domain CLF. In digital control systems, the continuous domain convergence requirement can be transformed into the following gated tracking error amplitude contraction constraint based on the control cycle:

[0107] (18);

[0108] In the formula, , The preset CLF discrete shrinkage parameters, This is the allowable contraction coefficient for the amplitude of the gated tracking error; For discrete control cycles; The discrete sampling time number; and The first The and the first +1 sampling time point of the gated variable value; For the first The expected gate value at each sampling time. The physical meaning of equation (18) is: the next gate variable. Compared with expected value The absolute value of the error between steps must not exceed the absolute value of the error in the current step. This results in a sequence of allowable intervals that gradually shrink towards the expected value.

[0109] make Then the feasible interval for CLF convergence is:

[0110] (19);

[0111] In the formula, For the first The convergence feasible interval is determined by the CLF convergence constraint at each sampling time. In actual control, the gated variable must simultaneously satisfy both the CBF safety constraint and the CLF convergence constraint. Therefore, the intersection of the CBF feasible interval and the CLF feasible interval is used to obtain the final gated feasible interval:

[0112] (20);

[0113] In the formula, For the first The final gated feasible interval at each sampling time. This is the lower bound of the final gated feasible interval. As the upper bound of the final gated feasible interval, equation (20) is the source of the gated feasible interval in the subsequent QP problem. In one implementation, when the intersection of the CBF safe feasible interval and the CLF convergent feasible interval is empty, the controller prioritizes satisfying the CBF safety constraint and projects the unconstrained minimum point onto the CBF safe feasible interval; then, in the next control cycle, the expected gate value, the gated variable safe interval, and the joint feasible interval are recalculated. In another optional implementation, the controller can keep the gated variable at the safe value of the previous control cycle.

[0114] The final feasible interval is obtained by intersecting the CBF safe interval and the CLF convergence interval. Within this interval, construct a one-dimensional strictly convex quadratic programming problem using the principle of minimum intervention:

[0115] (twenty one);

[0116] In the formula, This represents finding the value of the independent variable that minimizes the objective function; For optimization variables, represents the next gate value to be solved; The feasible interval is gated; the objective function consists of two weighted sums of squares: the first term... The second term represents the square of the deviation between the updated gate value and the expected value, which drives the optimization result to be as close to the expected value as possible; This represents the square of the deviation between the updated gate value and the current value. This term suppresses drastic jumps in the gate value between adjacent sampling periods. For smoothing weighting coefficients, ≥0 is used to make a trade-off between the goals of tracking expected values ​​and suppressing mutations.

[0117] When interval constraints are ignored, the unconstrained minimum point is:

[0118] (twenty two);

[0119] In the formula, This represents the value of the gate variable that minimizes the objective function under unconstrained conditions. The subscript 0 indicates that this point is the unconstrained optimal solution. The unconstrained minimum point is the expected value. and current value The weighted average, with weights of 1 and 2 respectively. .

[0120] The final optimal gated variable is the closed projection of the unconstrained minimum point onto the feasible interval:

[0121] (twenty three);

[0122] In the formula, For interval projection operators. Equation (23) only requires comparison operations and a small number of arithmetic operations, without the need for an iterative solver, and is suitable for real-time execution by MCU or ARM architecture controllers within a fixed control cycle.

[0123] When the unconstrained minimum point is within the final gated feasible interval, the closed projection result is equal to the unconstrained minimum point; when the unconstrained minimum point exceeds the final gated feasible interval, the closed projection result takes the nearest interval endpoint. This ensures that the gated variables in the next control cycle remain within the final gated feasible interval. The closed projection involves only comparison operations and a small number of arithmetic operations, requiring no online iterative solver, making it suitable for real-time execution by a digital controller within a fixed control cycle.

[0124] Furthermore, the coupling method between the asymptotic gain correction, the ADRC control law, and the ESO is as follows:

[0125] In ADRC speed loop, the ESO is used to estimate the speed state and total disturbance. In traditional fixed nominal input channel gain ADRC, Since the input channel gain of the actual motor changes with flux linkage, the input channel error will be included in the total disturbance term. At this time, the ESO not only needs to estimate the load disturbance, but also needs to compensate for the equivalent modeling error caused by flux linkage parameter mismatch.

[0126] This invention constructs the input channel gain by fusing magnetic flux identification using MRAS and gating, enabling the controller's nominal input channel to progressively correct for changes in actual magnetic flux. Ideally, as the identified magnetic flux converges and the gating variables are gradually released, the input channel error term decreases, the modeling error compensation burden on the ESO is reduced, and the disturbance estimation is more focused on load changes and unmodeled mechanical disturbances.

