A method for suppressing torque ripple of a permanent magnet synchronous motor

Abstract

The application discloses a permanent magnet synchronous motor torque ripple suppression method and belongs to the technical field of permanent magnet synchronous motor control. The method comprises the following steps: calculating average error voltage caused by the nonlinear characteristics of an inverter, obtaining initial compensation voltage according to the average error voltage, considering the air gap magnetic field distortion of the permanent magnet synchronous motor, deducing optimal harmonic current injection reference value, obtaining harmonic reference voltage by using the control strategy of optimal harmonic current injection, and calculating optimal compensation voltage based on the initial compensation voltage and the harmonic reference voltage; collecting operation data of the permanent magnet synchronous motor to construct a PSO-BP neural network agent model; and embedding the trained PSO-BP neural network agent model into a magnetic field oriented control system of the permanent magnet synchronous motor to suppress torque ripple. The application does not need to depend on specific motor parameters, is suitable for all working conditions, has better torque ripple suppression effect, and significantly improves system dynamic response.

Description

Technical Field

[0001] This invention belongs to the field of permanent magnet synchronous motor control technology, and specifically relates to a method for suppressing torque ripple in a permanent magnet synchronous motor. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are widely used in electric vehicle drive control systems due to their compact structure, high power density, strong air gap magnetic flux, and excellent torque-to-inertia ratio. However, the air gap magnetic field distortion caused by the motor itself and the inherent nonlinear characteristics of the inverter can generate a large number of high-order harmonics (especially the 5th and 7th harmonics) in the motor input current, which in turn causes torque pulsation, leading to increased operating noise, decreased control accuracy, and seriously affecting the motor's operating performance.

[0003] Existing torque ripple suppression methods are mainly divided into two categories: motor design optimization and control algorithm improvement. In terms of design optimization, the sinusoidal nature of the back EMF is improved and spatial harmonics are reduced by adjusting the stator winding structure and slot design, but the suppression effect on high-order harmonic currents is limited. Regarding control algorithms, traditional methods include the average voltage error compensation method and the harmonic current injection method. The average voltage error compensation method mainly addresses nonlinear issues such as inverter dead zones by compensating for voltage fluctuations, but it does not consider the harmonic effects of the motor itself. The harmonic current injection method cancels torque ripples by injecting harmonic currents of specific frequencies and phases, but it relies on precise motor parameters and has a slow dynamic response.

[0004] In recent years, data-driven surrogate model methods have gradually emerged, utilizing models such as neural networks to replace traditional control modules without relying on precise mathematical models. However, existing neural network-based torque ripple suppression methods still suffer from high data dependence, insufficient model generalization ability, and the need to improve dynamic response. Therefore, there is an urgent need to develop a torque ripple suppression method that balances suppression effectiveness, dynamic performance, and robustness. Summary of the Invention

[0005] The purpose of this invention is to provide a method for suppressing torque ripple in permanent magnet synchronous motors, which solves the problems of high data dependence, insufficient model generalization ability, and slow dynamic response in existing neural network-based torque ripple suppression methods.

[0006] To achieve the above objectives, this invention provides a method for suppressing torque ripple in a permanent magnet synchronous motor (PMSM). The method includes: Step S1, calculating the average error voltage caused by the nonlinear characteristics of the inverter, and obtaining an initial compensation voltage based on the average error voltage; Step S2, considering the air gap magnetic field distortion of the PMSM, deriving the optimal harmonic current injection reference value, and obtaining the harmonic reference voltage using the optimal harmonic current injection control strategy, and calculating the optimal compensation voltage based on the initial compensation voltage and the harmonic reference voltage; Step S3, collecting operating data of the PMSM using harmonic reference voltage compensation and optimal harmonic current injection to construct a particle swarm optimization backpropagation (PSO-BP) neural network surrogate model; Step S4, embedding the trained PSO-BP neural network surrogate model into the PMSM field-oriented control system to suppress torque ripple.

