Adaptive vector control method and system for PMSM based on double-constraint variable forgetting factor

Abstract

The application discloses a PMSM adaptive vector control method and system based on double-constraint variable forgetting factors. The method comprises the following steps: collecting motor position signals and three-phase current signals to obtain a mechanical angular velocity; sending a given value of a cross-axis current output by a speed loop to a current loop for closed-loop control; performing online identification of a rotational inertia and a viscous friction coefficient based on a double-constraint model variable forgetting factor recursive least square algorithm; and calculating real-time proportional gain and integral gain of the speed loop based on a closed-loop damping ratio and identification results and updating control parameters. The application can solve the contradiction between load identification convergence speed and steady-state accuracy, eliminate system overshoot and oscillation caused by sudden change of the inertia of the mechanical arm, and ensure high-precision and high-stability operation of overall control.

Description

Technical Field

[0001] This application relates to the field of motor vector control technology, specifically to a PMSM adaptive vector control method and system based on a dual-constraint variable forgetting factor. This application belongs to the IPC classification system H02P21 / 00. Background Technology

[0002] In the fields of industrial automation and intelligent manufacturing, when industrial collaborative robotic arms perform picking and placing tasks, the permanent magnet synchronous motors at their joints face frequent changes in rotational inertia and load torque. Current robotic arm servo systems typically employ a dual-loop or triple-loop structure based on field-oriented control, with the speed loop usually using a fixed-parameter proportional-integral (PI) controller. While the fixed-parameter PI controller can remain stable under rated operating conditions, when the robotic arm's posture changes significantly or when grasping heavy objects causes an increase in the overall rotational inertia, the actual bandwidth of the entire control system decreases. This can easily lead to speed response lag and position overshoot, affecting the robotic arm's working accuracy and safety.

[0003] To address the limitations of fixed-parameter controllers, existing adaptive control schemes often employ conventional recursive least squares algorithms for online identification of rotational inertia, adjusting controller parameters in real time accordingly. Conventional recursive least squares algorithms typically use a fixed forgetting factor. When this fixed forgetting factor approaches its upper limit, the algorithm's strong memory of old data leads to convergence times exceeding 200 milliseconds under sudden load changes. This delay cannot meet the 1-millisecond dynamic high-frequency response requirements of robotic arm joint control. When the forgetting factor is set too low, the system is prone to data saturation during steady-state operation, with a surge in the covariance matrix. Data saturation directly causes oscillations in the identified parameters, amplifies electromagnetic torque pulsations in the motor output, excites resonance in the robotic arm's mechanical body, and reduces system lifespan and operational reliability. Currently, it is difficult to effectively balance the contradiction between the dynamic convergence speed and steady-state identification accuracy of the load identification algorithm. Summary of the Invention

[0004] This application provides a PMSM adaptive vector control method and system based on a dual-constraint variable forgetting factor, which solves the technical problems of slow convergence and easy data saturation in steady state of conventional recursive least squares algorithms in the prior art.

[0005] This application provides an adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor, comprising: acquiring the rotor position signal and three-phase current signal of the target permanent magnet synchronous motor, performing a space vector coordinate transformation, and obtaining the quadrature-axis actual current, direct-axis actual current, and mechanical angular velocity. Obtain the quadrature-axis current setpoint output from the speed loop of the control system. Set the quadrature axis current value The data is sent to the current loop for closed-loop control, driving the permanent magnet synchronous motor. The current quadrature-axis current setpoint is acquired. The actual electromagnetic torque of the system and mechanical angular velocity Moment of inertia is calculated using a recursive least squares algorithm with a variable forgetting factor. With viscous friction coefficient Online identification. Among them, the variable forgetting factor. The calculation is based on a dual-constraint model, which includes a dynamic term based on prior estimation error and a term based on the quadrature-axis current setpoint. The abrupt feedforward term of the time derivative is configured to control the variable forgetting factor in the initial stage of a speed drop caused by a sudden load increase in the permanent magnet synchronous motor. It decreases exponentially. This is based on the system's preset closed-loop damping ratio and the moment of inertia identified during the current control cycle. and viscous friction coefficient Calculate the real-time proportional gain and integral gain of the speed loop, and update the control parameters of the speed loop within the current control cycle.

