A pre-optimization model-free adaptive control method for permanent magnet synchronous motor

Abstract

The application provides a permanent magnet synchronous motor preposition optimization model-free adaptive control method, which firstly establishes a discrete time nonlinear system for the permanent magnet synchronous motor and determines input and output parameters, and then processes the parameters by using a full-format dynamic linearization data model. Subsequently, the next time action voltage equation is derived from the full-format dynamic linearization data model and a control input criterion function, and a basic model-free adaptive control system is established. Finally, based on the target of suppressing torque ripple and reducing overshoot of the motor output parameters, the parameters in the model-free adaptive control are iteratively optimized by combining a fast non-dominated genetic algorithm, so that the optimal solution meeting the requirements is determined. The method is simple to realize and has more ideal control effect, and simultaneously avoids problems such as large fluctuation of output torque caused by parameter mismatch and artificial setting of parameters, and significantly improves the stability of motor control.

Description

Technical Field

[0001] This invention belongs to the field of adaptive control technology for permanent magnet synchronous motors, specifically relating to a model-free adaptive control method for permanent magnet synchronous motors based on a multi-input multi-output nonlinear system with pre-optimization. Background Technology

[0002] The mathematical equations of permanent magnet synchronous motors involve periodic parameters that change with rotor rotation and differential operations, including multiple input and output quantities, and these parameters exhibit highly nonlinear and strong coupling relationships. Furthermore, the electromagnetic relationships within the motor are complex and variable, and motor parameters constantly change with the external operating environment, motor lifespan, and operating conditions, resulting in severely insufficient accuracy for motor control methods based on simplified mathematical models and numerous assumptions. In existing permanent magnet synchronous motor drive control technologies implemented through deadbeat predictive control and model predictive control, the accuracy requirements for motor parameters are generally high. Inappropriate matching between actual parameters and the control algorithm can increase current harmonics, improve torque ripple, and affect motor output; therefore, these two control methods require additional parameter identification modules. However, motor control accuracy is related to identification accuracy, and this approach still cannot escape the influence of unmodeled dynamics. Summary of the Invention

[0003] In view of this, and to address the technical problems existing in this field, the present invention provides a model-free adaptive control method for pre-optimization of permanent magnet synchronous motors, which is based on a multi-input multi-output nonlinear system. The method specifically includes the following steps:

[0004] Step 1: Establish the voltage equation of the permanent magnet synchronous motor in the rotating coordinate system. Based on this, perform a first-order Taylor expansion on the stator current and rotor electric angular velocity, and use the unknown nonlinear functions f1(…), f2(…), f3(…) to establish a simplified expression for the relationship between the input parameter u(k) and the output parameter y(k) of the motor's discrete nonlinear system:

[0005] y(k+1)=f(y(k),...,y(kn y ),u(k),...,u(kn u ))

[0006] In the formula, f(...) is an unknown nonlinear vector function f(...)=[f1(...),f2(...)f3(...)] T ;n y n u is a constant greater than or equal to zero; k is a discrete-time variable;

[0007] Based on the combined vector H(k)=[y(k) T ,...,y(kL y+1) T ,u(k) T ,...,u(kL u +1) T ] T and the corresponding pseudo-block Jacobian matrix Φ f (k) transforms the discrete-time nonlinear system into the following full-format dynamic linearized data model:

[0008] Δy(k+1)=Φ f (k)ΔH(k)

[0009] In the formula, the symbol Δ represents the change of the corresponding parameter at the current time compared to the adjacent previous time; 0≤L y ≤n y 0≤L u ≤n u ;

[0010] Step 2: Combine the desired output parameter y of the system ref And a weighting factor λ is introduced to limit the variation of the input parameters to set the following control input criterion function:

[0011] J(u(k))=||y ref (k+1)-y(k+1)|| 2 +λ||u(k)-u(k-1)|| 2

[0012] Substituting the full-format dynamic linearized data model into the control input criterion function yields a model-free system control algorithm that does not involve inversion operations;

[0013] Set the following estimation criterion function;

[0014] J(Φ f (k))=||Δy(k)-Φ f (k)ΔH(k-1)|| 2 +μ||Φ f (k)-Φ f (k-1)|| 2

[0015] By minimizing the above equation, an algorithm for updating the pseudo-block Jacobian matrix without inversion operations is derived. Based on the algorithm's tracking capability and the actual need to avoid misalignment of the pseudo-block Jacobian matrix parameters, a corresponding reset mechanism is set for the pseudo-block Jacobian matrix algorithm.