[0127] In the specific implementation, the updated input channel gain applies to both the ADRC control law and the ESO input channel, ensuring that the controller and observer use consistent input channel gains. If only the control law is updated without updating the ESO, a new gain inconsistency will occur between the control law and the observer; if only the ESO is updated without updating the control law, the control input mapping will still fail to match the real object. Therefore, this invention emphasizes the synchronous updating of the control law and the ESO.

[0128] The synchronous update does not change the original ADRC structure and does not require the addition of higher-order observers or complex optimizers. Its core is to introduce a time-varying gain driven by fusion flux at the input channel gain position of the original ADRC framework, while ensuring the safety and asymptotic nature of this time-varying gain through gated fusion and closed projection.

[0129] In one implementation, the hardware and software embodiments are as follows:

[0130] like Figure 7 As shown, this invention can be implemented in software within the digital controller of a PMSM control system. The digital controller receives voltage sampling signals, current sampling signals, and position and speed feedback signals, and performs MRAS flux linkage identification, gating fusion and gain asymptotic correction, ADRC speed loop control, and PWM drive according to control parameters or speed commands sent from the host computer. The PWM drive signal acts on the three-phase inverter, which outputs a three-phase drive voltage to the PMSM, which then drives the load device via a mechanical connection.

[0131] In terms of software implementation, after the system powers on, it first completes the initialization of the clock, analog-to-digital converter, PWM, communication, encoder interface, and protection threshold; then it enters the main loop or interrupt-driven operation mode. In the high-speed interrupt, current sampling, coordinate transformation, current loop control, and PWM update are performed; in the speed loop cycle, speed error calculation, ADRC control, MRAS flux linkage identification, input channel gain asymptotic correction, and ESO status update are performed.

[0132] The system self-test includes bus voltage detection, phase current detection, driver temperature detection, motor temperature detection, encoder status detection, and overcurrent status detection. When an abnormal state is detected, the controller shuts down the bridge arm output signal, putting the inverter into a safe state. The gain progressive correction process of this invention can run in parallel with this safety monitoring mechanism; when abnormalities are detected in bus voltage, phase current, encoder, or temperature, the controller freezes the gated variables or restores the input channel gain to the nominal input channel gain determined by the nominal flux linkage, and suspends the continued input of MRAS identification results into the input channel. After the abnormality is resolved, the controller recalculates the MRAS error amplitude index and the safe range of the gated variables, gradually restoring the gated variables from the safe state to the normal closed-loop projection update process.

[0133] Since this invention does not alter the basic framework of the current loop, coordinate transformation, and PWM modulation, it can be embedded in existing PMSM vector control systems or implemented in software as an online gain correction module for the speed loop ADRC input channel. The above-described combined implementation methods do not affect the implementation of this invention as an independent gain asymptotic correction method.

[0134] In one implementation, the simulation example is as follows:

[0135] This embodiment verifies the self-disturbance rejection gain asymptotic correction method by constructing a permanent magnet synchronous motor vector control simulation model in MATLAB / Simulink. The simulation model is based on the permanent magnet synchronous motor vector control structure and includes a motor body model, an inverter and modulation module, a coordinate transformation module, a current loop control module, a speed loop ADRC control module, a MRAS flux linkage identification module, and an input channel gain asymptotic correction module. During the simulation, the d-axis current setpoint is set to zero, and the q-axis current command is generated by the speed loop ADRC controller. The MRAS flux linkage identification module obtains the identified flux linkage and MRAS flux linkage identification error based on voltage, current, speed, and control input signals. The input channel gain asymptotic correction module generates the desired gate value based on the MRAS error amplitude index, and obtains the gate variable for the next control cycle through gate fusion, safe interval construction, CBF / CLF constraint projection, and closed projection, thereby updating the fused flux linkage and input channel gain. The updated input channel gain synchronously acts on the speed loop ADRC control law and the observer input channel. By comparing with the ADRC (Advanced Dynamic Range) of a fixed nominal input channel gain, the suppressive effect of this invention on sudden changes in input channel gain and gain mismatch under conditions of speed command variation, load disturbance, and flux linkage identification fluctuation is verified. To illustrate the simulation conditions of this embodiment, the simulation model and main operating parameters are shown in Table 1. The parameters listed in Table 1 are those of a specific embodiment, used to illustrate the simulation implementation conditions of this invention, and do not constitute a limitation on the scope of protection of this invention.