[0006] In a preferred embodiment, the dual-constraint model calculates the variable forgetting factor. The process includes: calculating the actual electromagnetic torque The corresponding prior estimation error. Based on the preset initial forgetting factor, the absolute value of the prior estimation error, and the first gain coefficient, the value of the dynamic term is determined. This dynamic term is negatively correlated with the absolute value of the prior estimation error. For the given value of the quadrature-axis current... A first-order backward difference operation is performed to obtain the time derivative, and the absolute value of the time derivative is checked to see if it exceeds a preset derivative threshold. If the absolute value of the time derivative exceeds the threshold, the product of the time derivative and the second gain coefficient is input into a preset natural exponential function for calculation to obtain the decay coefficient of the abrupt feedforward term. The value of the dynamic term is multiplied by the decay coefficient to update and obtain the variable forgetting factor at the current time step. .

[0007] In a preferred embodiment, the moment of inertia is calculated using a recursive least squares algorithm based on a variable forgetting factor. With viscous friction coefficient The online identification process includes: obtaining the forgetting factor at the current moment based on the update. The system covariance matrix at the previous time step and the given quadrature-axis current value at the current time step. With mechanical angular velocity Given the data input vector, calculate the algorithm gain matrix at the current time step. Based on the system covariance matrix and variable forgetting factor from the previous time step... The algorithm gain matrix and data input vector are used to update the system covariance matrix at the current time step through matrix operations. Combining the parameter vector from the previous time step, the algorithm gain matrix, and prior estimation errors, the system iteratively obtains the moment of inertia identified at the current time step. and viscous friction coefficient The parameter vector.

[0008] In a preferred embodiment, the process of calculating the real-time proportional gain and integral gain of the velocity loop includes: extracting the pre-configured velocity loop pole parameters and closed-loop damping ratio of the control system; and calculating the velocity loop pole parameters, closed-loop damping ratio, and the identified moment of inertia. Double the product to determine the intermediate term, and then combine this intermediate term with the identified coefficient of viscous friction. The difference is calculated to obtain the real-time proportional gain. The squared values ​​of the velocity loop pole parameters are calculated, and this squared value is compared with the identified moment of inertia. Multiply to obtain the real-time integral gain.

[0009] In a preferred embodiment, before updating the control parameters of the speed loop within the current control cycle, the method further includes: determining whether the calculated real-time proportional gain is within a set proportional gain safety range, and determining whether the real-time integral gain is within a set integral gain safety range. In response to the real-time proportional gain or real-time integral gain exceeding the corresponding safety range, the exceeding gain parameter is clamped and limited to the boundary value of the corresponding safety range, and the control parameters of the speed loop are updated using the clamped and limited parameter.

[0010] In a preferred embodiment, the above control method is executed on a heterogeneous system-on-a-chip architecture including a field-programmable gate array (FPGA) and a microprocessor. It includes: running data acquisition logic and pulse width modulation logic in the FPGA to perform operations such as acquiring rotor position signals and three-phase current signals, performing spatial vector coordinate transformation, and driving the permanent magnet synchronous motor; and performing floating-point calculations in the microprocessor, including online identification and calculation of the real-time proportional gain and integral gain of the speed loop.

[0011] In a preferred embodiment, the control method further includes: after calculating the real-time proportional gain and integral gain within the microprocessor, packaging the updated real-time proportional gain and integral gain into control frame data. The control frame data is then sent from the microprocessor to a parameter control register allocated within the field-programmable gate array via an on-chip high-speed bus, overwriting the gain data from the previous control cycle.