[0016] Step 3: With the goal of minimizing torque ripple, speed, and torque overshoot, set the objective function of the fast non-dominated sorting genetic algorithm and set the corresponding stator current constraint function.

[0017] Initial values ​​are determined for each of the iterative parameter sets, including the pseudo-block Jacobian matrix, the control input criterion function weighting factor, the estimation criterion function weighting factor, the model-free system control algorithm step size, and the algorithm step size for updating the pseudo-block Jacobian matrix. Stator current and rotor electric angular velocity are collected in real time. After updating and resetting the pseudo-block Jacobian matrix at the current moment, the voltage at the current moment is calculated by the model-free system control algorithm, combined with the expected output parameters of the motor at the next moment. This voltage value, after a one-beat delay, is used as an input parameter to the motor. The pseudo-block Jacobian matrix at the current moment is synchronously used as the initial value of the iterative parameters for the next moment to estimate the voltage. Simultaneously, the fast non-dominated sorting genetic algorithm is used to iteratively calculate each of the iterative parameter sets until a set number of iterations is reached. The optimal iterative parameter set that minimizes all three objectives is selected as the preliminary optimization result for use in the aforementioned control process at subsequent moments.

[0018] Furthermore, the voltage equation of the permanent magnet synchronous motor established in the rotating coordinate system in step one specifically adopts the following form:

[0019]

[0020]

[0021]

[0022]

[0023] In the formula, u d u q and i d i q These represent the d-axis and q-axis stator voltages and stator currents, respectively; R s For stator resistance; L s Stator inductance; ψ f For rotor permanent magnet flux linkage; ω e and ω m These are the rotor electrical angular velocity and the rotor mechanical angular velocity, respectively, with ω e =pω m p is the extreme logarithm; T e T is the electromagnetic torque. L J is the load torque; J is the moment of inertia; t is time.

[0024] After performing the first-order Taylor expansion on the stator current and rotor electric angular velocity as described below, the following equations are obtained to predict the stator current and rotor electric angular velocity at time k+1 using the motor parameters at time k:

[0025]

[0026]

[0027]

[0028] Among them, i d (k+1),i q (k+1) represents the d-axis and q-axis stator currents at time k+1; i d (k), i q (k) represents the d-axis and q-axis stator currents at time k; u d (k), u q (k) represents the d-axis and q-axis stator voltages at time k; ω e (k+1) represents the rotor's electric angular velocity at time k+1; ω e (k) represents the rotor's electric angular velocity at time k.

[0029] For a discrete nonlinear system of a motor, the following relationship between input and output parameters can be established using the unknown nonlinear function:

[0030]

[0031] After describing the above relationship as the simplified expression, the input parameter u(k) and the output parameter y(k) are respectively expressed as u(k) = [u d (k),u q (k),0] T ∈R 3 ,y(k)=[i d (k),i q (k),ω e (k)] T ∈R 3 .

[0032] Furthermore, the establishment of the full-format dynamic linearized data model in step one is based on the following two assumptions:

[0033] (1) f1(…), f2(…), f3(…) have continuous partial derivatives with respect to each variable;

[0034] (2) The discrete nonlinear system of permanent magnet synchronous motor satisfies the generalized Lipschitz condition;

[0035] Let the combined vector H(k) have the following relationship for any k1≠k2, k1, k2≥0, and H(k1)≠H(k2):

[0036] ||y(k1+1)-y(k2+1)||≤b||H(k1)-H(k2)||

[0037] For any non-zero ΔH(k) = H(k) - H(k-1), there must exist a corresponding pseudo-block Jacobian matrix. And satisfy b > 0 is a constant; when k = 0, i = 1, L y +1, i = 2, ..., L y L y +2、...、L y +L u , These are the iteration parameters to be determined.