[0136] Table 1. Main parameters of the simulation model:

[0137]

[0138] like Figure 4 As shown in (a), under steady-state conditions, the MRAS error amplitude index gradually decreases, and the gated variable gradually increases from a small initial value and tends to stabilize; as Figure 4 As shown in (b), the identification flux linkage, fusion flux linkage, and input channel gain all exhibit a smooth transition process. This result indicates that as MRAS identification gradually converges, the gating variables can gradually loosen the identification flux linkage weights, rather than directly using the identification results in the initial stage.

[0139] like Figure 5 As shown in (a), under sinusoidal variable speed conditions, the MRAS error amplitude index fluctuates periodically with changes in the operating conditions; as Figure 5 As shown in (b), the identified magnetic flux exhibits periodic variations, while the fused magnetic flux remains relatively smooth; as Figure 5 (c) and Figure 5 As shown in (d), the input channel gain and gating variable are dynamically adjusted with the error amplitude, indicating that the gating fusion mechanism can tighten the identification flux weight near the peak of the identification error and gradually loosen it after the error falls back.

[0140] like Figure 6 As shown in (a), after introducing asymptotic correction for the input channel gain, the peak-to-peak oscillation of the q-axis current command decreases from approximately 9.05A to approximately 6.88A, a reduction of approximately 24%, and the oscillation duration is significantly shortened; Figure 6 As shown in (b), in constant speed variable load simulation, both the original ADRC and the incremental gain correction method of this invention can eventually track the speed command. However, the method of this invention has a faster start-up response, enters steady state earlier, and recovers faster after a sudden load change. This result shows that this invention does not simply change the fixed parameters of the controller, but rather, under the constraints of identification confidence and the preset safe range of input channel gain, makes the input channel gain gradually approach the actual object, thereby reducing the equivalent disturbance caused by flux mismatch.

[0141] like Figure 8 (a) to Figure 8 As shown in (c), under constant torque variable speed operation, the load torque remains constant, and the speed command undergoes periodic step switching. Compared with the original ADRC, the method of this invention has a faster speed tracking response and can suppress q-axis current command oscillation during the speed step switching process.

[0142] like Figure 9 (a) to Figure 9As shown in (c), under sinusoidal speed command conditions, the method of the present invention can reduce speed tracking phase lag and improve amplitude tracking performance. Specifically, the phase lag is reduced from 0.3937s to 0.1983s, the peak-to-peak speed is increased from 348.01rpm to 381.08rpm, and the q-axis current ripple during periodic speed changes is reduced, indicating that online correction of the input channel gain improves dynamic tracking capability.

[0143] In one implementation method, the experimental example is as follows:

[0144] In the experimental verification, the PMSM was used to simulate typical operating conditions such as constant torque speed change, sinusoidal speed command, and constant speed load change on the drag-load platform, and data such as phase current, dq axis current, speed command, speed feedback, fused flux linkage, and input channel gain were collected. To illustrate the experimental conditions of this embodiment, the main parameters of the experimental platform and operating conditions are shown in Table 2. The parameters listed in Table 2 represent a specific experimental platform configuration, used to illustrate the experimental implementation conditions of the present invention, and do not constitute a limitation on the scope of protection of the present invention. The experimental platform serves as a verification environment and does not limit the system structure of the present invention; the hardware connection relationships of the control system of the present invention are as follows: Figure 7 As shown.

[0145] Table 2 Main parameters of the experimental platform:

[0146]

[0147] For ease of explanation Figures 10 to 15 The experimental waveforms shown in the figure are explained below for the main variables: The d-axis current command value is used; in this embodiment, a control method with a d-axis current command value of zero is adopted. The d-axis current feedback value is obtained through coordinate transformation; This is the q-axis current command value output by the speed loop ADRC control unit; This is the q-axis current feedback value obtained through coordinate transformation; This is the q-axis current command limit value, used to restrict the q-axis current command from exceeding the allowable range of the drive system. Figure 11 (c) Figure 13 (c) and Figure 15 (c) The identified flux linkage is the output of the adaptive flux linkage identification module of the model reference. For the fused magnetic flux obtained according to equation (1), The input channel gain is obtained by the fusion flux mapping according to equation (7).

[0148] like Figure 10 and Figure 11As shown, in the constant torque speed change experiment, the load torque is kept at 5 N·m, and the speed command is switched from 500 rpm to 1000 rpm during the effective operation phase to examine the current regulation process and speed tracking performance under the condition of step change in speed command. Figure 10 The experimental waveforms are obtained under the condition of fixed nominal input channel gain. Figure 11 The waveforms are experimental waveforms obtained using the method of this invention.

