Permanent magnet synchronous motor adaptive weight control method and system based on gradient prediction pulse pattern

Abstract

This invention discloses a control method and system for permanent magnet synchronous motors based on adaptive weights in gradient prediction pulse modes. The method includes: 1) establishing a discretized current gradient prediction model considering delay compensation and motor DC voltage feedback, and reconstructing a reference current trajectory with harmonic characteristics using a nominal optimized pulse mode; 2) designing a degree-of-freedom control logic based on nonlinear mapping to address the performance mismatch problem caused by fixed weights in the prediction cost function; and 3) dynamically adjusting the weighting factor by evaluating the normalized composite index of motor torque and current in real time. During acceleration transients, low weights are used to release the degrees of freedom at the pulse switching moment of the three-phase inverter, accelerating motor torque establishment. During steady-state cruise, high weights are used to strengthen the inverter pulse position constraint, locking in the optimal harmonic distribution. This invention solves the problem of balancing dynamic response and steady-state harmonics in permanent magnet synchronous motors, significantly improving the transient quality and voltage disturbance robustness of the motor control system.

Description

Technical Field

[0001] This invention belongs to the field of automotive traction drive and motor control technology, and particularly relates to a gradient prediction pulse mode control (GP³C) method and system for permanent magnet synchronous motors (PMSM) applied to rail transit traction systems. Background Technology

[0002] Time delay and low switching frequency operation are major challenges faced by high-power traction drive systems. To reduce switching losses in high-voltage, high-power inverters, the system typically operates under low pulse number conditions. Gradient predictive pulse mode control (GP³C) achieves excellent stator current harmonic quality at low switching frequencies by adjusting the offline pre-calculated optimized pulse mode (OPP) switching timing online in real time.

[0003] However, the traditional GP³C algorithm typically uses a fixed penalty weight factor in the cost function. This limits the offset of the pulse moment relative to the nominal position. In practical engineering applications, this fixed weight setting presents a significant performance trade-off: if... If the setting is too large, the pulse movement during transients is severely constrained, resulting in slow voltage vector adjustment and a significant increase in torque response time, which cannot meet the needs of rapid vehicle acceleration or hill climbing; conversely, if the setting is significantly reduced to improve dynamic response... Under steady-state operation, the pulses may randomly shift due to measurement noise or DC voltage fluctuations, disrupting the optimal harmonic distribution optimized offline and leading to a significant increase in total harmonic distortion (TDD). Therefore, there is an urgent need for a control logic that can sense the system's operating status in real time and adaptively adjust the control strength to balance transient response speed and steady-state harmonic quality. Summary of the Invention

[0004] This invention addresses the shortcomings of existing technologies by proposing an adaptive weighted control method and system for permanent magnet synchronous motors based on gradient prediction pulse modes. The aim is to enable the active release of the control degrees of freedom of the three-phase inverter switching pulses during transient conditions such as vehicle start-up and hill climbing, and to flexibly lock the optimal position of the switching pulses during steady-state conditions such as constant speed cruising. This solves the problem of balancing dynamic torque response and steady-state current harmonics in traction motors, thereby significantly improving the transient power quality and robustness against battery voltage fluctuations in the permanent magnet synchronous motor (PMSM) traction drive system.