[0038] Furthermore, the weighting factor corresponding to the control input criterion function mentioned in step two is set to λ>0; the expected output parameter y at time k+1 ref The specific form is as follows:

[0039] Substituting the full-format dynamic linearized data model into the control input criterion function, and taking the derivative with respect to u(k) and setting the derivative to 0, yields the following model-free system control algorithm without inversion operations:

[0040]

[0041] In the formula, ρ is the step size factor constant.

[0042] Furthermore, the weight factor corresponding to the estimation criterion function mentioned in step two is set to μ>0, and its minimization derivation yields the following form of the algorithm for updating the pseudo-block Jacobian matrix:

[0043]

[0044] Where η is the step size factor constant, and 2≥η>0;

[0045] The reset mechanism is specifically configured as follows:

[0046] if or or So i = 1, 2, 3;

[0047] if or So i,j=1,2,3,i≠j.

[0048] Furthermore, the objective function of the fast non-dominated sorting genetic algorithm described in step three is specifically set in the following form:

[0049]

[0050] obj2:Δh n =|i q (max)-i q (∞)|

[0051] obj3:Δh T =|i q (max)-i q (∞)|

[0052] in, This is a torque ripple evaluation method, i. q The root mean square error; Δh n For the current to rise to its maximum value and remain there, so that the rotational speed changes linearly to a given value, i q The maximum deviation; similarly, Δh T Before the desired torque is reached, i q Maximum deviation of overshoot; i qn (∞) represents the theoretical value of the q-axis current during the linear increase of the rotational speed to the given value, i.e., the actual q-axis current should be at i qn (∞) fluctuates up and down; i qn (max) represents the deviation of the current from i when it fluctuates up and down before the rotational speed reaches the desired value. qn (∞) the farthest value; similarly, i qT (∞) represents the theoretical value of the q-axis current when the motor outputs its theoretical torque; i qT (max) represents the deviation of the current from i when it fluctuates up and down before the torque reaches the desired value. qT (∞) The farthest value.

[0053] The stator current constraint function is used to make i q It is negative only under reasonable fluctuations, and i q The pulsations at steady state are also within a limited range, and their specific form is as follows:

[0054] sti q (k)>-c1

[0055]

[0056] Where c1 and c2 are constants greater than 0.

[0057] Furthermore, the specific process of the pre-optimization described in step three is as follows:

[0058] (1) Randomly assign values ​​to the iterative parameter set within a given numerical range: Determine η, μ, λ, and ρ;

[0059] (2) Define the input parameters, output parameters, and their changes at adjacent time steps; design the changes at adjacent time steps of the combined vector; and define the initial values ​​of each iteration parameter and the pseudo-block Jacobian matrix:

[0060] u(k)=[u d (k),u q (k), 0] T Δu(k-1)=u(k-1)-u(k-2)

[0061] y(k)=[i d (k),i q (k),ω e (k)] T ,Δy(k)=y(k)-y(k-1),Δy(k-1)=y(k-1)-y(k-2)

[0062] ΔH(k-1)=[Δy(k-1) T ,Δu(k-1) T ] T

[0063] L y =L u =3

[0064] Φ f (0)=[Φ1(0),Φ2(0),...,Φ6(0)]

[0065]

[0066] (3) Calculate the pseudo-block Jacobian matrix corresponding to the current k-time;

[0067] (4) According to the reset mechanism, Φ i (k) Reset values ​​that are out of range;

[0068] (5) Calculate the voltage applied to the motor based on the expected output parameters of the motor at time k+1;

[0069] (6) Repeat steps (3) to (5) above until the permanent magnet synchronous motor completes the drive within the predetermined working cycle;

[0070] (7) For completed work cycles, simultaneously calculate the objective function values ​​of the fast non-dominated sorting genetic algorithm;

[0071] (8) Repeat the above steps (1) to (7) and select the iterative parameter set that minimizes the values ​​of the three objective functions as the result of the preliminary optimization;

[0072] (9) Substitute the pre-optimization result into the modelless system control algorithm, that is, apply it to steps (2) to (5) to avoid the interference of manually set parameters on the modelless system and eliminate the influence of motor parameters on control robustness.