[0007] The beneficial effects of this invention are as follows:

[0008] Combining the advantages of the average error voltage compensation method and the optimal reference harmonic current injection method, and considering the influence of inverter nonlinearity and motor magnetic field distortion, the torque ripple suppression effect is significant. The Particle Swarm Optimization-Back Propagation (PSO-BP) neural network surrogate model is adopted, which does not rely on precise motor parameters, is applicable to the entire operating range, has strong robustness, simplifies the control structure, and improves the dynamic response speed of the system. This invention is suitable for scenarios with high requirements for torque stability, such as electric vehicle drives. Attached Figure Description

[0009] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0010] Figure 1 A flowchart of the torque ripple suppression method for permanent magnet synchronous motors provided by the present invention;

[0011] Figure 2 The harmonic voltage compensation model provided by this invention;

[0012] Figure 3 This is a schematic diagram of torque ripple control based on the PSO-BP proxy model provided by the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1 The flowchart of the torque ripple suppression method for permanent magnet synchronous motors provided by the present invention is as follows: Figure 1 The method includes:

[0016] Step S1: Calculate the average error voltage caused by the nonlinear characteristics of the inverter, and obtain the initial compensation voltage based on the average error voltage;

[0017] Step S2: Considering the air gap magnetic field distortion of the permanent magnet synchronous motor, the optimal harmonic current injection reference value is derived, and the harmonic reference voltage is obtained by adopting the optimal harmonic current injection control strategy. The optimal compensation voltage is calculated based on the initial compensation voltage and the harmonic reference voltage.

[0018] Step S3: The operating data of the permanent magnet synchronous motor with harmonic reference voltage compensation and optimal harmonic current injection are collected to construct a particle swarm optimization backpropagation (PSO-BP) neural network surrogate model. The optimal network structure is determined by the number of hidden layer neurons in the PSO-BP neural network surrogate model. The operating data of the permanent magnet synchronous motor with harmonic voltage compensation model is used to train the PSO-BP neural network surrogate model after normalization preprocessing.

[0019] Step S4: The trained PSO-BP neural network surrogate model is embedded into the field orientation control system of the permanent magnet synchronous motor to replace the traditional harmonic current injection module, thereby achieving real-time output of the optimal compensation voltage to suppress torque pulsation.

[0020] Among them, the average error voltage The calculation is as follows:

[0021] ;

[0022] in, Dead time, For opening time, For shutdown delay time, For periodic time, For the voltage drop of the switching transistor, For diode voltage drop, This is the DC bus voltage.

[0023] Step S1 further includes: obtaining the average error voltage. coordinate system The initial compensation voltage for the shaft is calculated as follows:

[0024] ;

[0025] in, for The initial compensation voltage of the shaft, for The initial compensation voltage of the shaft, This refers to the phase current of phase A. This refers to the phase current of phase B. This refers to the C-phase current. It is a symbolic function.

[0026] Sign function The calculation is as follows:

[0027] ;

[0028] in, The electric angular velocity of the motor. The initial angle of the phase current. For time, The order of the harmonic current. For phase current, Values , , , Let A be the current in phase A of the three-phase system. Let B be the current in the three-phase system. This represents the current in phase C of the three-phase system.

[0029] Step S2 includes:

[0030] Considering the air gap magnetic field distortion of the permanent magnet synchronous motor, and setting the superposition of the 6th harmonic terms in the dq axis flux linkage and current to zero, the optimal harmonic current injection reference value is derived.

[0031] The optimal harmonic current injection reference value is injected by using a synchronous rotating PI controller to control the DC current, thereby obtaining the corresponding harmonic reference voltage in the dq coordinate system;

[0032] The harmonic reference voltage in the dq coordinate system is transformed to [the desired value] using the Clark transform. coordinate system, to obtain Harmonic reference voltage in coordinate system;

[0033] according to The initial compensation voltage, harmonic reference voltage, and shaft The fundamental voltage of the axis is calculated. Optimal compensation voltage for the shaft.