[0005] The present invention adopts the following technical solution to solve the technical problem: 1. A control method for a permanent magnet synchronous motor based on adaptive weights in gradient prediction pulse modes, characterized in that it is applied to a permanent magnet synchronous motor system consisting of a permanent magnet synchronous motor, a three-phase inverter, a DC power supply, and a sensor array, and includes the following steps: Step 1: Obtain the stator three-phase current of the permanent magnet synchronous motor at time k in real time using a sensor array. DC bus voltage Rotor position angle angular velocity and stator three-phase voltage ; Using coordinate transformation After projecting onto the stationary coordinate system, the current vector at time k is obtained. i s ( k Thus, the initial value of the predicted current state at time k+1 can be obtained using equation (1). i s ( k+ 1): (1) In equation (1), I It is the identity matrix; T s The sampling period; The continuous state matrix of the permanent magnet synchronous motor system; and , s This represents the stator inductance matrix of a permanent magnet synchronous motor. express s The inverse matrix, This represents the stator resistance matrix of a permanent magnet synchronous motor. B d This represents the inverse of the stator inductance matrix of a permanent magnet synchronous motor; Step 2: Calculate the instantaneous modulation ratio of the three-phase inverter at the current time k using equation (2). m : (2) In equation (2), The rated flux linkage of the stator of a permanent magnet synchronous motor; Step 3: Query the preset offline optimized pulse pattern library for... m Matched nominal switching time sequence , Indicates the first The nominal switching time corresponding to each switching action. Indicates transpose. Represents the prediction time domain T p The total number of switching actions contained therein, of which the prediction time domain T p It is an observation window that starts at time k+1; Step 4: Based on the given reference torque Determine the dq-axis current setpoint of the permanent magnet synchronous motor based on the maximum torque-to-current ratio curve. Thus, the prediction time domain is obtained through equation (3). T p The j-th nominal switching time Fundamental reference current component of a lower permanent magnet synchronous motor : (3) In equation (3), Indicates the j-th nominal switching time. The corresponding rotor position angle of the permanent magnet synchronous motor; Indicates the inverse Park transform; Step 5: Calculate the prediction time domain using equation (4) T p The j-th nominal switching time Current harmonic ripple components of a lower permanent magnet synchronous motor : (4) In equation (4), Indicates that the permanent magnet synchronous motor is in The pulse voltage it withstands at any given moment, Indicates that the permanent magnet synchronous motor is in The fundamental voltage at time t; Step 6: Construct the prediction time domain using equation (5) T p Reference current trajectory vector of internal permanent magnet synchronous motor r : (5) Step 7: Use equation (6) to obtain the permanent magnet synchronous motor at the current time. Reference current composite vector : = (6) Step 8: Construct the normalized composite index at time k using equation (7). : (7) In equation (7), The electromagnetic torque of the permanent magnet synchronous motor; This indicates the rated torque of the permanent magnet synchronous motor; This is the reference value for the rated current of the permanent magnet synchronous motor; , These are preset weighting coefficients for torque and current; Step 9: Calculate the weighting factor of the deviation at time k according to equation (8). : (8) In equation (7), 、 The maximum and minimum penalty weights are preset constants. The preset sensitivity coefficient, This is a preset error threshold for judging the switching of working conditions; Step 10: Obtain the lower triangular current gradient matrix using equation (9). M At the current time k, at the j-th nominal switching time The corresponding current gradient : (9) In equation (8), For the first Nominal switching time The corresponding voltage vector, For permanent magnet synchronous motors at rotor position angle The back electromotive force vector under, For the first At a nominal switching time The corresponding estimated rotor position angle; Step 11: Construct formula (9) to build the cost function J and solve it to obtain the prediction time domain. Internal nominal switching time sequence The optimal time offset sequence : st (10) In equation (9), This represents the time offset of the j-th switch action. The time offset sequence to be solved; Step 12: and After adding the corresponding values, the updated pulse switching time sequence is obtained. and will Transform into prediction time domain The gate drive signal is used to control the on and off of the power switching transistors in the three-phase inverter, thereby realizing the adaptive weighted predictive control of the permanent magnet synchronous motor.

