Non-cascaded model-free predictive current control method for axial flux switched permanent magnet motor, medium and equipment

Abstract

This invention discloses a non-cascaded model-free predictive current control method, medium, and device for axial flux switching permanent magnet motors, belonging to the field of electromechanical control. The method is based on non-cascaded model-free predictive current control, coordinate transformation, multi-state variable ESO considering load disturbances, and SVPWM modulation. It constructs a multi-state variable ESO including current, disturbance, and speed to estimate internal and external disturbances in real time, uses a gradient descent algorithm to adaptively optimize the ESO gain parameters, and combines a disturbance compensation mechanism to construct a discrete hyperlocal model. This eliminates the traditional cascaded structure and PI controller, thus removing dynamic coupling problems. This invention significantly improves the robustness, control accuracy, and anti-interference capability of the AFFSPMM control system, adapts to strong disturbance scenarios in the ocean, and provides a practical optimization strategy for high-performance motor control.

Description

Technical Field

[0001] This invention relates to the field of electromechanical control, and in particular to a non-cascaded model-free predictive current control method for axial flux switching permanent magnet motors. Background Technology

[0002] In recent years, with the rapid development of marine resource development and ship propulsion technology, shaftless rim propulsion motors have gradually become an ideal choice for propulsion systems of marine technical equipment due to their advantages such as compact structure, high propulsion efficiency, and low fluid noise. Axial field flux-switching permanent magnet motors (AFFSPMMs) are suitable for shaftless rim propulsion motors due to their high torque density, high efficiency, and good heat dissipation performance.

[0003] However, in a marine environment, various unknown disturbances (such as frictional torque and load torque) may occur, leading to load instability. This unstable load can cause system parameter mismatch, affecting the motor's speed and current, and consequently impacting the performance of the entire control system. Furthermore, even without considering load disturbances, the speed loop still uses a PI controller to process the speed error to obtain the quadrature-axis current setpoint i. q * PI controllers themselves have bandwidth limitations, and using the same set of PI parameters under various operating conditions will lead to a decrease in control performance.

[0004] To improve disturbance rejection capability, Extended State Observer (ESO) has been introduced into motor control. Its core idea is to treat unmodeled internal dynamics and external disturbances as a "total disturbance" and estimate and compensate for them in real time through state observation. However, traditional ESO parameters (such as error feedback gains β1, β2, and β3) are typically fixed values ​​or empirically adjusted, making it difficult to adapt to dynamic changes under different operating conditions. Especially under drastic load fluctuations, fixed-parameter ESOs may lead to decreased estimation accuracy, thus affecting control performance. Summary of the Invention

[0005] Purpose of the invention: In view of the above-mentioned prior art, a non-cascaded model-free predictive current control method for axial flux switching permanent magnet motors is proposed, which can effectively solve the problems of complex ESO calculation and poor real-time performance, while improving the robustness and dynamic performance of the motor control system under complex disturbance environments.

[0006] Technical solution: A non-cascaded model-predictive current control method for axial flux-switching permanent magnet motors, comprising the following steps: First, obtaining the rotor position angle estimate through sensorless control. Then, the estimated value of the motor speed is calculated. ;From the three-phase current signal of the motor The coordinates are obtained in the stationary reference coordinate system after Clarke transformation. , Axis current components Then, the d-axis and q-axis current components in the rotating coordinate system are obtained by Park coordinate transformation. ;Will , and The input is fed into an extended state observer that considers load disturbances, and the load disturbances are observed respectively. Lumped disturbances of the current along the d-axis and q-axis Then, the rotor position information is fused to drive a non-cascaded model-free predictive current control strategy to obtain the q-axis predicted current. And calculate the reference values ​​of the d-axis and q-axis stator voltages. and ; and Obtained through coordinate transformation , Shaft voltage reference value and The six pulse waveforms obtained through SVPWM are then applied to the inverter to control the operation of AFFSPMM.

[0007] Furthermore, the extended state observer considering load disturbances uses the d-axis current component... q-axis current component d-axis lumped disturbance q-axis lumped disturbance and motor speed Let be the state variables, and the state equation be expressed as:

[0008]

[0009]

[0010]

[0011] In the formula, , These are the current errors along the d and q axes, respectively. This refers to the rotational speed error; This represents the estimated value of the d-axis current; This represents the estimated value of the q-axis current; , These are the rates of change of the d-axis and q-axis currents, respectively. Indicates the rate of change of rotational speed; These represent the observer's observations of the d-axis and q-axis current disturbances, respectively. The observer's observation of the rotational speed; , These represent the rates of change of the d-axis and q-axis current disturbances, respectively. This represents the rate of change of the rotational speed disturbance; Indicates the stator voltage coefficient of the motor; The d-axis stator voltage component; is the q-axis stator voltage component; g is the design parameter. , For rotational inertia, This represents the number of pole pairs in an axial flux-switched permanent magnet motor. For permanent magnet flux linkage; , , These are the three error feedback gains of the observer.