[0034] In step S2, considering the air gap magnetic field distortion of the permanent magnet synchronous motor, and setting the superposition of the 6th harmonic terms in the dq-axis flux linkage and current to zero, the optimal harmonic current injection reference value is derived as follows:

[0035] ;

[0036] in, , These are the optimal harmonic current injection reference values ​​for the d-axis and q-axis, respectively. , and These represent the harmonic amplitudes of the fundamental, fifth, and seventh permanent magnet flux linkages, respectively. , The initial phase angles are those of the fifth and seventh harmonic components, respectively. , These are the fundamental currents along the d-axis and q-axis, respectively. For d-axis inductance, It is the q-axis inductance. The electric angular velocity of the motor. For time.

[0037] Step S2 also includes:

[0038] Inject the optimal harmonic current into the reference value and The injection process uses a synchronous rotating PI controller to control the DC current, thereby obtaining the corresponding harmonic reference voltage in the dq coordinate system.

[0039] The harmonic reference voltage is converted to [value] using the Clark transform. In coordinate system, we obtain Harmonic reference voltage in coordinate system Axial components and Axial components ;

[0040] The average error voltage compensation and the optimal harmonic current reference value were injected to obtain the result. Shaft harmonic reference voltage compensation model, based on Shaft harmonic reference voltage compensation model, based on The initial compensation voltage and harmonic reference voltage of the shaft. Axial components and Axial components , The fundamental voltage of the axis is calculated. Optimal compensation voltage for shaft and .

[0041] Optimal compensation voltage for shaft and Acquisition includes:

[0042] right The fundamental voltage of the shaft, The initial compensation voltage and harmonic reference voltage of the shaft. Axial components Summation yields Optimal compensation voltage for shaft ;

[0043] right The fundamental voltage of the shaft, The initial compensation voltage and harmonic reference voltage of the shaft. Axial components Summation yields Optimal compensation voltage for shaft .

[0044] In step S3, the operating data of the permanent magnet synchronous motor includes the model input dataset and the model output dataset for the PSO-BP neural network surrogate model. The model input dataset includes the d-axis current during the operation of the permanent magnet synchronous motor. q-axis current Rotor position electrical angle Electric angular velocity Electromagnetic torque command and dead zone time The model output dataset includes Optimal compensation voltage for shaft and .

[0045] Step S3 also includes:

[0046] Multiple groups were randomly selected to obtain the predicted optimal compensation voltage output by the PSO-BP neural network surrogate model under different operating conditions based on the input dataset;

[0047] The optimal compensation voltage calculated based on steps S1 and S2 is taken as the actual optimal compensation voltage.

[0048] The optimal network structure of the PSO-BP neural network surrogate model is determined with the goal of minimizing the root mean square error (RMSE) between the predicted optimal compensation voltage and the actual optimal compensation voltage.

[0049] Among them, the PSO-BP neural network surrogate model with the optimal network structure is the trained PSO-BP neural network surrogate model.

[0050] The model input and output data collected under different operating conditions of the permanent magnet synchronous motor (PMSM) are used as the operating data of the PMSM. After normalization preprocessing, the PSO-BP neural network surrogate model is trained and validated. To determine the structure of the neural network, the number of hidden layer neurons in the BP neural network is optimized using the particle swarm optimization algorithm. Operating data of a high-precision harmonic voltage compensation model is collected; for example, multiple data points can be randomly selected to compare the optimal compensation voltage predicted by the PSO-BP neural network surrogate model under different operating conditions. The optimal network structure is determined with the minimum root mean square error (RMSE) as the objective.

[0051] ;

[0052] in, For the first The predicted optimal compensation voltage for each output. For the first The actual optimal compensation voltage for each output. This represents the number of data points to be predicted.

[0053] In a specific embodiment of the present invention, for example, the final number of input layer nodes in the BP neural network is 6, the number of output layer nodes is 2, and the number of hidden layers is determined to be 6 using the particle swarm optimization algorithm. Finally, the network weights and neuron biases of the BP neural network are updated using the batch gradient descent method to output the optimal PSO-BP neural network surrogate model.