[0006] 2. A permanent magnet synchronous motor control system based on adaptive weights in gradient prediction pulse modes, characterized in that it is applied to a permanent magnet synchronous motor system consisting of a permanent magnet synchronous motor, a three-phase inverter, a DC power supply and a sensor array, wherein the permanent magnet synchronous motor control system includes: a signal analysis and prediction module, a reference trajectory reconstruction module, an error assessment and weight adjustment module, a gradient correction and optimization solution module, and a pulse generation and driving module; The signal analysis and prediction module includes a coordinate transformation unit and a discretization prediction unit, used to receive the stator three-phase current, DC bus voltage, rotor position angle, angular velocity and stator three-phase voltage at the current moment collected by the sensor array; The coordinate transformation unit converts the current stator three-phase current into the actual current vector in the stationary coordinate system. The discretization prediction unit calculates the initial value of the predicted current state at the next moment based on the actual current vector at the current moment. The reference trajectory reconstruction module calculates the instantaneous modulation ratio at the current moment based on the angular velocity and DC bus voltage, thereby matching the nominal switching time sequence in the prediction time domain in a preset offline optimized pulse mode lookup table; and under the nominal switching time sequence, calculates the fundamental reference current component and current harmonic ripple component of the permanent magnet synchronous motor, which are used to construct the reference current synthesis vector in the prediction time domain, and together with the initial value of the predicted current state at the next moment, constructs the reference current trajectory vector of the permanent magnet synchronous motor in the prediction time domain. The error assessment and weight adjustment module calculates the torque tracking error and current vector amplitude deviation based on the actual current vector and the reference current vector at the current moment. Then, based on the error, it calculates the normalized composite evaluation index at the current moment using the Sigmoid nonlinear mapping operator, and calculates the weight factor at the current moment through the nonlinear mapping relationship. The gradient correction and optimization solution module constructs and solves a cost function that satisfies the time constraints based on the current DC bus voltage, the reference current trajectory vector, and the current weight factor, to obtain the optimal time offset sequence. The pulse generation and drive module superimposes the optimal time offset sequence and the nominal switching time sequence to generate an updated pulse switching time sequence, which is then converted into a gate logic signal to control the on and off of the power switching transistors in the three-phase inverter, thereby achieving adaptive weighted predictive control of the permanent magnet synchronous motor.

[0007] Compared with existing technologies, the beneficial effects of this invention are reflected in: 1. This invention resolves the mismatch between dynamic response and steady-state quality, achieving intelligent dynamic allocation of traction performance. Addressing the performance trade-off problem of the traditional GP³C algorithm, which uses fixed penalty weights leading to either "fast but unstable" or "stable but too slow," this invention utilizes a nonlinear mapping logic based on the Sigmoid function designed in step 8 to calculate the normalized composite evaluation index in real time. This method can automatically adjust the weights during vehicle start-up or large-scale speed changes. By adjusting the weights to the release zone, the three-phase inverter pulse movement freedom is actively released, significantly accelerating the establishment speed of the traction motor's electromagnetic torque. In the steady-state constant-speed cruise, the weights are automatically adjusted to the lock zone, flexibly locking the pulse position. This "transient release, steady-state lock" process fundamentally solves the constraint effects in multi-objective optimization, balancing fast response and low distortion.

[0008] 2. This invention enhances the system's robustness to DC-side voltage fluctuations and reduces hardware dependence under complex operating conditions. Addressing the issue of inaccurate prediction models caused by instantaneous voltage drops at the battery terminal during rapid acceleration of electric vehicles, this invention introduces a real-time bus voltage feedback correction mechanism in step 10, directly affecting the online calculation of the current gradient matrix. By real-time correction of the lower triangular gradient elements, the control algorithm can perceive voltage disturbances in real time and automatically adjust pulse offset compensation. This feature overcomes the excessive reliance of traditional methods on a stable DC power supply, significantly improves the control accuracy of the permanent magnet synchronous motor drive system under severe battery voltage fluctuations, avoids current oscillations caused by predicted trajectory deviations, and enhances the robustness of the traction system under complex road conditions.

[0009] 3. This invention ensures the physical feasibility of the inverter switching logic and improves the reliability of system operation. To address the risk of inverter damage due to pulse overlap or timing errors that may result from pure mathematical optimization, this invention embeds physical switching timing constraints into the quadratic programming (QP) solution in step 11. This ensures that all optimized pulse offsets are within the physical safety range and maintain the correct action sequence. This technique tightly integrates abstract mathematical optimization with the physical constraints of the three-phase inverter, ensuring optimal control performance while eliminating the possibility of inverter shoot-through or illegal switching actions from the underlying logic, greatly improving the operational safety of the power electronic converter.

[0010] 4. This invention, by establishing a normalized deviation index system, changes the traditional control strategy's reliance on large-scale parameter traversal and debugging based on manual experience. Through the collaborative logic of steps 8 and 9, the controller can automatically identify the system's operating boundaries and select the optimal control strength. This method has strong versatility for permanent magnet synchronous motors of different power levels and power electronic converters with different topologies, significantly shortening the cycle of actual motor testing in electric vehicles and rail transit, reducing system integration costs, and providing a corresponding technical solution for the precision control of motor systems in modern intelligent transportation equipment. Attached Figure Description