[0012] This application also provides an adaptive vector control system applied to a robotic arm joint module. The adaptive vector control system includes a sensor module 201, a control processing module 202, and a control processing module 203. The sensor module 201 is configured to acquire the rotor position signal and three-phase current signal of the target permanent magnet synchronous motor. The control processing module 202 is configured to output the quadrature-axis current setpoint from the speed loop of the control system. The data is sent to the current loop for closed-loop control, driving the permanent magnet synchronous motor. The control processing module 203 is communicatively connected to the sensor module 201 and the control processing module 202. The control processing module 203 is configured to perform spatial vector coordinate transformation to obtain the actual current along the quadrature axis, the actual current along the direct axis, and the mechanical angular velocity. Collect quadrature-axis current setpoint The actual electromagnetic torque of the system and mechanical angular velocity Moment of inertia is calculated using a recursive least squares algorithm with a variable forgetting factor. With viscous friction coefficient The system performs online identification and uses the system's preset closed-loop damping ratio and the moment of inertia identified in the current control cycle. and viscous friction coefficient Calculate the real-time proportional gain and integral gain of the speed loop and update the control parameters. Among these, the variable forgetting factor... The calculation is based on a dual-constraint model, which includes a dynamic term based on prior estimation error and a term based on the quadrature-axis current setpoint. The abrupt feedforward term of the time derivative is configured to control the variable forgetting factor in the initial stage of a speed drop caused by a sudden load increase in the permanent magnet synchronous motor. It shows an exponential decline.

[0013] In a preferred embodiment, the control processing module 203 includes a system-on-a-chip (SoC) processor chip 301. The SoC processor chip 301 integrates a hardware programmable logic array region 302 and a hard-core microprocessor region 303 on the same silicon substrate. The hardware programmable logic array region 302 is configured with a high-speed input / output interface, connected to the sensor module 201 to continuously acquire current and position data. The hard-core microprocessor region 303 is configured to execute instruction code based on a recursive least squares algorithm with a variable forgetting factor and parameter updates.

[0014] This application also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by the processor of the control system, it implements the above-described adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor.

[0015] Based on the above technical solution, this application constructs a recursive least squares model of the variable forgetting factor driven by the time derivative of the quadrature-axis current setpoint and the state error, and maps the identification results to the real-time parameters of the speed controller. The quadrature-axis current setpoint, as a sudden change feedforward term, directly participates in the dynamic adjustment of the variable forgetting factor, providing a forward-looking prediction condition that transcends the physical mechanical motion delay for the system's parameter identification process. In the initial expansion stage when the robot arm's speed drops due to a sudden increase in load, the system directly forces the variable forgetting factor to decrease exponentially by monitoring the time derivative, quickly resetting the algorithm's covariance matrix, driving the identification parameter vector to approach the new state, resolving the contradiction between the load identification convergence speed and steady-state accuracy, eliminating system overshoot and oscillation caused by sudden changes in robot arm inertia, and ensuring high-precision and high-stability operation of the overall control. Attached Figure Description

[0016] Figure 1 This is a structural flowchart provided in an embodiment of the present invention.

[0017] Figure 2 This is a structural flowchart provided in an embodiment of the present invention.

[0018] Figure 3 This is a structural flowchart provided in an embodiment of the present invention. Detailed Implementation

[0019] The specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0020] Example 1

[0021] like Figure 1 As shown, this embodiment provides an adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor. This method runs on a heterogeneous microprocessor computing platform and achieves efficient control of the joint drive motors of industrial robotic arms by decoupling and jointly scheduling high-frequency low-level physical control with complex upper-level adaptive algorithms. The specific execution flow of this method includes the following:

[0022] In step S101, the control system acquires the rotor position signal and three-phase current signal of the target permanent magnet synchronous motor through the underlying data channel, and performs spatial vector coordinate transformation in a nanosecond-level execution environment to obtain the actual quadrature axis current, the actual direct axis current, and the mechanical angular velocity. For industrial robotic arm environments, the control system continuously reads serial position data from an absolute encoder mounted at the tail of the permanent magnet synchronous motor to obtain a high-precision rotor position signal. Simultaneously, it utilizes a high-precision current sensor positioned on the inverter bridge to synchronously capture three-phase analog sampling signals from the stator windings. The system performs standard Clark and Park transformations to map the stationary three-phase coordinate system parameters to an orthogonal two-phase coordinate system that rotates synchronously with the rotor, thus eliminating the influence of AC components on the control system. The aforementioned coordinate transformation and encoder acquisition methods are merely preferred solutions; those skilled in the art can also achieve the signal acquisition function of this application using incremental encoders or sensorless observation algorithms for position acquisition.

[0023] In step S102, the control system acquires the quadrature-axis current setpoint output by the speed loop. The quadrature axis current setpoint The data is sent to the current loop for closed-loop control, driving the permanent magnet synchronous motor. Based on the fundamental control topology, the system uses rotor field-oriented control as its foundation. The current loop employs fixed-parameter proportional-integral (PI) control logic to ensure high-frequency and rapid response of the internal electromagnetic torque output. The external speed loop uses variable-parameter PI control logic, with its control output serving as the target value for the internal current closed loop.

[0024] In step S103, the control system constructs a load identification channel and acquires the current quadrature-axis current setpoint. The actual electromagnetic torque of the system and mechanical angular velocity Moment of inertia is calculated using a recursive least squares algorithm with a variable forgetting factor. With viscous friction coefficient Online identification. The core physical basis of the load identification module is the discretized mechanical motion equations of the motor. Within any sampling period, the balance between electromagnetic torque, load torque, and the system's mechanical dynamic response can be expressed by the following formula: the difference between the actual electromagnetic torque and the external load torque is equal to the product of the quotient of the difference between the moment of inertia and the mechanical angular velocity, plus the product of the viscous friction coefficient and the mechanical angular velocity.

[0025]

[0026] in, This represents the actual electromagnetic torque calculated during the current control cycle; This represents the external load torque that the system is currently experiencing. Indicates the moment of inertia; This represents the mechanical angular velocity at the current moment; This represents the mechanical angular velocity of the previous control cycle; Indicates the discrete sampling period of the control system; This represents the coefficient of viscous friction during rotor operation. In this embodiment, It is set to 125 microseconds to match the pulse width modulation frequency of the hardware control layer.

[0027] External load torque during robotic arm operation Often containing unknown constant components, directly using the aforementioned physical equations for parameter identification would introduce DC deviation. The system uses differential transformation to extract the difference sequence of the control cycle to eliminate interference from constant load torque. Based on this, a standard linear regression model is established for recursive calculations. The model's operational logic is expressed by the following formula, where the observed measurements are calculated by multiplying the transpose of the data vector with the vector of parameters to be identified:

[0028]

[0029] Among them, the observations of the model The torque change between two adjacent sampling periods is expressed as follows: .

[0030] Data vectors in the model The data vector, which combines the second and first derivative characteristics of the system velocity, is represented by a column vector consisting of the second and first differences of the mechanical angular velocities. .

[0031] The parameter vector to be identified in the model It includes two core inertial indices currently dynamically acquired by the system. The parameter vector to be identified is represented by a column vector consisting of the estimated values ​​of the moment of inertia and the coefficient of viscous friction. .in, This is an estimate of the moment of inertia. This is an estimated value for the viscous friction coefficient.