[0149] Depend on Figure 10 (a) It can be seen that, under the condition of fixed nominal input channel gain, Keep it at zero. It fluctuates around zero. and Adjustments are made according to speed commands. The d-axis current feedback fluctuation, marked in the zoomed-in window, is 8.18A. (From...) Figure 10 (b) It is evident that during both the startup and speed command switching phases, there is a certain degree of callback in the speed feedback, with the callback amounts marked in the magnified window being 134.071 rpm and 106.286 rpm, respectively. This phenomenon indicates that when a fixed nominal input channel gain is used, the mismatch between the nominal input channel gain and the actual motor requires compensation through a significant current regulation process, thus affecting the compactness of the speed response.

[0150] Depend on Figure 11 (a) As can be seen, after adopting the method of the present invention, It still fluctuates around zero, and the d-axis current feedback fluctuation marked in the local magnified window has decreased to 6.54A; and They maintain mutual following during speed switching. Figure 11 (b) It can be seen that the callback amounts during the startup phase and the speed command switching phase decreased to 111.022 rpm and 85.824 rpm, respectively. Figure 11 (c) It is evident that identifying magnetic chains There are certain fluctuations when the operating status changes, while the fused magnetic flux... The transient changes in the identified flux linkage were not tracked with the same amplitude; input channel gain No significant abrupt changes occurred as the fusion flux was continuously adjusted. These results demonstrate that the present invention progressively incorporates MRAS identification results into the input channel gain through gated fusion and safety constraint projection, which helps reduce gain mismatch during speed switching and improves current regulation and speed tracking performance.

[0151] like Figure 12 and Figure 13As shown, in the sinusoidal speed command experiment, the peak speed command was 150 rpm with a period of 5 seconds, and the load torque was maintained at 5 N·m. Under this condition, the motor periodically switched between acceleration and deceleration states to examine the control method's ability to track continuously changing speed commands. Figure 12 The experimental waveforms are obtained under the condition of fixed nominal input channel gain. Figure 13 The waveforms are experimental waveforms obtained using the method of this invention.

[0152] Depend on Figure 12 (a) It can be seen that, under the condition of fixed nominal input channel gain, Keep it at zero. It fluctuates around zero. and It changes periodically with the sinusoidal speed command. (From...) Figure 12 (b) It can be seen that there is a significant phase lag between the speed feedback and the speed command. In the three typical cycles, the marked lag times are 0.5814s, 0.6465s and 0.5491s, respectively, with an average of about 0.5923s.

[0153] Depend on Figure 13 (a) As can be seen, after adopting the method of the present invention, the dq-axis current still maintains periodic adjustment with the sinusoidal speed command. Figure 13 (b) It can be seen that, within the three typical cycles, the lag time of the speed feedback relative to the speed command decreases to 0.5491s, 0.5168s, and 0.4522s, respectively, with an average of approximately 0.5060s. Compared to the fixed nominal input channel gain condition, the average phase lag is reduced. Figure 13 (c) It is evident that identifying magnetic chains Significant fluctuations occur during the periodic speed change process, while the fused magnetic flux... Maintain a relatively smooth, gradual change trajectory; input channel gain The flux linkage is continuously adjusted. The above results show that the method of the present invention can suppress the direct impact of transient fluctuations in the identification flux linkage on the input channel gain under periodic variable speed conditions, and improve the phase characteristics of speed tracking.

[0154] like Figure 14 and Figure 15 As shown, in the constant speed variable load test, the speed command was set to 450 rpm, and the load torque was increased from 5 N·m to 15 N·m to examine the dynamic disturbance rejection capability under the condition of sudden change in external load. Figure 14 The experimental waveforms are obtained under the condition of fixed nominal input channel gain. Figure 15 The waveforms are experimental waveforms obtained using the method of this invention.

[0155] Depend on Figure 14 (a) It can be seen that, under the condition of fixed nominal input channel gain, Keep it at zero. It fluctuates around zero; as the load increases... and It is then raised again to provide the additional electromagnetic torque required for the increased load. Figure 14 (b) It can be seen that the peak overshoot after the speed reaches near steady state during the startup phase is 133.791 rpm; after the load increases, the speed experiences a transient dip and enters a recovery process, with a speed dip of 25.38 rpm. This result indicates that, under the condition of a fixed nominal input channel gain, the dynamic response of the system is affected by the degree of matching between the nominal input channel gain and the actual motor object.