[0073] The model-free adaptive control method for permanent magnet synchronous motors (PMSMs) provided by this invention enables drive control of PMSMs without involving motor parameters. This method first establishes a discrete-time nonlinear system for the PMSM and determines the input and output parameters. Then, it processes these parameters using a full-format dynamic linearized data model. Subsequently, the voltage equation for the next time step is derived from the full-format dynamic linearized data model and the control input criterion function, establishing a basic model-free adaptive control system. Finally, based on the objectives of suppressing torque ripple and reducing motor output parameter overshoot, a fast non-dominated genetic algorithm is used to iteratively optimize the parameters in the model-free adaptive control, thereby determining the optimal solution that meets the requirements. This method is simple to implement and achieves more ideal control results. While avoiding problems such as motor parameter mismatch and large fluctuations in output torque caused by manually set parameters, it also significantly improves the stability of motor control. Attached Figure Description

[0074] Figure 1 A flowchart of the method provided by the present invention;

[0075] Figure 2 This is a block diagram of the model-free adaptive control of a permanent magnet synchronous motor.

[0076] Figure 3 The output current and target current of the permanent magnet synchronous motor driven by this invention are given. Detailed Implementation

[0077] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] The model-free adaptive control method for pre-optimization of permanent magnet synchronous motors provided in this invention is based on a multi-input multi-output nonlinear system, such as... Figure 1 As shown, the specific steps include:

[0079] Step 1: Establish the voltage equation of the permanent magnet synchronous motor in the rotating coordinate system. Based on this, perform a first-order Taylor expansion on the stator current and rotor electric angular velocity, and use the unknown nonlinear functions f1(…), f2(…), f3(…) to establish a simplified expression for the relationship between the input parameter u(k) and the output parameter y(k) of the motor's discrete nonlinear system:

[0080] y(k+1)=f(y(k),...,y(kn y ),u(k),...,u(kn u ))

[0081] In the formula, f(...) is an unknown nonlinear vector function f(...)=[f1(...),f2(...)f3(...)] T ;n y n u is a constant greater than or equal to zero; k is a discrete-time variable;

[0082] Based on the combined vector H(k)=[y(k) T ,...,y(kL y +1) T ,u(k) T ,...,u(kL u +1) T ] T and the corresponding pseudo-block Jacobian matrix Φ f (k) transforms the discrete-time nonlinear system into the following full-format dynamic linearized data model:

[0083] Δy(k+1)=Φ f (k)ΔH(k)

[0084] In the formula, the symbol Δ represents the change of the corresponding parameter at the current time compared to the adjacent previous time; 0≤L y ≤n y 0≤L u ≤n u ;

[0085] Step 2: Combine the desired output parameter y of the system ref And a weighting factor λ is introduced to limit the variation of the input parameters to set the following control input criterion function:

[0086] J(u(k))=||y ref (k+1)-y(k+1)|| 2 +λ||u(k)-u(k-1)|| 2

[0087] Substituting the full-format dynamic linearized data model into the control input criterion function yields a model-free system control algorithm that does not involve inversion operations;

[0088] Set the following estimation criterion function;

[0089] J(Φ f (k))=||Δy(k)-Φ f (k)ΔH(k-1)|| 2 +μ||Φ f (k)-Φ f (k-1)|| 2

[0090] By minimizing the above equation, an algorithm for updating the pseudo-block Jacobian matrix without inversion operations is derived. Based on the algorithm's tracking capability and the actual need to avoid misalignment of the pseudo-block Jacobian matrix parameters, a corresponding reset mechanism is set for the pseudo-block Jacobian matrix algorithm.

[0091] Step 3: With the goal of minimizing torque ripple, speed, and torque overshoot, set the objective function of the fast non-dominated sorting genetic algorithm and set the corresponding stator current constraint function.

[0092] For the iterative parameter set including the pseudo-block Jacobian matrix, the control input criterion function weighting factor, the estimation criterion function weighting factor, the model-free system control algorithm step size, and the algorithm step size for updating the pseudo-block Jacobian matrix, initial values ​​are determined respectively. Stator current and rotor electric angular velocity are collected in real time. After updating and resetting the pseudo-block Jacobian matrix at the current moment, and combining it with the expected output parameters of the motor at the next moment, the voltage at the current moment is calculated by the model-free system control algorithm. This voltage value, after a one-beat delay, is used as an input parameter to the motor. The pseudo-block Jacobian matrix at the current moment is synchronously used as the initial value of the iterative parameters for the next moment to estimate the voltage. Simultaneously, the fast non-dominated sorting genetic algorithm is used to iteratively calculate each of the iterative parameter sets until a set number of iterations is reached. The optimal iterative parameter set that minimizes all three objectives is selected as the pre-optimization result and used in the aforementioned control process at subsequent moments. The control principle framework corresponding to this method is as follows: Figure 2 As shown.

