A torque control system for a permanent magnet synchronous motor

Abstract

The application discloses a kind of permanent magnet synchronous motor torque control systems. The optimal current instruction corresponding to the adjustment parameter of motor under each working condition is obtained by table lookup and given module, including given voltage amplitude, preset current instruction amplitude, preset current instruction angle, given mechanical power. The feedback voltage amplitude and feedback mechanical power are obtained by selecting corresponding signal through detection adjustment module, voltage amplitude calculation link and power prediction neural network model, which are compared with given voltage amplitude and given mechanical power respectively, and the compensation of preset current instruction angle and amplitude is output through PI controller. The application adopts data-driven mode to construct the nonlinear mapping of various effective signals to mechanical power parameters, and cooperates with the voltage and power collaborative control of preset current instruction double degree of freedom, which can effectively suppress the influence of temperature fluctuation, parameter uncertainty and other non-ideal factors on torque control accuracy during motor operation.

Description

Technical Field

[0001] This invention belongs to the field of permanent magnet synchronous motor control, and in particular relates to a torque control system to suppress the interference of non-ideal factors such as temperature fluctuations on torque control accuracy. This invention can be applied to various scenarios such as electric vehicle drive systems and automated industrial production. Background Technology

[0002] In recent years, the advanced manufacturing sector has experienced rapid development amidst the wave of electrification and intelligentization. In particular, the development of industries related to electric vehicles and automated industrial production has received high attention from both enterprises and governments. Enterprises have increased investment in related technology research and development, and governments have issued numerous preferential policies to promote technological innovation. During the promotion of electrification, permanent magnet synchronous motors have been widely used due to their significant advantages in efficiency, power density, and cost. Thanks to the efforts of many researchers, numerous innovative research and developments have emerged around them. Besides improving the design of the motor itself, effective control of the motor is also an important issue.

[0003] The design of electric drive systems for electric vehicles and automated industrial production require certain performance indicators for motor control in order to achieve their predetermined goals. Since a motor is a nonlinear system and there is a strong coupling between electromagnetic variables, it is very difficult to control and analyze it in the natural coordinate system. Therefore, some researchers have proposed a field-oriented control system, which transforms the motor model into a synchronous rotating coordinate system through coordinate transformation and controls the currents of the quadrature and direct axes respectively, thus realizing the control of a three-phase permanent magnet synchronous motor similar to a DC motor.

[0004] Within the framework of field-oriented control, in order to maximize the torque output per unit current and thus improve the operating efficiency of the motor, some researchers have proposed the MTPA control system. This system essentially transforms the determination of the AC and DC axis current commands into an optimization problem to be solved. It can be solved online using formulas, but due to the uncertainty of motor parameters and limited computing resources, the lookup table method is more commonly used in engineering. The optimal current command under various operating conditions is obtained through calibration experiments.

[0005] During motor operation, the magnetic field of the rotor permanent magnet cuts the stator winding, generating back electromotive force (EMF). The higher the speed, the greater the back EMF. The generation of back EMF limits the speed range of the permanent magnet motor. To address this, some scholars have proposed a field weakening control system, which introduces a negative direct-axis current to weaken the excitation magnetic field, thereby reducing the back EMF and breaking through the motor's base speed limit to achieve higher speed operation.

[0006] While achieving high torque control efficiency and a wide operating speed range, torque control accuracy is also a crucial motor performance indicator. Besides the inherent nonlinearity of the motor itself, it is also affected by non-ideal factors such as temperature fluctuations during operation. Machine learning and deep learning, from a data-driven perspective, have achieved nonlinear mapping between input and output, making significant progress in numerous research fields. This invention constructs a neural network model and applies it to the current command adjustment process of a permanent magnet motor to achieve high-precision torque control. Summary of the Invention

[0007] The purpose of this invention is to provide a torque control system for a permanent magnet synchronous motor. Based on the classic MTPA and field weakening control algorithms to determine the preset current command, this system introduces a neural network model for power prediction and designs a two-degree-of-freedom cooperative closed-loop control method to dynamically adjust the preset current command. This invention can suppress the influence of non-ideal factors such as temperature fluctuations on the torque control accuracy.