[0054] After embedding the PSO-BP neural network surrogate model into the field orientation control system of permanent magnet synchronous motor, torque ripple suppression can be achieved over a wide speed range and under different load conditions without relying on precise motor parameters.

[0055] The following is combined Figure 2 and Figure 3 The present invention will be further described below. Figure 2 The harmonic voltage compensation model provided by this invention Figure 3 This is a schematic diagram of torque ripple control based on the PSO-BP proxy model provided by the present invention.

[0056] In the simulation, a joint simulation of the motor model and MATLAB Simulink is adopted. Traditional analytical models based on the dq axes in Simulink usually ignore factors such as spatial harmonics, magnetic saturation, cogging effect, and magnetic circuit nonlinearity, and cannot realistically reflect the sources of 5th / 7th order current harmonics and torque pulsations generated in actual motor operation. The reduced-order motor model can accurately calculate the changes in the air gap magnetic field, allowing the compensation algorithm to be verified close to the real operating conditions during the simulation stage.

[0057] The torque ripple of the permanent magnet synchronous motor is reduced by using a harmonic voltage compensation model. A dual closed-loop vector control (inner current loop + outer speed loop) is adopted, and two types of compensation strategies are superimposed: average error voltage compensation and optimal harmonic current injection strategy.

[0058] Speed ​​outer loop combined with field weakening control, input speed command With actual speed control The field weakening control outputs dq-axis current commands based on the speed difference modulation signal. , This enables a smooth switching between the constant torque region and the weak magnetic region.

[0059] In the inner current loop, the dq-axis current command from the speed loop is input. , With the actual dq-axis current of the feedback , The system is controlled by a PI controller, which adjusts the current deviation to zero steady-state error and generates a control voltage command. The feedforward decoupling module improves the system's dynamic response speed by injecting cross-coupling terms.

[0060] The voltage command in the dq coordinate system is converted to a voltage command in the stationary coordinate system. This voltage command is then superimposed with compensation voltages from two compensation strategies and used to generate the three-phase inverter circuit drive signal via SVPWM. The three-phase inverter circuit is driven by a bus voltage of... The DC power source provides power to drive the permanent magnet synchronous motor.

[0061] It is a mechanical angle that can be directly measured as the physical position of the motor rotor. Through The electrical angle used for coordinate transformation is indirectly obtained through pole-logarithmic conversion P. Average error voltage compensation is based on current polarity and electrical angle. Calculate the nonlinearity of the inverter Voltage error in coordinate system , ,exist It is superimposed on the voltage command in the coordinate system.

[0062] The optimal harmonic current injection strategy first involves controlling the three-phase current... , , The coordinates were transformed to the 5th and 7th synchronous dq coordinate systems using a synchronous rotating coordinate system transformation, and the 5th and 7th harmonic current components along the dq axis were extracted using a low-pass filter module. , , , In the harmonic voltage model with optimal reference current, the control target for the 5th and 7th harmonic current components is the optimal reference harmonic current. Based on the harmonic current error, the corresponding compensation voltage along the dq axis is calculated. , , , Finally, the compensation voltage under the dq axis... , , , Transformed to the reverse synchronous rotating coordinate system coordinate system obtained , This is then superimposed on the voltage command. The final result after superposition is... , .

[0063] During data acquisition and model training, the core objective is to iteratively optimize the weights and thresholds of the backpropagation (BP) model through PSO, using the root mean square error (RMSE) of the validation set as the fitness function to minimize the prediction error and ultimately output the optimal BP model parameters.

[0064] In the algorithm initialization phase, the particle swarm is initialized: PSO parameters such as particle swarm size, particle dimension, number of iterations, inertia weight and learning factor are set, and the initial position and velocity of the particles are randomly generated.

[0065] Construct the initial BP structure: Randomly generate the initial topology of the backpropagation neural network (number of nodes in the input layer, hidden layer, and output layer), and randomly initialize the weights and thresholds of the network.

[0066] In the fitness calculation and model training phase, the backpropagation neural network model is trained using the BP weights and thresholds corresponding to the current particle position. The root mean square error (RMSE) of the validation set is used as the fitness function of the particle to calculate the fitness value of the current particle and quantify the model's predictive performance.