[0093] In a preferred embodiment of the present invention, the voltage equation of the permanent magnet synchronous motor established in step one in the rotating coordinate system specifically takes the following form:

[0094]

[0095]

[0096]

[0097]

[0098] In the formula, u d u q and i d i q These represent the d-axis and q-axis stator voltages and stator currents, respectively; R s For stator resistance; L s Stator inductance; ψ f For rotor permanent magnet flux linkage; ω e and ω m These are the rotor electrical angular velocity and the rotor mechanical angular velocity, respectively, with ω e =pω m p is the extreme logarithm; T e T is the electromagnetic torque. L J is the load torque; J is the moment of inertia; t is time.

[0099] After performing the first-order Taylor expansion on the stator current and rotor electric angular velocity as described below, the following equations are obtained to predict the stator current and rotor electric angular velocity at time k+1 using the motor parameters at time k:

[0100]

[0101]

[0102]

[0103] Among them, i d (k+1),i q (k+1) represents the d-axis and q-axis stator currents at time k+1; i d (k), i q (k) represents the d-axis and q-axis stator currents at time k; u d (k), u q (k) represents the d-axis and q-axis stator voltages at time k; ω e (k+1) represents the rotor's electric angular velocity at time k+1; ω e (k) represents the rotor's electric angular velocity at time k.

[0104] For a discrete nonlinear system of a motor, the following relationship between input and output parameters can be established using the unknown nonlinear function:

[0105]

[0106] After describing the above relationship as the simplified expression, the input parameter u(k) and the output parameter y(k) are respectively expressed as u(k) = [u d (k),u q(k),0] T ∈R 3 ,y(k)=[i d (k),i q (k),ω e (k)] T ∈R 3 .

[0107] In a preferred embodiment of the present invention, the establishment of the full-format dynamic linearized data model in step one is based on the following two assumptions:

[0108] (1) f1(…), f2(…), f3(…) have continuous partial derivatives with respect to each variable;

[0109] (2) The discrete nonlinear system of permanent magnet synchronous motor satisfies the generalized Lipschitz condition;

[0110] Let the combined vector H(k) have the following relationship for any k1≠k2, k1, k2≥0, and H(k1)≠H(k2):

[0111] ||y(k1+1)-y(k2+1)||≤b||H(k1)-H(k2)||

[0112] For any non-zero ΔH(k) = H(k) - H(k-1), there must exist a corresponding pseudo-block Jacobian matrix. And satisfy b > 0 is a constant; when k = 0, , i = 2, ..., L y L y +2、...、L y +L u , These are the iteration parameters to be determined.

[0113] In a preferred embodiment of the present invention, the weighting factor corresponding to the control input criterion function in step two is set to λ>0; the expected output parameter y at time k+1 ref The specific form is as follows:

[0114] Substituting the full-format dynamic linearized data model into the control input criterion function, and taking the derivative with respect to u(k) and setting the derivative to 0, yields the following model-free system control algorithm without inversion operations:

[0115]

[0116] In the formula, ρ is the step size factor constant.

[0117] In a preferred embodiment of the present invention, the weight factor corresponding to the estimation criterion function in step two is set to μ>0, and the following algorithm for updating the pseudo-block Jacobian matrix is ​​derived by minimizing it:

[0118]

[0119] Where η is the step size factor constant, and 2≥η>0;

[0120] The reset mechanism is specifically configured as follows:

[0121] if or or So i = 1, 2, 3;

[0122] if or So i,j=1,2,3,i≠j.