[0008] The objective of this invention is achieved through the following technical solution: a torque control system for a permanent magnet synchronous motor, which mainly consists of three modules: a lookup table and setpoint module, a detection and adjustment module, and a motor and drive module.

[0009] (1) Table lookup and given module

[0010] The input to the lookup table module is the given torque T. ref Measured inverter DC voltage V dc The measured speed n of the permanent magnet synchronous motor is given by the voltage vector amplitude u. sref Preset current command angle θ ref Preset current command amplitude I sref Given mechanical power P mref Where the given voltage vector magnitude u sref Preset current command angle θ ref and preset current command amplitude I sref These values ​​were obtained through separate lookup table models, which established the parameter values ​​corresponding to the optimal current command under each operating condition based on calibration experimental data. The operating conditions were determined using torque as the horizontal index and speed as the vertical index. Considering the fluctuation of the motor's power supply voltage, multiple DC voltages were selected during the calibration experiment; therefore, the vertical index covers the speed under different DC voltages. Given the mechanical power P... mref It is obtained through a given mechanical power calculation process, specifically calculated using formula (1) to obtain P. mref :

[0011]

[0012] (2) Detection and adjustment module

[0013] The detection and adjustment module receives four parameters mapped to each operating condition from the lookup table and the given module as reference values, namely the given voltage vector amplitude u. sref Preset current command angle θ ref Preset current command amplitude I sref Given mechanical power P mref Simultaneously obtain the direct-axis voltage u d Cross-axis voltage u q Feedback direct-axis current i d Feedback quadrature-axis current i q Temperature T of permanent magnet synchronous motor r To achieve the actual voltage vector magnitude u s and actual mechanical power P m Calculation and prediction. Among them, the voltage vector magnitude u... s The result is obtained by formula (2):

[0014]

[0015] Actual mechanical power P m The power prediction neural network model was used. This model uses eight parameters as input layer variables, which are given torque T. ref Measured inverter DC voltage A dc Measured speed n and direct-axis voltage u of permanent magnet synchronous motor d Cross-axis voltage u q Feedback direct-axis current i d Feedback quadrature-axis current i q and the temperature T of the permanent magnet synchronous motor r With actual mechanical power P m As an output layer variable, two hidden layers are constructed between the input and output layers. A dataset is built based on the data from the calibration experiment, and the constructed neural network model is trained, its hyperparameters are optimized and verified, and its power prediction performance is tested.

[0016] The actual voltage vector amplitude u obtained from the voltage amplitude calculation step s With a given voltage vector magnitude u sref The comparison is performed and the preset current command angle θ is obtained through the PI controller. ref The compensation angle Δθ; the actual mechanical power P obtained from the power prediction neural network model. m With a given mechanical power P mref The values ​​are compared and processed by the PI controller to obtain the preset current command amplitude I. sref The compensation amplitude ΔI s The preset current command is compensated for angle and amplitude to obtain the actual current command amplitude I.s And the actual current command angle θ, and then the current command is calculated using equation (3) to obtain the direct axis current setpoint i in the synchronous rotating coordinate system. dref quadrature axis current setpoint i qref .

[0017]

[0018] Direct-axis current setpoint i dref With feedback direct-axis current i d The direct-axis voltage u is obtained by comparison and processing via a PI controller. d quadrature axis current given value i qref With feedback cross-axis current i q The cross-axis voltage u is obtained by comparing and passing through a PI controller. q The direct-axis voltage u d u q It is then used as the input for the motor and drive module.

[0019] (3) Motor and drive module

[0020] The motor and drive module detect the direct-axis voltage u output by the regulating module. d and quadrature axis voltage u q As input, the rotor angle θ fed back from the speed sensor r The voltage u in the stationary coordinate system is obtained by performing an inverse Park transformation according to formula (4). α and u β .