[0011] Figure 1. Logical architecture block diagram of the control method of the present invention; Figure 2 is a flowchart of the entire logic execution of the control method of the present invention; Figure 3 is a block diagram of the hardware and algorithm architecture of an 800V electric vehicle drive system; Figure 4 shows the penalty weights. With composite error The changing Sigmoid characteristic curve; Figure 5 is a schematic diagram illustrating the physical meaning of the nominal pulse offset in the predicted time domain; Figure 6 shows a comparison of the torque response waveforms of the present invention and the traditional fixed-weight method under acceleration conditions; Figure 7. FFT spectrum analysis of stator current under steady-state cruise conditions. Detailed Implementation

[0012] The invention will be further described below with reference to the accompanying drawings.

[0013] In this embodiment, an adaptive weighted control method and system for permanent magnet synchronous motors based on gradient prediction pulse modes is presented. Its core logic lies in nonlinearly adjusting the penalty weight coefficient related to the switching moment offset in the prediction cost function by real-time monitoring of the dynamic deviation of the electric vehicle drive system. This allows for the coordinated optimization of the system's torque build-up speed and steady-state energy efficiency at low switching frequencies. This method possesses three core technical features: current state evolution based on digital delay compensation, online correction of the lower triangular sensitivity matrix considering bus voltage disturbances, and dynamic adjustment of control degrees of freedom based on nonlinear mapping functions.

[0014] In this embodiment, a logical architecture for a permanent magnet synchronous motor control system based on adaptive weights in gradient prediction pulse patterns is as follows: Figure 1 As shown. The system mainly consists of a torque estimator, a normalized deviation evaluation index calculation module, an adaptive weight mapping operator, a reference trajectory reconstruction module, a gradient matrix real-time correction module, a one-step prediction delay compensation module, and a constrained quadratic programming optimization module; as shown. Figure 2As shown, the closed-loop implementation and calibration process of the adaptive weighted control method for permanent magnet synchronous motors includes the following steps: Step 1: Obtain the stator three-phase current of the permanent magnet synchronous motor at time k in real time using a sensor array. DC bus voltage Rotor position angle angular velocity and stator three-phase voltage ; Using coordinate transformation After projecting onto the stationary coordinate system, the current vector at time k is obtained. i s ( k Thus, the initial value of the predicted current state at time k+1 can be obtained using equation (1). i s ( k+ 1): (1) In equation (1), I It is the identity matrix; T s The sampling period; The continuous state matrix of the permanent magnet synchronous motor system; and , s This represents the stator inductance matrix of a permanent magnet synchronous motor. express s The inverse matrix, This represents the stator resistance matrix of a permanent magnet synchronous motor. B d This represents the inverse of the stator inductance matrix of a permanent magnet synchronous motor; In practical high-voltage electric vehicle drive systems, such as Figure 3 As shown, there is a computational delay of approximately one sampling cycle from the time the current sensor samples the data to the time the microprocessor (MCU) calculates and executes the control command. To compensate for this delay, this step utilizes discretized state-space equations for one-step prediction.

[0015] Set sampling period Compared to the traditional Euler forward difference method, this embodiment employs the Tustin bilinear transform (structure shown in Equation 1), which exhibits better numerical stability when handling back EMF disturbances under high-speed rotation. The motor physical parameters are shown in Table 1, and these parameters are injected into the prediction model in real time to ensure the initial prediction values. The accuracy.

[0016] Table 1 Implementation parameters of electric vehicle drive motor Step 2: Calculate the instantaneous modulation ratio of the three-phase inverter at the current time k using equation (2). m : (2) In equation (2), Rated flux linkage of the stator of a permanent magnet synchronous motor; modulation ratio This reflects the utilization rate of DC voltage at the motor's operating point. Real-time calculation of the modulation ratio enables the control algorithm to detect and automatically compensate for drops in battery pack voltage (common during rapid acceleration), thereby maintaining the consistency of control gain.

[0017] Step 3: Query the preset offline optimized pulse pattern library for... m Matched nominal switching time sequence , Indicates the first The nominal switching time corresponding to each switching action. Indicates transpose. Represents the prediction time domain T p The total number of switching actions contained therein, of which the prediction time domain T p The observation window starts at time k+1; the optimized pulse mode library OPP is obtained offline by minimizing the total harmonic distortion (THD) of the current across the entire range. Step 3 aims to extract the "theoretically perfect" pulse position for the current speed and voltage from the lookup table. The prediction time domain is then set. By sliding the window, you can ensure that the next 2-5 switch points are always covered.