[0032] The recursive least squares algorithm is endowed with adaptive tracking capability for handling time-varying loads, and a variable forgetting factor is set in the system. The calculation is based on a dual-constraint model. This dual-constraint model performs parallel calculations based on the dynamic term of the prior estimation error and the given value of the quadrature-axis current. The abrupt change feedforward term of the time derivative. The abrupt change feedforward term is specifically configured to: control the forgetting factor in the initial stage when a sudden increase in external load on the permanent magnet synchronous motor causes a slight drop in rotor speed. It exhibits an exponential decreasing trend. In a preferred embodiment, the logic for calculating the forgetting factor using the dual-constraint model is implemented through the following comprehensive formula: the variable forgetting factor is calculated by adding the difference between the upper and lower limits of the steady-state accuracy to the lower limit of the lower limit, and then multiplying it by the decay term of the natural exponential function. The decay term consists of the absolute value of the prior estimation error and the absolute value of the time derivative.

[0033]

[0034] in, This represents the variable forgetting factor for the current control cycle. The lower limit protection parameter for the forgetting factor is set to 0.900 in this embodiment to prevent the covariance matrix from diverging out of control. This represents the upper limit of the steady-state accuracy forgetting factor, which is set to 0.995 to ensure extremely high recognition accuracy of the system under static or uniform speed conditions without sudden changes, and to prevent data drift. Represents actual electromagnetic torque The corresponding prior estimation error is the difference between the model prediction at the previous moment and the current actual observation. This represents the error sensitivity coefficient, which is set to a value range of 10 to 50, and determines the system's response magnitude to normal state errors. This represents the torque derivative sensitivity coefficient, with a value range of 0.1 to 0.5. The time derivative of the quadrature-axis current setpoint is obtained by performing a first-order backward difference operation on the quadrature-axis current setpoint output by the speed loop. The specific value range of the above parameters is an optimized range derived from extensive simulations and experiments. Those skilled in the art can adjust it proportionally according to the rated power of the motor.

[0035] Based on the obtained forgetting factor The system updates the algorithm gain matrix at the current time step using a recursive matrix formula. This gain matrix is ​​then iteratively calculated and combined with the old system covariance matrix to reconstruct the current system covariance matrix carrying the new weight distribution. Combining the parameter vector from the previous time step, the new algorithm gain matrix, and the calculated prior estimation error, the system derives a new moment of inertia containing the optimal identification result for the current control cycle. With viscous friction coefficient The parameter vector.

[0036] In step S104, the rotational inertia is identified based on the system's preset closed-loop damping ratio and the current control cycle. and viscous friction coefficient The system calculates the real-time proportional gain and integral gain of the speed loop and updates the speed loop control parameters within the current control cycle. The system uses the pole placement principle to map the identification results output in step S103 to the controller. The system extracts the pre-configured desired speed loop pole parameters. and closed-loop damping ratio Real-time proportional gain and real-time integral gain The operational logic is calculated using the following formulas: the system calculates the desired closed-loop natural frequency, twice the product of the damping ratio and the moment of inertia, and subtracts the viscous friction coefficient to obtain the real-time proportional gain; the system also calculates the product of the square of the desired closed-loop natural frequency and the moment of inertia to obtain the real-time integral gain.

[0037]

[0038]

[0039] in, This represents the proportional gain value calculated and issued in the k-th control cycle; This represents the corresponding integral gain value; ξ is the damping ratio target set in the undamped oscillation control of the system. The configured pole bandwidth parameters; and This is the identification result output in step S103.

[0040] To ensure the safety of the control system during high-frequency dynamic updates, this embodiment also includes forced clamping logic. After calculating the two gain parameters mentioned above, the system determines whether the real-time proportional gain is within the set proportional gain safety range and whether the real-time integral gain is within the integral gain safety range. If either gain parameter exceeds the range due to special noise interference, the system will trigger an abnormal response, forcibly clamping and limiting the excess gain parameter to the upper or lower boundary value of the corresponding safety range. After ensuring safety, the system updates the internal operating variables of the speed loop using the final parameters.