[0021]

[0022] u α and u β The input is fed to the SVPWM module for pulse width modulation to generate six pulse signals to drive the inverter. The inverter generates equivalent three-phase voltage and current, which are applied to the permanent magnet synchronous motor to achieve its operation. The current sensor monitors the three-phase current i. a i b i c The rotor angle θ fed back by the speed sensor is detected. r The current i in the synchronous rotating coordinate system is obtained by performing Clark-Park transformation according to formula (5). d andi q .

[0023]

[0024] The beneficial effects of this invention are as follows:

[0025] (1) The control system described in this invention adopts a data-driven approach to establish a nonlinear mapping from various signals to mechanical power by building a neural network model. Compared with the method of using theoretical models to calculate various motor losses to obtain mechanical power, it gets rid of complex mechanism analysis and has a stronger adaptability to changes in motor parameters.

[0026] (2) The control system described in this invention adopts voltage and power coordinated control, and realizes dynamic adjustment of two degrees of freedom in the current vector space. When non-ideal factors such as temperature fluctuations interfere, the preset current command angle and amplitude can be corrected simultaneously according to the feedback signal, and the precise control of motor torque has strong robustness. Attached Figure Description

[0027] Figure 1 This is a structural diagram of the torque control system for a permanent magnet synchronous motor;

[0028] Figure 2 It is a power prediction neural network model; Detailed Implementation

[0029] The present invention will now be described in further detail with reference to the accompanying drawings:

[0030] This invention discloses a torque control system for a permanent magnet synchronous motor, the structure of which is shown in the figure below. Figure 1 As shown, the system mainly includes a lookup table and setpoint module, a detection and adjustment module, and a motor and drive module. Before using this system for torque control, calibration experiments are required to obtain basic data, and two preliminary tasks need to be completed: establishing a lookup table model and establishing a power prediction neural network model.

[0031] The calibration experiment, based on the MTPA and field weakening control principles, involved a fixed-point scanning experiment in the current vector space. Considering the fluctuations in the inverter's supply voltage, multiple DC-side voltage values ​​were selected for the corresponding experimental process. Ultimately, the optimal current command for the permanent magnet synchronous motor under various operating conditions was determined, and the corresponding basic data were recorded, including: the motor torque T measured by the torque sensor during calibration, and the measured inverter DC voltage V. dc Measured speed n and temperature T of the permanent magnet synchronous motor r Direct-axis voltage u d Cross-axis voltage u q Feedback direct-axis current i d Feedback quadrature-axis current i q Temperature T of permanent magnet synchronous motor r and the actual feedback mechanical power P of the dynamometer mtest .

[0032] To establish a lookup table model for the voltage amplitude parameters corresponding to the optimal current command under different operating conditions of a permanent magnet synchronous motor, a reasonable data granularity is set based on the processor's computing and storage performance. The lookup torque is set with the motor torque T as the lateral index and a certain step size, and different DC voltages V are used as the lookup parameters. dc The motor speed n is used as the vertical index, and the lookup speed is set with a certain step size. The u corresponding to each two-dimensional lookup index point is obtained by linear interpolation. d u q The value is obtained by calculating the corresponding voltage amplitude parameter u using equation (2). s .

[0033] To establish a lookup table model for the preset current command amplitude and angle parameters corresponding to the optimal current command under different operating conditions of a permanent magnet synchronous motor, a reasonable data granularity is set according to the processor's computing and storage performance. The lookup torque is set with the motor torque T as the lateral index and a certain step size, and different DC voltages V are used as the lookup parameters. dc The motor speed n is used as the vertical index, and the lookup speed is set with a certain step size. The i corresponding to each two-dimensional lookup index point is obtained by linear interpolation. d i q The value is obtained by calculating the corresponding preset current command amplitude parameter I using equations (6) and (7). s , Angle parameter θ.