[0067] In the error convergence judgment stage, the error of the current model is judged. If the RMSE of the validation set is less than the set threshold, it means that the model has reached the expected accuracy requirement, and the process directly jumps to the "output optimal parameters" stage. If the root mean square error (RMSE) does not meet the convergence condition, the particle swarm update process is entered. The convergence condition is met when the root mean square error is less than 0.0001.

[0068] During the particle swarm position and velocity update phase, a termination condition is determined for the iteration process. If the termination condition is met (such as reaching the maximum number of iterations or algorithm convergence), the loop is exited, and the globally optimal BP weights and threshold parameters are output. If the termination condition is not met, the process jumps back to the "training the backpropagation neural network" step and enters the next round of iteration optimization until the convergence or termination condition is met.

[0069] In experimental verification, a trained PSO-BP surrogate model was embedded into the DSP controller, replacing the dual-loop vector control and two types of compensation strategies. The control model was simplified to a 6-input, 2-output surrogate model. The inputs acquired the motor's operating data in real time, and the optimal control voltage was output in real time to suppress the motor's torque ripple. The control system includes: the surrogate model ( Figure 3 The surrogate model in the model consists of a PSO-BP neural network surrogate model, SVPWM (Space Vector Pulse Width Modulation), a three-phase inverter circuit, a permanent magnet synchronous motor (PMSM), coordinate transformation, and a feedback loop. The three-phase inverter circuit is based on a bus voltage of... The DC power source provides power to drive the permanent magnet synchronous motor.

[0070] The system collects real-time operating condition information such as motor speed, current, and rotor position. Simultaneously, it combines this information with known dead time and device characteristic parameters as input to a Particle Swarm Optimization (PSO-BP) neural network surrogate model. The surrogate model's input data is the d-axis current of the permanent magnet synchronous motor during operation. q-axis current Rotor position electrical angle Electric angular velocity Electromagnetic torque command and dead zone time The proxy model quickly calculates the input working condition characteristics. Optimal compensation voltage for shaft and The signal is then fed into the SVPWM module to generate a PWM signal to drive the inverter. The three-phase inverter circuit is powered by a bus voltage of... The DC power source provides power, outputting three-phase voltage. The motor generates three-phase current during operation. It is a mechanical angle that can be directly measured as the physical position of the motor rotor. Through The electrical angle indirectly obtained by pole-logarithm P conversion is used for coordinate transformation. After coordinate transformation, it is fed back to the surrogate model to form a closed loop and realize dynamic compensation.

[0071] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0072] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0073] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A method for suppressing torque ripple in a permanent magnet synchronous motor, characterized in that, The method includes: Step S1: Calculate the average error voltage caused by the nonlinear characteristics of the inverter, and obtain the initial compensation voltage based on the average error voltage; Step S2: Considering the air gap magnetic field distortion of the permanent magnet synchronous motor, the optimal harmonic current injection reference value is derived, and the harmonic reference voltage is obtained by adopting the optimal harmonic current injection control strategy. The optimal compensation voltage is calculated based on the initial compensation voltage and the harmonic reference voltage. Step S3: Collect the operating data of the permanent magnet synchronous motor with harmonic reference voltage compensation and optimal harmonic current injection to construct a particle swarm optimization backpropagation (PSO-BP) neural network surrogate model. Step S4: The trained PSO-BP neural network surrogate model is embedded into the field orientation control system of the permanent magnet synchronous motor to suppress torque pulsation.

2. The method of claim 1, wherein, In the step S1, the average error voltage The calculation is as follows: ； in, Dead time, For opening time, For shutdown delay time, For periodic time, For the voltage drop of the switching transistor, For diode voltage drop, This is the DC bus voltage.

3. The method of claim 2, wherein, Step S1 further includes: obtaining the average error voltage. coordinate system The initial compensation voltage for the shaft is calculated as follows: ； in, for The initial compensation voltage of the shaft, for The initial compensation voltage of the shaft, This refers to the phase current of phase A. This refers to the phase current of phase B. This refers to the C-phase current. It is a symbolic function.