[0018] Step 4: Based on the given reference torque Determine the dq-axis current setpoint of the permanent magnet synchronous motor based on the maximum torque-to-current ratio curve. Thus, the prediction time domain is obtained through equation (3). T p The j-th nominal switching time Fundamental reference current component of a lower permanent magnet synchronous motor : (3) In equation (3), Indicates the j-th nominal switching time. The corresponding rotor position angle of the permanent magnet synchronous motor; Indicates the inverse Park transform; Step 5: Calculate the prediction time domain using equation (4) T p The j-th nominal switching time Current harmonic ripple components of a lower permanent magnet synchronous motor : (4) In equation (4), Indicates that the permanent magnet synchronous motor is in The pulse voltage it withstands at any given moment, Indicates that the permanent magnet synchronous motor is in The fundamental voltage at a given moment; traditional control methods only track sine waves, but this is ineffective at low switching frequencies. This invention extracts the "harmonic characteristics" corresponding to the optimal pulse in steps 4 and 5 and superimposes them back into the fundamental voltage. This means that the online controller is not forcibly smoothing out harmonics, but actively tracking a reference trajectory of "predicted ripple," which greatly reduces control pressure and system losses in steady state.

[0019] Step 6: Construct the prediction time domain using equation (5) T p Reference current trajectory vector of internal permanent magnet synchronous motor r : (5) Step 7: Use equation (6) to obtain the permanent magnet synchronous motor at the current time. Reference current composite vector : = (6) Step 8: Construct the normalized composite index at time k using equation (7). : (7) In equation (7), The electromagnetic torque of the permanent magnet synchronous motor; This indicates the rated torque of the permanent magnet synchronous motor; This is the reference value for the rated current of the permanent magnet synchronous motor; , These are preset weighting coefficients for torque and current; The system's "deviation" was quantified in real time. When the driver depresses the pedal deeply, the torque reference value... When the torque increases from 50 Nm to 300 Nm (e.g.) Figure 6 As shown in the figure, the torque residual term dominates, and the calculated value is... It quickly rose to around 0.85.

[0020] Step 9: Calculate the weighting factor of the deviation at time k according to equation (8). : (8) In equation (7), 、 The maximum and minimum penalty weights are preset constants. The preset sensitivity coefficient, This is a preset error threshold for judging the switching of working conditions; Using the Sigmoid function shown in equation (4), the penalty weights in the objective function are adjusted in real time according to the evaluation index in step 3. A preset upper limit for the steady-state strong constraint weights is established. Transient free release weight lower bound : This is the control core of the present invention, and its characteristic curve is as follows: Figure 4 As shown. The steady-state high-weight λ is set. max =10 10 Transient low weight Switching threshold E th =0.05.

[0021] In accelerated transients (Enorm ≥ 0.05): weights It automatically adjusts to the release zone. At this point, the pulse position can be significantly shifted to generate the required instantaneous voltage vector, significantly accelerating torque build-up.

[0022] In steady-state cruise (Enorm < 0.05): weights The pulse is then locked in the locked area to prevent timing jitter caused by sampling noise.

[0023] Regarding the above acceleration conditions ( =0.85), weight Immediately adaptively downgraded to 1.2×10 6 At this point, the constraint of the cost function on the pulse deviation from the nominal position is actively released, allowing the algorithm to change the terminal voltage vector by moving the pulse position over a large range. Its mapping curve characteristics are as follows: Figure 5 As shown.

[0024] Step 10: Obtain the lower triangular current gradient matrix using equation (9). M At the current time k, at the j-th nominal switching time The corresponding current gradient : (9) In equation (8), For the first Nominal switching time The corresponding voltage vector, For permanent magnet synchronous motors at rotor position angle The back electromotive force vector under, For the first At a nominal switching time The corresponding estimated rotor position angle.

[0025] If the battery in the motor drive system experiences a voltage drop during discharge, the gradient matrix M immediately detects this change by correcting its gradient elements. This allows it to automatically adjust the pulse shift step size in subsequent optimizations, enhancing the system's robustness to DC voltage fluctuations. The influence of the switching time offset on the current trajectory is as follows: Figure 5 As shown.