[0156] Depend on Figure 15 (a) As can be seen, after adopting the method of the present invention, the dq-axis current can be adjusted accordingly with changes in load. Figure 15 (b) It can be seen that the peak speed overshoot during the startup phase decreased to 111.253 rpm; the speed slump after the load increased decreased to 20.66 rpm, and the speed recovery process became more compact. Figure 15 (c) It can be seen that, during the load change process, the magnetic flux can be identified. There are certain transient fluctuations, and the fused magnetic flux... The input channel gain remains continuously variable. The flux linkage is gradually adjusted. The above results show that the method of the present invention can reduce the mismatch between the fixed nominal input channel gain and the actual motor object when external load disturbances occur, and improve the speed recovery process.

[0157] The experimental results of constant torque variable speed, sinusoidal speed command, and constant speed variable load demonstrate that the method of this invention can operate in actual controllers, three-phase inverters, and PMSM-assisted drag-and-drop platforms. Compared with the fixed nominal input channel gain control method, the method of this invention achieves fused flux linkage through MRAS flux linkage identification, error amplitude evaluation, gating fusion, safety interval construction, and closed-loop projection update. and input channel gain The system is gradually adjusted according to the operating conditions, thereby reducing the risk of sudden changes in input channel gain caused by transient fluctuations in flux linkage, and improving the speed tracking process, current regulation process, and dynamic recovery process after load changes.

[0158] The experimental results and simulation results corroborate each other, demonstrating that the method described in this invention can not only improve speed response and current regulation in a simulation environment, but also operate online in an experimental system composed of a real digital controller, a three-phase inverter, a permanent magnet synchronous motor, and a drag-load platform. Under experimental conditions such as constant torque speed change, sinusoidal speed command, and constant speed load change, when using a fixed nominal input channel gain, the dq-axis current exhibits significant fluctuations during speed changes or load disturbances, and the speed response is prone to backtracking, dips, or tracking lag. Using the method of this invention, the identified flux linkage is first obtained through gated fusion, and then the input channel gain is mapped from the fused flux linkage, allowing the input channel gain to be gradually adjusted according to the operating state. In the experimental waveforms, the fused flux linkage and input channel gain show a synchronous and smooth changing trend without significant jumps, indicating that the gated fusion, CBF / CLF constraint projection, and synchronous update process of the input channel gain can be executed within the actual control cycle. Therefore, this invention can reduce the risk of gain abrupt changes caused by the direct entry of transient flux fluctuations into the ADRC input channel in actual motor control platforms, and improve the smoothness of speed regulation and current response.

[0159] In one alternative implementation, the parameter tuning method is as follows.

[0160] In practical engineering applications, the gating sensitivity coefficient CLF contraction factor CBF barrier function coefficients and smoothing weights It can be adjusted according to the motor power rating, control cycle, allowable current surge and speed response requirements. The larger the value, the more sensitive the gated variable is to identification errors, and the more conservative the controller becomes. The smaller the value, the more the controller tends to quickly generate the recognition result.

[0161] Smoothing weights Used to balance the desired gating value for tracking and suppress gating jumps between adjacent cycles. If A larger value results in a smoother change in the gating variable, but a slower asymptotic correction of the input channel gain; if... A smaller value results in the gating variable approaching the desired value more quickly, but may lead to more significant gain adjustments when there are fluctuations in the identification error. In actual tuning, a more conservative value can be selected initially. Then, based on the q-axis current waveform and velocity response, the current is gradually reduced.

[0162] Gain safety boundary and The safety margin should be determined based on the motor's rated parameters, controller bandwidth, allowable current range, and system stability margin. An excessively narrow safety margin will limit the asymptotic correction effect of the input channel gain, while an excessively wide safety margin will weaken the protective effect of the CBF constraint. Preferably, the upper and lower limits can be set comprehensively based on the nominal gain, the experimental motor flux variation range, the allowable current range, and the system stability margin.

[0163] To illustrate the specific implementation conditions of the input channel gain asymptotic correction algorithm, the main algorithm parameters used in this embodiment are shown in Table 3. The parameters listed in Table 3 are values ​​for a specific embodiment, used to illustrate the algorithm implementation conditions of the present invention, and do not constitute a limitation on the scope of protection of the present invention.

[0164] Table 3. Main parameters of the input channel gain progressive correction algorithm:

[0165]

[0166] In the stages of low speed and weak back EMF, high sampling noise, or unstable current loop, the gated variable can be frozen or restricted to a small range; after the MRAS error amplitude enters a stable small range, the gated variable can be gradually increased. This freezing or gradual opening strategy is consistent with the ideas of bounded mapping, saturation constraints, and frozen gating, which helps to improve engineering feasibility.