[0012] Furthermore, the gradient descent algorithm is used to adaptively solve for and optimize the error feedback gain of the extended state observer. , , The objective function is the sum of squared estimation errors of the extended state observer. The objective function is calculated iteratively. , , The partial derivatives are calculated and the parameters are dynamically updated until the convergence criterion is met, at which point the optimal value is output. , , .

[0013] Furthermore, the specific iterative process of the gradient descent algorithm includes:

[0014] S31: Initialization parameters , , And set the initial learning rate. ;

[0015] S32: Calculate the objective function under the current parameters. The value of , and the objective function right , , partial derivatives , , ;

[0016] S33: Update the formula according to gradient descent. , , Perform iterative updates to obtain the first... Next iteration value , , ;

[0017] S34: According to the updated , , Recalculate the objective function value Determine the updated objective function value Compared with the previous objective function value Is the change less than the set value? If so, stop iterating and output. , , Find the optimal value; otherwise, return to S32 and continue iterating.

[0018] Furthermore, the method for constructing the extended state observer that considers load disturbances includes the following steps:

[0019] Step 1: Establish a hyperlocal mathematical model of the axial flux switching permanent magnet motor, and use the d-axis current component as the model. q-axis current component and d-axis lumped disturbance q-axis lumped disturbance Using current error as the feedback, a linear extended state observer with current error feedback is constructed, and the state equation is expressed as:

[0020]

[0021]

[0022]

[0023] Step 2: Based on the motor motion equations, treat the load torque as a load disturbance. And as an independent new state variable, construct with , , , , To extend the state observer to a multi-state variable state variable, enabling online estimation of lumped disturbances along the d-axis and q-axis, as well as rotational speed disturbances;

[0024] Step 3: Classify the motor body and parameter mismatch as a lumped disturbance. Using the disturbance estimation results of the multi-state variable extended state observer, construct a discrete hyperlocal model based on disturbance compensation, specifically expressed as:

[0025]

[0026]

[0027]

[0028] In the formula, Indicates the discrete time step number; The sampling period; This represents the equivalent disturbance of the d-axis current. This represents the equivalent disturbance of the q-axis current. This represents an estimated value of the rotational speed. The equivalent disturbance representing the rotational speed is the load disturbance observed by the extended state observer that takes load disturbance into account.

[0029] Furthermore, the fused rotor position information drives a non-cascaded model-free predictive current control strategy to obtain the q-axis predicted current. And calculate the reference values ​​of the d-axis and q-axis stator voltages. and Specifically, it is expressed as:

[0030]

[0031] In the formula, Indicates the discrete time step number; This is the reference value for the d-axis current. This is a reference value for the motor speed; The sampling period.

[0032] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors.

[0033] A computer-readable storage medium storing the method for performing the non-cascaded model-free predictive current control of the axial flux-switched permanent magnet motor.

[0034] Beneficial effects: To address the load disturbances and parameter uncertainties faced by motor control systems under complex operating conditions, this invention provides a non-cascaded model-free predictive current control method for axial flux switching permanent magnet motors. This method constructs an extended state observer containing multiple state variables such as current, disturbance, and speed to estimate and compensate for internal and external disturbances in the system in real time, effectively solving the performance problems caused by the limited bandwidth and fixed parameters of traditional PI controllers.

[0035] The gradient descent algorithm is used to adaptively optimize the observer gain parameters. With minimizing the sum of squared estimated errors as the objective function, the partial derivatives are calculated iteratively and the parameters are dynamically adjusted, enabling the system to automatically adapt to sudden load changes and parameter variations, which significantly improves control accuracy and anti-interference capability.

[0036] Compared to traditional methods, this method eliminates the dynamic coupling problem in cascaded control through a non-cascaded structure and disturbance compensation mechanism; by solving the adaptive coupling gain coefficient, it effectively reduces the impact of rotor position error and nonlinear disturbance on motor speed and current, and has strong robustness.

[0037] This method is applicable to motor control in highly disturbed environments such as marine environments, and provides a novel and practical parameter optimization strategy for high-performance motor control systems. Attached Figure Description

[0038] Figure 1 This is a block diagram of the control system corresponding to the method in the embodiment of the present invention;

[0039] Figure 2 This is a structural block diagram of the extended state observer in this invention;

[0040] Figure 3 This is a flowchart of the adaptive solution of gain parameters in this invention. Detailed Implementation

[0041] The invention will now be further explained with reference to the accompanying drawings.