[0026] Step 11: Construct formula (9) to build the cost function J and solve it to obtain the prediction time domain. Internal nominal switching time sequence The optimal time offset sequence : st (10) In equation (9), This represents the time offset of the j-th switch action. The time offset sequence to be solved; Adaptive weights Substituting into the objective function shown in equation (10), the optimal switching timing correction sequence is solved within the framework of rolling optimization. In this acceleration example, the QP solver satisfies the physical switching timing constraints. Under the premise of this, the maximum forward shift of the pulse switching point is calculated to be 420. μs ,like Figure 6 As shown, the torque is 2.8. ms The target value tracking was completed within 4.5 seconds, compared to the traditional fixed-weight scheme. ms The power build-up speed has been improved by 37.8%, significantly enhancing the dynamic quality of acceleration response.

[0027] Step 12: and After adding the corresponding values, the updated pulse switching time sequence is obtained. and will Transform into prediction time domain The gate drive signal is used to control the on and off of the power switching transistors in the three-phase inverter, thereby realizing the adaptive weighted predictive control of the permanent magnet synchronous motor.

[0028] As the vehicle speed stabilizes, the torque deviation... Decrease to bring it back below 0.05, mapping logic control. Orders of magnitude. At this point, the switching pulse is flexibly locked at the offline-optimized nominal position, preventing timing jitter caused by sampling noise. For example... Figure 7 As shown, the total harmonic distortion (TDD) of the stator current stabilizes at around 1.3%. This step, by strengthening pulse position constraints, effectively suppresses the magnetic pull fluctuations of the permanent magnet synchronous motor excited by high-order harmonics at low switching frequencies, reduces electromagnetic noise, and improves the operational stability of the permanent magnet synchronous motor.

[0029] This embodiment, through the above steps, without changing the physical upper limit of the inverter's switching frequency, successfully solves the problem of balancing the power performance and energy efficiency of permanent magnet synchronous motors in electric vehicles under complex dynamic environments by intelligently allocating the control degrees of freedom.

Claims

1. A control method for a permanent magnet synchronous motor based on adaptive weights in gradient prediction pulse modes, characterized in that, It is applied to a permanent magnet synchronous motor system consisting of a permanent magnet synchronous motor, a three-phase inverter, a DC power supply, and a sensor array, and includes the following steps: Step 1: Obtain the stator three-phase current of the permanent magnet synchronous motor at time k in real time using a sensor array. DC bus voltage Rotor position angle angular velocity and stator three-phase voltage ; Using coordinate transformation After projecting onto the stationary coordinate system, the current vector at time k is obtained. i s ( k Thus, the initial value of the predicted current state at time k+1 can be obtained using equation (1). i s ( k+ 1): (1) In equation (1), I It is the identity matrix; T s The sampling period; The continuous state matrix of the permanent magnet synchronous motor system; and , s This represents the stator inductance matrix of a permanent magnet synchronous motor. express s The inverse matrix, This represents the stator resistance matrix of a permanent magnet synchronous motor. B d This represents the inverse of the stator inductance matrix of a permanent magnet synchronous motor; Step 2: Calculate the instantaneous modulation ratio of the three-phase inverter at the current time k using equation (2). m : (2) In equation (2), The rated flux linkage of the stator of a permanent magnet synchronous motor; Step 3: Query the preset offline optimized pulse pattern library for... m Matched nominal switching time sequence , Indicates the first The nominal switching time corresponding to each switching action. Indicates transpose. Represents the prediction time domain T p The total number of switching actions contained therein, of which the prediction time domain T p It is an observation window that starts at time k+1; Step 4: Based on the given reference torque Determine the dq-axis current setpoint of the permanent magnet synchronous motor based on the maximum torque-to-current ratio curve. Thus, the prediction time domain is obtained through equation (3). T p The j-th nominal switching time Fundamental reference current component of a lower permanent magnet synchronous motor : (3) In equation (3), Indicates the j-th nominal switching time. The corresponding rotor position angle of the permanent magnet synchronous motor; Indicates the inverse Park transform; Step 5: Calculate the prediction time domain using equation (4) T p The j-th nominal switching time Current harmonic ripple components of a lower permanent magnet synchronous motor : (4) In equation (4), Indicates that the permanent magnet synchronous motor is in The pulse voltage it withstands at any given moment, Indicates that the permanent magnet synchronous motor is in The fundamental voltage at time t; Step 6: Construct the prediction time domain using equation (5) T p Reference current trajectory vector of internal permanent magnet synchronous motor r : (5) Step 7: Use equation (6) to obtain the permanent magnet synchronous motor at the current time. Reference current composite vector : = (6) Step 8: Construct the normalized composite index at time k using equation (7). : (7) In equation (7), The electromagnetic torque of the permanent magnet synchronous motor; This indicates the rated torque of the permanent magnet synchronous motor; This is the reference value for the rated current of the permanent magnet synchronous motor; , These are preset weighting coefficients for torque and current; Step 9: Calculate the weighting factor of the deviation at time k according to equation (8). : (8) In equation (7), 、 The maximum and minimum penalty weights are preset constants. The preset sensitivity coefficient, This is a preset error threshold for judging the switching of working conditions; Step 10: Obtain the lower triangular current gradient matrix using equation (9). M At the current time k, at the j-th nominal switching time The corresponding current gradient : (9) In equation (8), For the first Nominal switching time The corresponding voltage vector, For permanent magnet synchronous motors at rotor position angle The back electromotive force vector under, For the first At a nominal switching time The corresponding estimated rotor position angle; Step 11: Construct formula (9) to build the cost function J and solve it to obtain the prediction time domain. Internal nominal switching time sequence The optimal time offset sequence : s.t (10) In equation (9), This represents the time offset of the j-th switch action. The time offset sequence to be solved; Step 12: and After adding the corresponding values, the updated pulse switching time sequence is obtained. and will Transform into prediction time domain The gate drive signal is used to control the on and off of the power switching transistors in the three-phase inverter, thereby realizing the adaptive weighted predictive control of the permanent magnet synchronous motor.