[0041] By constructing a dual-constraint control logic driven by the derivative of the quadrature-axis current setpoint and the state error, this embodiment can utilize the responsiveness of the quadrature-axis current setpoint to the system's electromagnetic torque demand. When the load suddenly increases, the step change in the current setpoint directly triggers an exponential reset of the forgetting factor, compressing the parameter convergence time under sudden changes in moment of inertia to within 5 steps of the control cycle. Proactive parameter adaptation eliminates the dead zone of acquiring new inertia, strictly controlling the system's speed overshoot within 2% of the given target value, achieving a balance between speed and stability.

[0042] Example 2

[0043] Based on the core control method described in Embodiment 1 above, this embodiment provides an adaptive vector control system. For example... Figure 2 As shown, this adaptive vector control system is applied to the robotic arm joint module in industrial automation scenarios. The system mainly includes a sensor module 201, a control processing module 202, and a control processing module 203 at the physical level.

[0044] Sensor module 201 is configured via hardwired connection to acquire rotor position signals and three-phase current signals of the target permanent magnet synchronous motor. Control processing module 202 is configured to process the quadrature-axis current setpoint output from the speed loop of the control system. The current loop at the bottom layer performs closed-loop control calculations and uses power devices to drive the permanent magnet synchronous motor. The control processing module 203 communicates with the sensor module 201 and the control processing module 202. It is the core of the control system's computing power and is responsible for tasks such as spatial vector coordinate transformation, parameter acquisition, dual-constraint model calculation, online recursive identification, and gain mapping update.

[0045] like Figure 3 As shown, to meet the technical requirements of high-frequency updates and complex floating-point operations proposed in this application, the control processing module 203 is implemented based on a heterogeneous microprocessor architecture. The hardware core is built on a Zynq-7020 system-on-a-chip processor chip 301. This system-on-a-chip processor chip 301 integrates a hardware programmable logic array region 302 and a hard-core microprocessor region 303 on a unified silicon substrate.

[0046] The hardware programmable logic array region 302 is internally configured with customized high-speed input / output interfaces, enabling continuous acquisition of analog sampling signals from the sensor module 201. The hardware programmable logic array region 302 is primarily defined as the underlying hardware acceleration layer, responsible for executing parallel deterministic logic within nanosecond cycles, including acquiring absolute encoder position signals, coordinate transformation logic, and generating pulse-width modulation signals used to drive the inverter bridge.

[0047] The hard-core microprocessor region 303 is defined as the upper-level algorithm computation layer, configured to periodically read the state variables in the registers and use the built-in floating-point unit to perform complex iterative matrix calculations for the recursive least squares algorithm with a variable forgetting factor. Before the end of each 125-microsecond control cycle, the hard-core microprocessor region 303 internally performs real-time proportional gain calculations. With integral gain After the calculation, the floating-point data is packaged to generate control frame data that conforms to the on-chip communication protocol. Through the high-speed parallel bus network built into the chip, the control frame data is sent and written to the dedicated parameter control register inside the hardware programmable logic array area 302, directly overwriting the old gain data left over from the previous control cycle, thus completing the adaptive parameter delivery between the two-layer architecture.

[0048] As support for the operation of the hardware entity, the system hardware structure of this embodiment also includes a series of key peripheral driving and power supply circuits. The system includes a power supply unit 304 electrically connected to the on-chip system processor chip 301. This power supply unit 304 includes multiple voltage regulator modules such as 5V and 3.3V, providing low-ripple operating power for the control digital chip and analog amplification channels. The system is equipped with an external clock oscillation circuit 305, providing a reference 100 MHz operating clock source for the chip. An inverter bridge gate drive circuit 306 is connected in parallel to the output of the control processing module 202. This circuit amplifies and boosts the weak-level pulse signal emitted by the processor through an optocoupler isolation mechanism, converting it into a high-energy drive signal capable of directly controlling the on / off state of insulated-gate bipolar transistors or silicon carbide elements. The specific chip selection and power supply specifications are examples of a physical solution. Those skilled in the art can use an equivalent computing architecture to replace the monolithic heterogeneous solution to achieve the same system control effect.