[0167] Furthermore, the differences between this invention and conventional control methods are as follows:

[0168] Unlike the Fixed Nominal Input Channel Gain (ADRC), this invention does not simply change a constant in the original controller parameter table. Instead, it reconstructs the input channel gain based on the flux linkage identification state within each control cycle. The ADRC relies on the input channel gain given during the design phase. When the flux linkage of the permanent magnet or the equivalent moment of inertia of the system changes, a deviation may occur between the actual input channel and the nominal channel. When the load conditions change, this input channel mismatch will also increase the disturbance compensation burden of the ESO, which can only compensate for this deviation as part of the total disturbance. This invention, by fusing the physical mapping between flux linkage and input channel gain, enables the nominal input channel to be progressively calibrated with online flux linkage information, reducing the proportion of this deviation entering the total disturbance term.

[0169] Compared to direct injection of identification results, this invention adds an identification reliability evaluation and a gating fusion process. Direct injection, before identification convergence, easily maps transient fluctuations in the identification flux to abrupt changes in the input channel gain, further manifesting as abrupt changes in the q-axis current command and electromagnetic torque impact. This invention constructs the desired gating value using the MRAS error amplitude index and obtains the gating variable for the next control cycle through CBF, CLF, and closed projection, ensuring that the identification result must be jointly filtered by reliability and a safety interval before entering the control channel.

[0170] Compared to low-pass filtering, this invention not only considers signal smoothing but also explicitly considers the safety boundary and convergence target of the gating variable. Low-pass filtering can only smooth the identification flux in the frequency domain and cannot guarantee that the filtered gain will meet certain requirements. and Constraints alone cannot guarantee that the gated variable converges to the desired value at a preset rate. In this invention, the CBF constraint directly corresponds to the feasible boundary, the CLF constraint corresponds to the convergence rate, and the QP objective function corresponds to smooth updates. The three together determine the gate value for the next control cycle.

[0171] Unlike simple rate limiting, this invention does not impose a fixed upper limit on the input channel gain. Instead, it updates the feasible region in real time based on the current identified flux linkage, nominal flux linkage, preset input channel gain safety range, and desired gate value. Rate limiting thresholds typically rely on human experience, making it difficult to adapt to different flux linkage deviation directions and operating conditions. The feasible region of this invention changes in conjunction with the deviation between the identified and nominal flux linkages, the lower bound of the gate variable, the upper bound of the gate variable, and the desired gate value, thus achieving an adaptive trade-off between safety and speed.

[0172] Compared to typical CBF / CLF-QP safety control, the constraint object of this invention is not the motor control input itself, but rather the gated variable in the ADRC input channel gain update process. This design allows safety control theory to directly serve controller parameter updates, rather than replacing the original current or speed loop. Since the gated variable is a one-dimensional variable, this invention can simplify the QP problem into a closed-range projection, avoiding the difficulty of real-time implementation of multi-dimensional online optimization in motor controllers.

[0173] Compared to methods that add extra high-order observers or complex intelligent optimizers, this invention retains the original ADRC structure and MRAS identification framework, adding only gating fusion and constraint projection stages before the identification results enter the control channel. Therefore, this method requires minimal modification to existing PMSM control systems and can be easily implemented through software upgrades in existing FOC, SVPWM, and speed loop ADRC programs.

[0174] In a typical control cycle, the execution sequence of this invention is as follows:

[0175] In a typical control cycle, the controller first performs synchronous sampling with the ADC to obtain the phase current or dq-axis current feedback values, and then reads the rotational speed information output by the encoder or speed estimation module. Subsequently, it performs coordinate transformation and current feedback updates, ensuring that the MRAS reference model and the adjustable model receive the voltage, current, and rotational speed inputs at the same sampling time. This step guarantees that the data used for flux linkage identification is consistent in time with the feedback quantities used for speed loop control.

[0176] Subsequently, the MRAS flux linkage identification unit calculates the identified flux linkage and flux linkage error, and sends the error amplitude to the low-pass smoothing stage to obtain the MRAS error amplitude index. The gate value generation module generates the desired gate value according to equation (13). Based on this, the safety constraint projection module... and In addition, the current identified flux linkage and nominal flux linkage are used to calculate the safe interval of the gate variable. The CLF unit calculates the convergence interval based on the expected gate value and the gate variable of the previous cycle.

[0177] Next, the closed projection loop in the safety constraint projection module calculates the intersection of the two feasible intervals and calculates the gate variable for the next control cycle according to equations (22) and (23). If the intersection is valid, the projection result is adopted; if an abnormality is detected in current, bus voltage, encoder, or temperature, the gate variable can be frozen and restored to the nominal gain to maintain the safety of the drive system. This freezing method does not change the core calculation logic of the present invention, but only suspends adaptive updates in the protection state.