[0123] In a preferred embodiment of the present invention, the objective function of the fast non-dominated sorting genetic algorithm in step three is specifically set in the following form:

[0124]

[0125]

[0126] obj3:Δh T =|i q (max)-i q (∞)

[0127] in, This is a torque ripple evaluation method, i. q The root mean square error; Δh n For the current to rise to its maximum value and remain there, so that the rotational speed changes linearly to a given value, i q The maximum deviation; similarly, Δh T Before the desired torque is reached, i q Maximum deviation of overshoot; i qn (∞) represents the theoretical value of the q-axis current during the linear increase of the rotational speed to the given value, i.e., the actual q-axis current should be at i qn (∞) fluctuates up and down; i qn (max) represents the deviation of the current from i when it fluctuates up and down before the rotational speed reaches the desired value. qn (∞) the farthest value; similarly, i qT (∞) represents the theoretical value of the q-axis current when the motor outputs its theoretical torque; i qT(max) represents the deviation of the current from i when it fluctuates up and down before the torque reaches the desired value. qT (∞) The farthest value.

[0128] The stator current constraint function is used to make i q It is negative only under reasonable fluctuations, and i q The pulsations at steady state are also within a limited range, and their specific form is as follows:

[0129] sti q (k)>-c1

[0130]

[0131] Where c1 and c2 are constants greater than 0.

[0132] In a preferred embodiment of the present invention, the specific process of the pre-optimization in step three is as follows:

[0133] (1) Randomly assign values ​​to the iterative parameter set within a given numerical range: Determine η, μ, λ, and ρ;

[0134] (2) Define the input parameters, output parameters, and their changes at adjacent time steps; design the changes at adjacent time steps of the combined vector; and define the initial values ​​of each iteration parameter and the pseudo-block Jacobian matrix:

[0135] u(k)=[u d (k),u q (k), 0] T Δu(k-1)=u(k-1)-u(k-2)

[0136] y(k)=[i d (k),i q (k),ω e (k)] T ,Δy(k)=y(k)-y(k-1),Δy(k-1)=y(k-1)-y(k-2)

[0137] ΔH(k-1)=[Δy(k-1) T ,Δu(k-1) T ] T

[0138] L y =L u =3

[0139] Φ f (0)=[Φ1(0),Φ2(0),...,Φ6(0)]

[0140]

[0141] (3) Calculate the pseudo-block Jacobian matrix corresponding to the current k-time;

[0142] (4) According to the reset mechanism, Φ i (k) Reset values ​​that are out of range;

[0143] (5) Calculate the voltage applied to the motor based on the expected output parameters of the motor at time k+1;

[0144] (6) Repeat steps (3) to (5) above until the permanent magnet synchronous motor completes the drive within the predetermined working cycle;

[0145] (7) For completed work cycles, simultaneously calculate the objective function values ​​of the fast non-dominated sorting genetic algorithm;

[0146] (8) Repeat the above steps (1) to (7) and select the iterative parameter set that minimizes the values ​​of the three objective functions as the result of the preliminary optimization;

[0147] (9) Substitute the pre-optimization result into the modelless system control algorithm, that is, apply it to steps (2) to (5) to avoid the interference of manually set parameters on the modelless system and eliminate the influence of motor parameters on control robustness.

[0148] In a specific embodiment of the present invention, the motor parameters are as follows: rated voltage U is 120V, rated current I is 20A, and stator resistance R... s The Ω is 0.365Ω, and the stator inductance L is... s The flux linkage ψ of the rotor permanent magnet is 0.001225H. f The value is 0.1667 Wb. The motor model is given a torque step of (0 N to 5 N) at 0.02 s and a speed step of (400 r / min to 600 r / min) at 0.06 s.

[0149] In the pre-optimization model-free adaptive control method, the initial value of the pseudo-block Jacobian matrix is: Φ f (0)=[Φ1(0),...,Φ6(0)], i = 1, 4, i = 2, 3, 5, 6, where When updating the pseudo-block Jacobian matrix, the weight factor μ is 5.24 and the step size factor η is 0.19. In the reset mechanism, a, b1, and b2 are 1.0001, 2 × 10⁻⁶, and 2 × 10⁻⁶, respectively. -7 9.999. When calculating the applied voltage at the next moment, the weighting factor λ is 42.70 and the step factor ρ is 33.74. The output value and expected value of the motor stator current are as follows: Figure 3 As shown in the figure, it can be seen that the output value has a good tracking effect on the expected value.