[0049] The construction of a heterogeneous hardware-software integrated structure provides sufficient floating-point computing power resources for the variable forgetting factor least squares method, isolates the interference of complex algorithm calculation process on the stability of the underlying high-frequency current loop control cycle, and realizes high efficiency and robustness at the control system architecture level.

[0050] Example 3

[0051] This application provides a computer-readable storage medium storing a computer program. When the computer program is read and executed by the central processing unit of an industrial servo control system, it implements the adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor described in Embodiment 1. The computer-readable storage medium can be a non-volatile storage device, such as a read-only memory, random access memory, electrically erasable programmable read-only memory, flash memory chip, or solid-state memory chip soldered onto an industrial control board.

[0052] All readily conceivable variations or substitutions described above should be included within the scope of protection of this application. The scope of protection of this application shall be determined by the scope of the claims.

Claims

1. An adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor, characterized in that, include: The rotor position signal and three-phase current signal of the target permanent magnet synchronous motor are acquired, and a space vector coordinate transformation is performed to obtain the actual quadrature axis current, the actual direct axis current, and the mechanical angular velocity. ; The actual electromagnetic torque of the system is calculated based on the actual cross-axis current. ; Obtain the quadrature-axis current setpoint from the speed loop output of the control system. Set the quadrature axis current value The signal is sent to the current loop for closed-loop control, driving the permanent magnet synchronous motor to run; Collect the current quadrature-axis current setpoint. The actual electromagnetic torque of the system and mechanical angular velocity Moment of inertia is calculated using a recursive least squares algorithm with a variable forgetting factor. With viscous friction coefficient Online identification; among which, the variable forgetting factor The calculation is based on a dual-constraint model, which includes a dynamic term based on prior estimation error and a term based on the quadrature-axis current setpoint. The abrupt feedforward term of the time derivative is configured to control the forgetting factor during the initial stage of a speed drop caused by a sudden increase in load on the permanent magnet synchronous motor. It declined exponentially; Based on the system's preset closed-loop damping ratio and the moment of inertia identified in the current control cycle. and viscous friction coefficient Calculate the real-time proportional gain and integral gain of the speed loop, and update the control parameters of the speed loop within the current control cycle.

2. The adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor as described in claim 1, characterized in that, Calculating the variable forgetting factor using a dual-constraint model The process includes: Calculate the actual electromagnetic torque The corresponding prior estimation error; The value of the dynamic term is determined based on the preset initial forgetting factor, the absolute value of the prior estimation error, and the first gain coefficient. The value of the dynamic term is negatively correlated with the absolute value of the prior estimation error. For quadrature axis current setpoint Perform a first-order backward difference operation to obtain the time derivative, and determine whether the absolute value of the time derivative is greater than a preset derivative threshold. In response to the absolute value of the time derivative being greater than the derivative threshold, the product of the time derivative and the second gain coefficient is input into a preset natural exponential function for calculation to obtain the attenuation coefficient of the abrupt feedforward term; The dynamic term is multiplied by the decay coefficient to update and obtain the variable forgetting factor at the current time step. .

3. The adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor as described in claim 2, characterized in that, Rotational inertia is calculated using a recursive least squares algorithm with a variable forgetting factor. With viscous friction coefficient The online identification process includes: Obtain the current forgetting factor based on the update. The system covariance matrix at the previous time step and the given quadrature-axis current value at the current time step. With mechanical angular velocity Given the data input vector, calculate the algorithm gain matrix at the current time step; Based on the system covariance matrix and the variable forgetting factor from the previous time step The algorithm gain matrix and data input vector are used to update the system covariance matrix at the current time through matrix operations; By combining the parameter vector from the previous time step, the algorithm gain matrix, and the prior estimation error, the moment of inertia identified at the current time step is iteratively obtained. and viscous friction coefficient The parameter vector.