2. A permanent magnet synchronous motor control system based on adaptive weights in gradient prediction pulse modes, characterized in that, It is applied to a permanent magnet synchronous motor system consisting of a permanent magnet synchronous motor, a three-phase inverter, a DC power supply and a sensor array. The permanent magnet synchronous motor control system includes: a signal analysis and prediction module, a reference trajectory reconstruction module, an error assessment and weight adjustment module, a gradient correction and optimization solution module, and a pulse generation and driving module. The signal analysis and prediction module includes a coordinate transformation unit and a discretization prediction unit, used to receive the stator three-phase current, DC bus voltage, rotor position angle, angular velocity and stator three-phase voltage at the current moment collected by the sensor array; The coordinate transformation unit converts the current stator three-phase current into the actual current vector in the stationary coordinate system. The discretization prediction unit calculates the initial value of the predicted current state at the next moment based on the actual current vector at the current moment. The reference trajectory reconstruction module calculates the instantaneous modulation ratio at the current moment based on the angular velocity and DC bus voltage, thereby matching the nominal switching time sequence in the prediction time domain in a preset offline optimized pulse mode lookup table; and under the nominal switching time sequence, calculates the fundamental reference current component and current harmonic ripple component of the permanent magnet synchronous motor, which are used to construct the reference current synthesis vector in the prediction time domain, and together with the initial value of the predicted current state at the next moment, constructs the reference current trajectory vector of the permanent magnet synchronous motor in the prediction time domain. The error assessment and weight adjustment module calculates the torque tracking error and current vector amplitude deviation based on the actual current vector and the reference current vector at the current moment. Then, based on the error, it calculates the normalized composite evaluation index at the current moment using the Sigmoid nonlinear mapping operator, and calculates the weight factor at the current moment through the nonlinear mapping relationship. The gradient correction and optimization solution module constructs and solves a cost function that satisfies the time constraints based on the current DC bus voltage, the reference current trajectory vector, and the current weight factor, to obtain the optimal time offset sequence. The pulse generation and drive module superimposes the optimal time offset sequence and the nominal switching time sequence to generate an updated pulse switching time sequence, which is then converted into a gate logic signal to control the on and off of the power switching transistors in the three-phase inverter, thereby achieving adaptive weighted predictive control of the permanent magnet synchronous motor.

Images