[0034]

[0035]

[0036] To establish a power prediction neural network model, calibration experiments were conducted on multiple motors to obtain basic data and build a dataset. The dataset was then divided into training, validation, and test sets in a 7:2:1 ratio. Figure 2 As shown, a neural network input layer is constructed with the following 8 parameters as input: given torque T ref Measured DC voltage V of inverter dc Measured speed n and direct-axis voltage u of permanent magnet synchronous motor d Cross-axis voltage u q Feedback direct-axis current i d Feedback quadrature-axis current i q and the temperature T of the permanent magnet synchronous motor r Build a system based on actual mechanical power P m The output layer of the neural network serves as the output layer, with two hidden layers built between the input and output layers. The neural network model is trained using training data, and hyperparameters such as the number of neurons in the hidden layers are optimized using validation data. Finally, the model's prediction accuracy is validated using a test set, and the superior model is selected for application in the torque control system proposed in this invention.

[0037] After establishing the lookup table model and the power prediction neural network model, the permanent magnet synchronous motor torque control system involved in this invention is implemented as follows: Figure 1 The three modules shown process and transmit the corresponding signals.

[0038] The lookup table and given module receives the following three signals: given torque T ref Measured inverter DC voltage V dc The measured speed n of the permanent magnet synchronous motor is used to transmit the above signal to the voltage amplitude parameter lookup model, the preset current command amplitude, and the angle parameter lookup model. After lookup and linear interpolation, the given voltage vector amplitude u is obtained. sref Preset current command angle θ ref Preset current command amplitude I sref The given mechanical power P is calculated according to formula (1). mref .

[0039] The detection and adjustment module receives the given parameter signal from the lookup table and the given parameter output from the given parameter module: given voltage vector amplitude u. sref Preset current command angle θ ref Preset current command amplitude I sref Given mechanical power P mref . will u sref The actual voltage amplitude u fed back from the voltage amplitude calculation stage s Comparison is made by outputting the PI controller to θ. ref The compensation angle Δθ will P mref The actual mechanical power P fed back by the power prediction neural network model m Comparison, via the PI controller output to I sref The compensation amplitude ΔI s The voltage amplitude calculation step uses the direct-axis voltage u. d Cross-axis voltage u q u is obtained by calculation according to formula (2) s The power prediction neural network model is used to predict the T value during the control process. ref V dc 、n、u d u q i d i q T r P obtained by input prediction of the signal m The actual current command amplitude I after angle and current compensation. s The actual current command angle θ is calculated according to formula (3) to obtain the direct-axis current setpoint i in the synchronous rotating coordinate system. dref quadrature axis current setpoint iqref i dref The direct-axis current i fed back from the motor and drive module d After comparison, the PI controller outputs a direct-axis voltage to control the vector u. d i qref The quadrature-axis current i fed back from the motor and drive module q After comparison, the quadrature-axis voltage output by the PI controller controls the vector u. q .

[0040] The motor and drive module receives the AC and DC axis voltage control vector signals output by the detection and adjustment module: u d u q And acquire the rotor speed signal θ fed back by the speed sensor. r Using equation (4), the inverse Park transformation is performed to obtain the voltage component u in the stationary coordinate system. α and u β The two voltage vector signals are then transmitted to SVPWM for pulse width modulation, outputting six pulse signals that apply to each phase arm of the inverter. The inverter outputs equivalent three-phase voltage and current, which are then applied to the permanent magnet synchronous motor. A current sensor collects the three-phase current i. a i b i c Rotor speed signal θ fed back from the speed sensor r The Clark-Park transformation is performed using equation (5) to obtain the quadrature and direct axis current feedback value i in the synchronous rotating coordinate system. d i q The feedback value is then transmitted to the detection and adjustment module.

[0041] This invention is not limited to the embodiments described above. All other embodiments obtained by those skilled in the art without creative effort, using the same or similar methods as the embodiments described above, are within the protection scope of this invention.