4. The method of claim 3, wherein, The sign function The calculation is as follows: ； in, The electric angular velocity of the motor. The initial angle of the phase current. For time, The order of the harmonic current. For phase current, Values , , , Let A be the current in phase A of the three-phase system. Let B be the current in the three-phase system. This represents the current in phase C of the three-phase system.

5. The method of claim 4, wherein, Step S2 includes: Considering the air gap magnetic field distortion of the permanent magnet synchronous motor, and setting the superposition of the 6th harmonic terms in the dq axis flux linkage and current to zero, the optimal harmonic current injection reference value is derived. The optimal harmonic current injection reference value is injected by using a synchronous rotating PI controller to control the DC current, thereby obtaining the corresponding harmonic reference voltage in the dq coordinate system; The harmonic reference voltage in the dq coordinate system is transformed to [the desired value] using the Clark transform. coordinate system, to obtain Harmonic reference voltage in coordinate system; According to the initial compensation voltage of the axis, the harmonic reference voltage and the fundamental voltage of the axis, the optimal compensation voltage of the axis is calculated the axis.

6. The method of claim 5, wherein, In step S2, the optimal harmonic current injection reference value is calculated as follows: ； in, , These are the optimal harmonic current injection reference values ​​for the d-axis and q-axis, respectively. , and These represent the harmonic amplitudes of the fundamental, fifth, and seventh permanent magnet flux linkages, respectively. , The initial phase angles are those of the fifth and seventh harmonic components, respectively. , These are the fundamental currents along the d-axis and q-axis, respectively. For d-axis inductance, It is the q-axis inductance. The electric angular velocity of the motor. For time.

7. The method of claim 6, wherein the method further comprises: Step S2 further includes: injecting optimal harmonic current reference value and The synchronous rotating PI controller is adopted to control the direct current component, and the corresponding harmonic reference voltage in the dq coordinate system is obtained. The harmonic reference voltage is converted to [value] using the Clark transform. In coordinate system, we obtain Harmonic reference voltage in coordinate system Axial components and Axial components ; The average error voltage compensation and the optimal harmonic current reference value were injected to obtain the result. Shaft harmonic reference voltage compensation model, based on Shaft harmonic reference voltage compensation model, based on The initial compensation voltage and harmonic reference voltage of the shaft. Axial components and Axial components , The fundamental voltage of the axis is calculated. Optimal compensation voltage for shaft and .

8. The method of claim 7, wherein, axis optimal compensation voltage and acquiring includes: right The fundamental voltage of the shaft, The initial compensation voltage and harmonic reference voltage of the shaft. Axial components Summation yields Optimal compensation voltage for shaft ; right The fundamental voltage of the shaft, The initial compensation voltage and harmonic reference voltage of the shaft. Axial components Summation yields Optimal compensation voltage for shaft .

9. The method of claim 8, wherein, In step S3, the operating data of the permanent magnet synchronous motor includes the model input dataset and the model output dataset for the PSO-BP neural network surrogate model. The model input dataset includes the d-axis current during the operation of the permanent magnet synchronous motor. q-axis current Rotor position electrical angle Electric angular velocity Electromagnetic torque command and dead zone time The model output dataset includes Optimal compensation voltage for shaft and .

10. The method of claim 9, wherein, Step S3 further includes: Multiple groups were randomly selected to obtain the predicted optimal compensation voltage output by the PSO-BP neural network surrogate model under different operating conditions based on the input dataset; The optimal compensation voltage calculated based on steps S1 and S2 is taken as the actual optimal compensation voltage. The optimal network structure of the PSO-BP neural network surrogate model is determined with the goal of minimizing the root mean square error (RMSE) between the predicted optimal compensation voltage and the actual optimal compensation voltage. Among them, the PSO-BP neural network surrogate model with the optimal network structure is the trained PSO-BP neural network surrogate model.

Images