[0042] A non-cascaded model-predictive current control method for axial flux-switching permanent magnet motors, such as... Figure 1 As shown, it consists of non-cascaded model-free predictive current control (N_MFPCC), inverse Park coordinate transformation, SVPWM (space vector pulse width modulation), three-phase inverter, axial flux-switched permanent magnet motor (AFFSPMM), Clarke coordinate transformation, Park coordinate transformation, and extended state observer (ESO) considering load disturbances.

[0043] First, the rotor position angle estimate is obtained through sensorless control. Then, the estimated value of the motor speed is calculated. Combined with the acquired phase current signal After Clarke coordinate transformation, the coordinates in the stationary reference system are obtained. , Axis current components Furthermore, it is transformed into d-axis and q-axis current components in a rotating coordinate system via Park coordinate transformation. Subsequently, based on the extended state observer (ESO) considering load disturbances, the lumped disturbances on the d-axis and q-axis are... and load disturbance Perform online estimation.

[0044] Among them, a mathematical model is performed on the axial flux switching permanent magnet motor to achieve the following: and Using current error as feedback and the state variable as the input, a linear extended state observer is established. The specific block diagram is shown below. Figure 2 As shown. To implement an extended state observer that considers the load, the state equations need to be discretized to obtain the optimal reference control quantity. .

[0045] The specific steps of the method in this embodiment are as follows:

[0046] Step 1: Establish the hyperlocal mathematical model of AFFSPMM:

[0047] (1)

[0048] Equation (1) can be rewritten in matrix form as follows:

[0049] (2)

[0050] In the formula, The current component is the d-axis component; This refers to the q-axis current component. The d-axis stator voltage component; This refers to the q-axis stator voltage component. The d-axis lumped disturbance of the axial flux switching permanent magnet motor; The q-axis lumped disturbance of an axial flux-switched permanent magnet motor. It is the d-axis inductance; It is the q-axis inductance; Stator resistance; For stator inductance; Electric angular velocity; For permanent magnet flux linkage; The d-axis stator voltage coefficient; This is the q-axis stator voltage coefficient.

[0051] Based on the hyperlocal mathematical model of AFFSPMM, an extended state observer with current error feedback is constructed:

[0052] (3)

[0053] (4)

[0054] (5)

[0055] In the formula, The current error along the d-axis; This represents the current error along the q-axis. This represents the estimated value of the d-axis current; This represents the estimated value of the q-axis current; , These are the rates of change of the d-axis and q-axis currents, respectively. These represent the observer's observations of the d-axis and q-axis current disturbances, respectively. This represents the stator voltage coefficient of the motor, and is generally taken as the reciprocal of the stator inductance. , These represent the rates of change of the d-axis and q-axis current disturbances, respectively. and These are the two error feedback gains of the observer.

[0056] The state-space model of the extended state observer is as follows: Figure 2 As shown, the state equation and output equation can be written as:

[0057] (6)

[0058] In the formula, These are estimates of the system state. They represent rate of change, For reference current input, To control the voltage input.

[0059] Step 2: Under load disturbances (such as load changes caused by uneven propeller force), motor operation may be affected by the disturbance. To cope with load disturbances, the extended state observer needs to treat the load torque as a load disturbance and estimate and compensate for it as a new state variable.

[0060] The equation of motion for the electric motor can be written as:

[0061] (7)

[0062] In the formula, This refers to the motor's rotational speed; This represents the number of pole pairs in an axial flux-switched permanent magnet motor. B is the magnetic flux linkage of the permanent magnet; B is the coefficient of viscous friction. This is the load torque; Let be the moment of inertia.

[0063] The extended state observer will handle load disturbances. Treat it as an independent new state variable and incorporate it into the state estimation equation. Construct a... , , Extended state observer for state variables:

[0064] (8)

[0065] (9)

[0066] (10)

[0067] In the formula, , These are the current errors along the d and q axes, respectively. This refers to the rotational speed error; This represents the estimated value of the d-axis current; This represents the estimated value of the q-axis current; This represents an estimated value of the rotational speed. Indicates the motor speed; , These are the rates of change of the d-axis and q-axis currents, respectively. Indicates the rate of change of rotational speed; These represent the observer's observations of the d-axis and q-axis current disturbances, respectively. The observer's observation of the rotational speed; , These represent the rates of change of the d-axis and q-axis current disturbances, respectively. This represents the rate of change of the rotational speed disturbance; g is the design parameter. ; This represents the error feedback gain associated with load disturbance.