[0150] It should be understood that the sequence number of each step in the embodiments of the present invention does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0151] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A model-free adaptive control method for pre-optimization of permanent magnet synchronous motors, characterized in that: Specifically, the following steps are included: Step 1: Establish the voltage equations of the permanent magnet synchronous motor in the rotating coordinate system. Based on this, perform a first-order Taylor expansion on the stator current and rotor electric angular velocity, and utilize the unknown nonlinear function. f 1(...), f 2(...), f 3(...) to establish the input parameters of the discrete nonlinear system of the motor u ( k ) and output parameters y ( k A simplified expression for the relationship between ) In the formula, f (...) is an unknown nonlinear vector function f (…) = [ f 1(...), f 2(...) f 3(…)] T ; n y , n u A constant that is greater than or equal to zero; k For discrete-time variables; Based on combined vectors and the corresponding pseudo-block Jacobian matrix Φ f ( k The discrete-time nonlinear system is transformed into the following full-format dynamic linearized data model: In the formula, the symbol Δ represents the change of the corresponding parameter at the current time compared to the adjacent previous time; 0 ≤ L y ≤ n y 0 ≤ L u ≤ n u ; Step 2: Combine the desired output parameters of the system y ref And introduce weighting factors to limit the variation of input parameters. λ To set the following control input criterion functions: Substituting the full-format dynamic linearized data model into the control input criterion function yields a model-free system control algorithm that does not involve inversion operations; Set the following estimation criterion function; By minimizing the above equation, an algorithm for updating the pseudo-block Jacobian matrix without inversion operations is derived. Based on the algorithm's tracking capability and the actual need to avoid misalignment of the pseudo-block Jacobian matrix parameters, a corresponding reset mechanism is set for the pseudo-block Jacobian matrix algorithm. Step 3: With the goal of minimizing torque ripple, speed, and torque overshoot, set the objective function of the fast non-dominated sorting genetic algorithm and set the corresponding stator current constraint function. For the iterative parameter set including the pseudo-block Jacobian matrix, the control input criterion function weight factor, the estimation criterion function weight factor, the step size of the model-free system control algorithm, and the step size of the algorithm for updating the pseudo-block Jacobian matrix, determine the corresponding initial values ​​respectively. The stator current and rotor electric angular velocity are collected in real time. After updating the pseudo-block Jacobian matrix at the current moment and resetting it as necessary, the voltage at the current moment is calculated by the model-free system control algorithm in combination with the expected output parameters of the motor at the next moment. The voltage value is then used as an input parameter to the motor after a one-beat delay. The pseudo-block Jacobian matrix at the current moment is synchronously used as the initial value of the iteration parameters for the next moment to estimate the voltage at the next moment. At the same time, the fast non-dominated sorting genetic algorithm is used to iteratively calculate each of the iteration parameter sets until the set number of iterations is reached. The optimal iteration parameter set that minimizes all three objectives is selected as the preliminary optimization result and used in the aforementioned control process at subsequent moments.

2. The method as described in claim 1, characterized in that: The voltage equation for a permanent magnet synchronous motor in a rotating coordinate system, established in step one, takes the following form: In the formula, u d , u q and i d , i q They are respectively d , q Stator voltage and stator current; R s Stator resistance; L s For stator inductance; ψ f For rotor permanent magnet flux linkage; ω e and ω m These are the rotor electrical angular velocity and the rotor mechanical angular velocity, respectively. ω e = pω m , p It is the extreme logarithm; T e Electromagnetic torque; T L This is the load torque; J It is the moment of inertia; t For time; After performing the following first-order Taylor expansion on the stator current and rotor electric angular velocity, the following is obtained: k Motor parameter prediction at any time k+ The equations for the stator current and rotor electric angular velocity at time 1 are as follows: in, i d ( k +1) i q ( k +1) is k +1 moment d , q Shaft stator current; i d ( k ), i q ( k )for k time d , q Shaft stator current; u d ( k ), u q ( k )for k time d , q Shaft-stator voltage; ω e ( k +1) is k Rotor electric angular velocity at time +1; ω e ( k )for k The rotor's electrical angular velocity at any given moment; For a discrete nonlinear system of a motor, the following relationship between input and output parameters is established using the unknown nonlinear function: After describing the above relationship as the simplified expression, the input parameter u ( k ) and output parameters y ( k ) are respectively represented as .