[0178] Finally, the gating fusion stage in the gating fusion and gain mapping module calculates the fusion flux using the gating variables of the next control cycle. The active disturbance rejection gain synchronization update module calculates the input channel gain according to equation (7) and writes the input channel gain into the speed loop ADRC control law and the ESO input channel. After the speed loop outputs the q-axis current command, the inverter switching signal is generated by the current loop and the PWM modulation unit. The entire execution link makes the gain correction result effective in subsequent control updates, thereby realizing the closed-loop connection between identification, correction and drive execution.

[0179] The PMSM active disturbance rejection gain progressive correction method, apparatus and system provided by this invention address the problems of fixed nominal input channel gain being difficult to adapt to flux changes and online identification of flux direct injection easily causing current surges. It establishes a complete technical link from MRAS flux identification to ADRC input channel gain safe progressive update.

[0180] This method is based on the gating fusion of identified flux linkage and nominal flux linkage. It evaluates the reliability of identification using the MRAS error amplitude index, constructs a gated safety interval using a preset input channel gain safety range, ensures convergence using CLF, guarantees safety using CBF, and generates the gated variables for the next control cycle through one-dimensional QP closed projection. This scheme has both clear physical meaning and the conditions for embedded real-time implementation.

[0181] Simulation and experimental results show that this invention can improve the dynamic response of the speed loop, reduce q-axis current oscillation, shorten the recovery process after load abrupt changes, and reduce tracking phase lag under sinusoidal speed commands. Therefore, this invention can be used for online input channel gain correction in motor drive systems employing PMSM speed loop ADRC.

[0182] The embodiments described above are for illustrating the technical solutions of the present invention, and not for limiting the scope of protection of the present invention. Without departing from the technical concept of the present invention, those skilled in the art can make adaptive adjustments to the relevant parameters, sampling methods, modulation methods, and execution cycles. The scope of protection of the present invention is defined by the claims.

Claims

1. A device for adaptive gain tuning of active disturbance rejection control for permanent magnet synchronous motor, characterized in that: The gain progressive correction device includes: The model reference adaptive system flux linkage identification module is used to obtain the identification flux linkage and the model reference adaptive system flux linkage identification error based on the current, voltage, speed of the permanent magnet synchronous motor and the control input signal for voltage reconstruction. The control input signal includes the PWM duty cycle or the inverter switching state. The gating value generation module is used to generate a desired gating value based on the model reference adaptive system error amplitude index obtained by amplitude calculation and smoothing of the model reference adaptive system flux linkage identification error. The gated fusion and gain mapping module is used to perform gated fusion on the identified flux linkage and the nominal flux linkage according to the gated variable to obtain the fused flux linkage, and to map the fused flux linkage as the gain of the active disturbance rejection control input channel. The safety constraint projection module is used to convert the safety range of the active disturbance rejection control input channel gain into a fused flux linkage safety interval, and to obtain the gated variable safety interval based on the fused flux linkage safety interval and the gated fusion relationship between the identified flux linkage and the nominal flux linkage; it is also used to generate a joint feasible interval of gated variables based on the safety constraints of the control barrier function and the convergence constraints of the control Lyapunov function, and to perform a closed projection of the unconstrained minimum point within the joint feasible interval of gated variables to obtain the gated variables for the next control cycle; The active disturbance rejection gain synchronization update module is used to recalculate the fused flux linkage based on the gated variables of the next control cycle, and to synchronously apply the same input channel gain determined by the fused flux linkage to the input channel gain term in the velocity loop active disturbance rejection control law and the input gain term of the extended state observer.

2. The progressive correction device for the self-disturbance rejection gain of a permanent magnet synchronous motor according to claim 1, characterized in that: The gating value generation module determines the expected gating value according to a model reference adaptive system error amplitude index, the model reference adaptive system error amplitude index is obtained by amplitude calculation and smoothing processing of the model reference adaptive system flux linkage recognition error, and the expected gating value is represented as: ； In the formula, is the model reference adaptive system error amplitude index, , is the gate sensitivity coefficient, which is used to adjust the sensitivity of the expected gate value to the identification error.

3. The device for gain progressive correction of active disturbance rejection in permanent magnet synchronous motor according to claim 1, characterized in that: The gated fusion and gain mapping module generates a fused magnetic link based on the identified magnetic link, the nominal magnetic link, and the gated variable. The fused magnetic link is represented as follows: ； wherein is the flux linkage value after fusion at the moment, is the gating variable at the moment, is the identified flux linkage value at the moment, is the nominal flux linkage value, and the input channel gain is expressed as: ； wherein is the number of motor pole pairs, is the system equivalent moment of inertia, is the active disturbance rejection control input channel gain.