[0068] Step 3: Due to nonlinear load disturbances under complex operating conditions, rotor position and parameter errors will affect... , and There is a certain coupling relationship among the three, therefore an adaptive solution for adjusting the gain coefficient is adopted. , and First, define the objective function, J, which is defined as the sum of squared estimation errors of the extended state observer:

[0069] (11)

[0070] In the formula, y represents the actual output of the observer; This represents the estimated output of the observer.

[0071] For the whole , and Find the partial derivatives:

[0072] (12)

[0073] Then, β is updated using gradient descent, with the following update formula:

[0074] (13)

[0075] in, It is the learning rate; , , The first In the next iteration update , , The value of . For example Figure 3 As shown, the specific iteration method is as follows: First, initialize the parameters. , , And set an initial learning rate, typically 0.01 or 0.05; under the current parameters, calculate the value of the objective function J and its relation to... , , The partial derivatives are updated according to the gradient descent formula. , , The value of .

[0076] According to the updated , , Recalculate the objective function value and check the updated objective function value. Compared with the previous objective function value Is the change less than the set value? That is, whether it satisfies If the value is less than the set value, the iteration stops and the optimal parameter is output; otherwise, the iteration continues, thus adaptively adjusting the parameter. , , .

[0077] Step 4: Classify potential parameter mismatches between the motor body and the system as lumped disturbances. Use an extended state observer to estimate these disturbances in real time and incorporate them as compensation variables into the system. The discrete hyperlocal model based on disturbance compensation is constructed as follows:

[0078] (14)

[0079] (15)

[0080] (16)

[0081] In the formula, Indicates the discrete time step number; The sampling period; This represents the equivalent disturbance of the d-axis current. This represents the equivalent disturbance of the q-axis current, used to compensate for parameter perturbations in the d- and q-axis currents; This represents an estimated value of the rotational speed. The equivalent disturbance representing the rotational speed is the load disturbance observed by the extended state observer considering the load disturbance, used to compensate for the influence of factors such as load disturbance on the rotational speed.

[0082] According to the control box Figure 1 A simulation model of the control system was established using Matlab software. First, the estimated rotor position angle was obtained through sensorless control. Then, the estimated value of the motor speed is calculated. ;From the three-phase current signal of the motor The coordinates are obtained in the stationary reference coordinate system after Clarke transformation. , Axis current components Then, the d-axis and q-axis current components in the rotating coordinate system are obtained by Park coordinate transformation. ;Will , and The load disturbance is input into an extended state observer that considers load disturbances. The extended state observers designed using equations (8) to (10) observe the load disturbances respectively. Lumped disturbances of the current along the d-axis and q-axis Then, the rotor position information is fused to drive a non-cascaded model-free predictive current control strategy, and the q-axis predicted current is obtained by equation (17). And calculate the reference values ​​of the d-axis and q-axis stator voltages. and ; and Obtained through coordinate transformation , Shaft voltage reference value and The six pulse waveforms obtained through SVPWM are then applied to the inverter to control the operation of AFFSPMM.

[0083] Deadbeat predictive current control fails to adequately consider the impact of parameter variations and various disturbances on the accuracy of the motor model, leading to a deviation between the control model and the actual motor mathematical model. This invention treats the motor itself and its potential parameter mismatches as lumped disturbance terms and estimates these disturbances in real time using an extended state observer to dynamically compensate the motor model, thereby effectively reducing the impact of disturbances on the control system performance. This allows for the acquisition of the optimal reference control quantity. as follows:

[0084] (17)

[0085] In the formula, , These are the reference values ​​for the d-axis and q-axis stator voltages, respectively. , These are the reference values ​​for the d-axis and q-axis currents, respectively. This is a reference value for the motor speed. This is the stator voltage coefficient of the motor.

[0086] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors.

[0087] A computer-readable storage medium storing a non-cascaded model-predictive current control method for an axial flux-switched permanent magnet motor, as described above.

[0088] Through the above steps, the non-cascaded model-predictive current control strategy based on axial flux switching permanent magnet motor improves the robustness and wide speed range stability of sensorless control, avoids the dynamic interaction problem between speed loop and current loop in traditional cascaded control, quickly responds to load changes and parameter mismatch, reduces coupling effects, and improves the dynamic performance of the system.