4. The adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor as described in claim 1, characterized in that, The process of calculating the real-time proportional gain and integral gain of the velocity loop includes: Extract the pre-configured velocity loop pole parameters and closed-loop damping ratio of the control system; Calculate the velocity loop pole parameters, closed-loop damping ratio, and identified moment of inertia. Double the product to determine the intermediate term, and then combine the intermediate term with the identified coefficient of viscous friction. Difference is calculated to obtain the real-time proportional gain; Calculate the squared values ​​of the velocity loop pole parameters, and then combine these squared values ​​with the identified moment of inertia. Multiply to obtain the real-time integral gain.

5. The adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor as described in claim 4, characterized in that, Before updating the speed loop control parameters within the current control cycle, the following steps are also included: Determine whether the calculated real-time proportional gain is within the set proportional gain safety range, and determine whether the real-time integral gain is within the set integral gain safety range. In response to the real-time proportional gain or real-time integral gain exceeding the corresponding safe range, the excess gain parameter is clamped and limited to the boundary value of the corresponding safe range, and the control parameters of the speed loop are updated using the clamped and limited parameters.

6. The adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor as described in claim 1, characterized in that, Execution is based on a heterogeneous system-on-a-chip architecture that includes field-programmable gate arrays and microprocessors, including: The data acquisition logic and pulse width modulation logic are run in the field programmable gate array to perform operations such as acquiring rotor position signals and three-phase current signals, performing spatial vector coordinate transformation, and driving the permanent magnet synchronous motor. Execute floating-point calculations in the microprocessor, including online identification and calculation of the real-time proportional gain and integral gain of the speed loop.

7. The adaptive vector control method for permanent magnet synchronous motors based on a variable forgetting factor as described in claim 6, characterized in that, Also includes: After the real-time proportional gain and integral gain are calculated within the microprocessor, the updated real-time proportional gain and integral gain are packaged to generate control frame data. The control frame data is sent from the microprocessor to the parameter control register allocated in the field programmable gate array via the on-chip high-speed bus, overwriting the gain data of the previous control cycle.

8. An adaptive vector control system, characterized in that, This adaptive vector control system, applied to robotic arm joint modules, includes: The sensor module is configured to acquire the rotor position signal and three-phase current signal of the target permanent magnet synchronous motor; The motor drive module is configured to output the quadrature-axis current setpoint from the speed loop of the control system. The signal is sent to the current loop for closed-loop control, driving the permanent magnet synchronous motor to run; The control processing module communicates with the sensor module and the motor drive module. The control processing module is configured to perform spatial vector coordinate transformation to obtain the actual current along the quadrature axis, the actual current along the direct axis, and the mechanical angular velocity. Collect quadrature-axis current setpoint The actual electromagnetic torque of the system and mechanical angular velocity Moment of inertia is calculated using a recursive least squares algorithm with a variable forgetting factor. With viscous friction coefficient The system performs online identification and uses the system's preset closed-loop damping ratio and the moment of inertia identified in the current control cycle. and viscous friction coefficient Calculate the real-time proportional gain and integral gain of the speed loop and update the control parameters; Among them, the variable forgetting factor The calculation is based on a dual-constraint model, which includes a dynamic term based on prior estimation error and a term based on the quadrature-axis current setpoint. The abrupt feedforward term of the time derivative is configured to control the forgetting factor during the initial stage of a speed drop caused by a sudden increase in load on the permanent magnet synchronous motor. It shows an exponential decline.

9. The adaptive vector control system as described in claim 8, characterized in that, The control and processing module includes a system-on-a-chip (SoC) processor chip, which integrates a hardware programmable logic array area and a hard-core microprocessor area on the same silicon substrate. The hardware programmable logic array area is equipped with high-speed input / output interfaces to connect to sensor modules and continuously acquire current and position data; the hard-core microprocessor area is configured to execute instruction code based on a recursive least squares algorithm with variable forgetting factor and parameter updates.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor of the control system, it implements the adaptive vector control method for permanent magnet synchronous motors based on variable forgetting factor as described in claim 1.

Images