3. The method as described in claim 2, characterized in that: The establishment of the full-format dynamic linearized data model in step one is based on the following two assumptions: (1) There are continuous partial derivatives with respect to each variable; (2) The discrete nonlinear system of permanent magnet synchronous motor satisfies the generalized Lipschitz condition; Let the combined vector H ( k For any ,and The following relationships exist: For nonzero Δ H ( k )= H ( k )- H ( k -1) There must exist a corresponding pseudo-block Jacobian matrix. And satisfy ; b > 0 is a constant; k When =0, , , φ 1. φ 2 represents the iteration parameters to be determined.

4. The method as described in claim 3, characterized in that: The weighting factor corresponding to the control input criterion function mentioned in step two is set as follows: λ >0; k The expected output parameter at time +1 y ref The specific form is as follows: ; Substituting the full-format dynamic linearized data model into the control input criterion function, for Taking the derivative and setting it to zero, we obtain the following model-free system control algorithm without inversion operations: In the formula, ρ This is the step size factor constant.

5. The method as described in claim 4, characterized in that: The weighting factor corresponding to the estimation criterion function mentioned in step two is set as follows: μ The algorithm for updating the pseudo-block Jacobian matrix is ​​derived by minimizing the value >0, resulting in the following form: in, η Let be the step size factor constant, then 2 ≥ η >0; The reset mechanism is specifically configured as follows: if ,or ,or ,So , ; if ,or ,So , .

6. The method as described in claim 5, characterized in that: The objective function of the fast non-dominated sorting genetic algorithm described in step three is specifically set in the following form: in, This is a torque ripple assessment method, namely i q The root mean square error; Δ h n For the process of the current rising to its maximum value and remaining there, so that the rotational speed changes linearly to a given value... i q The maximum deviation; similarly, Δ h T Before the desired torque is achieved, i q Maximum deviation of overshoot; i qn (∞) represents the process of the rotational speed linearly increasing to the given value. q The theoretical value of shaft current, i.e., the actual value... q The shaft current should be in i qn (∞) fluctuates up and down; i qn (max) represents the deviation of the current from the desired value when it fluctuates up and down before the rotational speed reaches the desired value. i qn (∞) The farthest value; similarly, i qT (∞) represents the theoretical output torque of the motor. q Theoretical values ​​for shaft current; i qT (max) represents the deviation of the current from the desired torque value during fluctuations. i qT (∞) The farthest value; The stator current constraint function is used to make... i q It is negative only under reasonable fluctuations, and i q The pulsations at steady state are also within a limited range, and their specific form is as follows: in, c 1. c 2 is a constant greater than 0.

7. The method as described in claim 6, characterized in that: The specific process of the pre-optimization described in step three is as follows: (1) Randomly assign values ​​to the iterative parameter set within a given numerical range: Determine φ 1. φ 2. η , μ , λ and ρ ; (2) Define the input parameters, output parameters, and their changes at adjacent time points; design the changes at adjacent time points of the combined vector; and define the initial values ​​of each iteration parameter and the pseudo-block Jacobian matrix: (3) Calculate the current k The pseudo-block Jacobian matrix corresponding to time step; (4) According to the reset mechanism Reset values ​​that are out of range; (5) According to k+ Calculate the voltage value applied to the motor based on the expected output parameters of the motor at time 1; (6) Repeat steps (3) to (5) above until the permanent magnet synchronous motor completes the drive within the predetermined working cycle; (7) For completed work cycles, simultaneously calculate the objective function values ​​of the fast non-dominated sorting genetic algorithm; (8) Repeat steps (1) to (7) above, and select the iterative parameter set that minimizes the values ​​of the three objective functions as the result of the preliminary optimization; (9) Substitute the pre-optimization result into the modelless system control algorithm, that is, apply it to steps (2) to (5) to avoid interference from manually set parameters to the modelless system, while eliminating the influence of motor parameters on control robustness.

Images