Claims

1. A non-cascaded model-predictive current control method for axial flux-switching permanent magnet motors, characterized in that, Includes the following steps: First, the rotor position angle estimate is obtained through sensorless control. Then, the estimated value of the motor speed is calculated. ;From the three-phase current signal of the motor The coordinates are obtained in the stationary reference coordinate system after Clarke transformation. , Axis current components Then, the d-axis and q-axis current components in the rotating coordinate system are obtained by Park coordinate transformation. ;Will , and The input is fed into an extended state observer that considers load disturbances, and the load disturbances are observed respectively. Lumped disturbances of the current along the d-axis and q-axis Then, the rotor position information is fused to drive a non-cascaded model-free predictive current control strategy to obtain the q-axis predicted current. And calculate the reference values ​​of the d-axis and q-axis stator voltages. and ; and Obtained through coordinate transformation , Shaft voltage reference value and The six pulse waveforms obtained through SVPWM are then applied to the inverter to control the operation of AFFSPMM.

2. The non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors according to claim 1, characterized in that, The extended state observer considering load disturbances uses the d-axis current component. q-axis current component d-axis lumped disturbance q-axis lumped disturbance and motor speed Let be the state variables, and the state equation be expressed as: In the formula, , These are the current errors along the d and q axes, respectively. This refers to the rotational speed error; This represents the estimated value of the d-axis current; This represents the estimated value of the q-axis current; , These are the rates of change of the d-axis and q-axis currents, respectively. Indicates the rate of change of rotational speed; These represent the observer's observations of the d-axis and q-axis current disturbances, respectively. The observer's observation of the rotational speed; , These represent the rates of change of the d-axis and q-axis current disturbances, respectively. This represents the rate of change of the rotational speed disturbance; Indicates the stator voltage coefficient of the motor; The d-axis stator voltage component; is the q-axis stator voltage component; g is the design parameter. , For rotational inertia, This represents the number of pole pairs in an axial flux-switched permanent magnet motor. For permanent magnet flux linkage; , , These are the three error feedback gains of the observer.

3. The non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors according to claim 2, characterized in that, The gradient descent algorithm is used to adaptively solve and optimize the error feedback gain of the extended state observer. , , The objective function is the sum of squared estimation errors of the extended state observer. The objective function is calculated iteratively. , , The partial derivatives are calculated and the parameters are dynamically updated until the convergence criterion is met, at which point the optimal value is output. , , .

4. The non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors according to claim 3, characterized in that, The specific iterative process of the gradient descent algorithm includes: S31: Initialize parameters , , And set the initial learning rate. ; S32: Calculate the objective function under the current parameters. The value of , and the objective function right , , partial derivatives , , ; S33: Update the formula according to gradient descent. , , Perform iterative updates to obtain the first... Next iteration value , , ; S34: According to the updated , , Recalculate the objective function value Determine the updated objective function value Compared with the previous objective function value Is the change less than the set value? If so, stop iterating and output. , , Find the optimal value; otherwise, return to S32 and continue iterating.

5. The non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors according to claim 2, characterized in that, The method for constructing the extended state observer that considers load disturbances includes the following steps: Step 1: Establish a hyperlocal mathematical model of the axial flux switching permanent magnet motor, and use the d-axis current component as the model. q-axis current component and d-axis lumped disturbance q-axis lumped disturbance Using current error as the feedback, a linear extended state observer with current error feedback is constructed, and the state equation is expressed as: Step 2: Based on the motor motion equations, treat the load torque as a load disturbance. And as an independent new state variable, construct with , , , , To extend the state observer to a multi-state variable state variable, enabling online estimation of lumped disturbances along the d-axis and q-axis, as well as rotational speed disturbances; Step 3: Classify the motor body and parameter mismatch as a lumped disturbance. Using the disturbance estimation results of the multi-state variable extended state observer, construct a discrete hyperlocal model based on disturbance compensation, specifically expressed as: In the formula, Indicates the discrete time step number; The sampling period; This represents the equivalent disturbance of the d-axis current; This represents the equivalent disturbance of the q-axis current. This represents an estimated value of the rotational speed. The equivalent disturbance representing the rotational speed is the load disturbance observed by the extended state observer that takes load disturbance into account.

6. The non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors according to any one of claims 2-5, characterized in that, The fused rotor position information drives a non-cascaded model-free predictive current control strategy to obtain the q-axis predicted current. And calculate the reference values ​​of the d-axis and q-axis stator voltages. and Specifically, it is expressed as: In the formula, Indicates the discrete time step number; This is the reference value for the d-axis current. This is a reference value for the motor speed; The sampling period.

7. A computer device, characterized in that: The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the non-cascaded model-free predictive current control method for axial flux-switching permanent magnet motors as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a method for non-cascaded model-predictive current control of an axial flux-switched permanent magnet motor as described in any one of claims 1-6.

Images