Claims

1. A torque control system for a permanent magnet synchronous motor, characterized in that: The control system includes a lookup and setpoint module, a detection and adjustment module, and a voltage and drive module. The setpoint torque, the measured inverter DC voltage, and the measured permanent magnet synchronous motor speed are used as inputs to the lookup and setpoint module. After lookup and interpolation, the setpoint voltage vector amplitude parameter, the preset current command amplitude, and the angle parameter corresponding to the optimal current command under each operating condition are obtained. The setpoint mechanical power parameter is calculated by formula and then input to the detection and adjustment module. The detection and adjustment module acquires the direct and quadrature axis voltage vectors in a synchronous rotating coordinate system. After voltage amplitude calculation, it obtains the feedback voltage vector amplitude and compares it with the given voltage vector amplitude parameters transmitted by the lookup table and the given module. The PI controller outputs the compensation angle for the preset current command. This module also acquires eight signals, including the given torque, measured inverter DC voltage, measured permanent magnet synchronous motor speed, direct axis voltage, quadrature axis voltage, feedback direct axis current, feedback quadrature axis current, and permanent magnet synchronous motor temperature, and inputs them into the power prediction neural network model to obtain the feedback mechanical power. This feedback mechanical power is compared with the given mechanical power parameters transmitted by the lookup table and the PI controller outputs the compensation amplitude for the preset current command. After compensation, the angle and amplitude of the actual current command are used to obtain the direct-axis current command value and quadrature-axis current command value in the synchronous rotating coordinate system through the current command calculation step. The given values ​​of the quadrature and direct axis currents are compared with the feedback quadrature and direct axis currents transmitted by the motor and drive module, respectively. The quadrature and direct axis voltage vectors are output by the PI controller and then input to the motor and drive module. The motor and drive module receives the quadrature and direct-axis voltage vectors input from the detection and regulation module. Together with the rotor speed feedback from the speed sensor, they undergo an inverse Park transformation to obtain the voltage vector in the stationary coordinate system, which is then applied to the SVPWM module for pulse width modulation. Each phase bridge arm of the inverter is subjected to six pulses, outputting equivalent voltage and current to drive the permanent magnet synchronous motor. The current sensor detects the three-phase current and, together with the rotor speed, undergoes a Clark-Park transformation to obtain the quadrature and direct-axis voltage vectors in the synchronous rotating coordinate system, which are then fed back to the detection and regulation module.

2. The torque control system for a permanent magnet synchronous motor as described in claim 1, characterized in that, A power prediction neural network is constructed to detect and provide feedback on actual mechanical power. This neural network model uses given torque, measured inverter DC voltage, measured permanent magnet synchronous motor speed, direct-axis voltage, quadrature-axis voltage, feedback direct-axis current, feedback quadrature-axis current, and permanent magnet synchronous motor temperature as input layer variables, and feedback mechanical power as output layer variables. Hidden layers are introduced to establish connections, enabling a nonlinear mapping of multiple signals to feedback mechanical power. A dataset is constructed based on extensive calibration experimental data from multiple motors, and divided into training, validation, and test sets according to a certain ratio. The training set data is used to train the model to converge, the validation set data is used for hyperparameter optimization, and the test set data is used to select models with high actual prediction accuracy to test the permanent magnet synchronous motor torque control system.

3. The torque control system for a permanent magnet synchronous motor as described in claim 1, characterized in that, The amplitude and angle of the preset current command are dynamically corrected simultaneously through coordinated control of mechanical power and voltage amplitude, specifically as follows: The feedback mechanical power output by the power prediction neural network model is compared with the given mechanical power calculated by the formula, and the preset current command amplitude is compensated by the PI controller. The feedback voltage vector amplitude output by the voltage amplitude calculation stage is compared with the given voltage vector amplitude output by the lookup table module, and the preset current command angle is compensated by the PI controller.

Images