4. The progressive gain correction device for permanent magnet synchronous motor self-disturbance rejection according to claim 1, characterized in that: The safety constraint projection module converts the input channel gain safety range into a fused flux linkage safety interval, and obtains the gated variable safety interval based on the fused flux linkage safety interval, the identified flux linkage, the nominal flux linkage, and the gated fusion relationship. When identifying the deviation between the magnetic flux and the nominal magnetic flux At that time, among them This indicates the deviation between the identified flux linkage and the nominal flux linkage. To preset the flux linkage deviation threshold, the gated variable safety interval is set to [0,1]. When the deviation between the identified flux linkage and the nominal flux linkage is greater than the preset flux linkage deviation threshold, the safety constraint projection module calculates the gated variable boundary candidate values ​​according to the upper and lower boundaries of the fused flux linkage safety interval, and intersects the boundary candidate values ​​with [0,1] to obtain the gated variable safety interval. The safety constraint projection module intersects the safe feasible interval corresponding to the safety constraint of the control barrier function with the convergent feasible interval corresponding to the convergent constraint of the control Lyapunov function to obtain the final gated feasible interval, which is expressed as follows: ； In the formula, For the first The final gated feasible interval at each sampling time. This is the lower bound of the final gated feasible interval. This is the upper bound of the final gated feasible interval. For the first The convergence feasible interval at each sampling time is determined by the convergence constraint of the control Lyapunov function. For the first The safe and feasible interval is determined by the safety constraints of the control obstacle function at each sampling time. When the intersection of the safe and feasible interval and the convergent feasible interval is empty, the safety constraint projection module will preferentially project the unconstrained minimum point to the safe and feasible interval, or maintain the safety gate variable value of the previous control cycle. The unconstrained minimum point is the weighted average of the expected gate value and the current gate variable under the influence of the smoothing weighting coefficient, expressed as: ； In the formula, ≥0 indicates a smoothing weighting coefficient. This represents the value of the gate variable that minimizes the objective function under unconstrained conditions. For the first The expected gate value at each sampling time. For the first The gated variable values ​​at each sampling time are used to project the unconstrained minimum point onto the final gated feasible interval by the safety constraint projection module, thus obtaining the gated variable for the next control cycle, expressed as: ； In the formula, For interval projection operators, This is the value of the gate variable for the next control cycle.

5. A method for progressively correcting the self-disturbance rejection gain of a permanent magnet synchronous motor using the progressive correction device for self-disturbance rejection gain of a permanent magnet synchronous motor as described in claim 1, characterized in that: The method includes: The current, voltage, speed, and control input signals for voltage reconstruction of the permanent magnet synchronous motor are acquired. The control input signals include the PWM duty cycle or the inverter switching state. The identified flux and the flux identification error of the model reference adaptive system are obtained through model reference adaptive flux identification. The flux linkage identification error of the model reference adaptive system is calculated and smoothed to obtain the error amplitude index of the model reference adaptive system, and the desired gate value is determined based on the error amplitude index of the model reference adaptive system. The safe range of the active disturbance rejection control input channel gain is converted into the safe range of the fused flux linkage, and the safe range of the gated variable is determined by inverse solution based on the deviation between the identified flux linkage and the nominal flux linkage and the gating fusion relationship. Generate the joint feasible interval of the gated variables based on the safety constraints of the control barrier function and the convergence constraints of the control Lyapunov function; The unconstrained minimum point is projected in a closed loop within the joint feasible interval of the gated variables to obtain the gated variables for the next control cycle. A fused flux linkage is generated based on the gating variables, identified flux linkage, and nominal flux linkage of the next control cycle, and the input channel gain is determined by the fused flux linkage. The input channel gain is synchronously applied to the input channel of the velocity loop active disturbance rejection control law and the extended state observer.

6. A progressive gain correction system for a permanent magnet synchronous motor with self-disturbance rejection, characterized in that: The system includes a permanent magnet synchronous motor, a three-phase inverter, a current sampling unit, a voltage sampling unit, a position and speed feedback unit, a PWM drive / modulation unit, and a digital controller. The digital controller is equipped with the permanent magnet synchronous motor self-disturbance rejection gain progressive correction device as described in claim 1; The digital controller is used to perform flux linkage identification, gate value generation, gate fusion and gain mapping, safety constraint projection and active disturbance rejection gain synchronization update of the model reference adaptive system, and to perform permanent magnet synchronous motor speed loop control based on the updated input channel gain.

Images