A three-vector model predictive torque control method for permanent magnet synchronous motor

Abstract

A three-vector model predictive torque control method for a permanent magnet synchronous motor belongs to the technical field of motor control, and includes the following steps: step S01, setting the running speed of the motor; step S02, obtaining the current motor speed and calculating the speed error; step S03, designing an improved anti-disturbance controller, inputting the speed error into the improved anti-disturbance controller, and thus obtaining the corresponding reference torque. Step S04: generating a reference flux through a reference torque; step S05: obtaining feedback current, the current motor running torque, and the motor angle; step S06: calculating the predicted values ​​of the torque and flux at the next moment; step S07: applying the obtained optimal voltage vector combination to the motor. The present invention improves the overall robustness and control accuracy of the control system by setting a new model predictive torque control strategy of an improved anti-disturbance controller.

Description

Technical Field

[0001] The present invention belongs to the technical field of motor control, and in particular relates to a three-vector model predictive torque control method for a permanent magnet synchronous motor. Background Art

[0002] In the grinding and selection process of mines, ball mills are important equipment for re-crushing ore. Their working state directly affects the efficiency of the grinding system. Most ball mills are driven by permanent magnet synchronous motors (PMSM). For low-speed and high-torque permanent magnet synchronous motors, simply optimizing the structure of the permanent magnet synchronous motor cannot achieve the best results, and improvements are also needed in the direction of motor control. Traditional motor control strategies still have certain defects and deficiencies in solving the problems caused by the multivariable, strong coupling, and nonlinear characteristics of permanent magnet synchronous motor systems. In particular, the reference torque generated by the PI controller in the traditional model predictive torque control (MPTC) is sensitive to load torque changes and internal parameter disturbances, resulting in poor robustness, and the weight factor selection in the cost function is difficult. If the weight factor is not selected properly, it will cause the motor torque and speed pulsation, reducing the control accuracy. Therefore, under the above conditions, how to optimize the motor control strategy to improve the motor control accuracy and the robustness of the motor system has become a major difficulty. Summary of the invention

[0003] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a three-vector model predictive torque control method for a permanent magnet synchronous motor, and proposes a new model predictive torque control strategy with an improved anti-disturbance controller to improve the overall robustness and control accuracy of the control system.

[0004] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0005] A three-vector model predictive torque control method for a permanent magnet synchronous motor comprises the following steps:

[0006] Step S01, setting the running speed of the motor, and grinding different ores at different speeds;

[0007] Step S02: Obtain the current motor speed through the speed sensor, and subtract it from the set speed to obtain the speed error;

[0008] Step S03, designing an improved active disturbance rejection controller, inputting the speed error into the improved active disturbance rejection controller, thereby obtaining a corresponding reference torque;

[0009] The improved active disturbance rejection controller is:

[0010]

[0011] Where: e 1D (k) is the speed error at time k; ω ref is the speed given signal; x 1D (k) Tracking input signal ω * (k) value, x 1D (k+1) is the value of x at time k+1 1D (k) predicted value; x 2D (k) is x 1D The differential of (k), x 2D (k+1) is the value of x at time k+1 2D (k) predicted value; T s is the discrete control period; r is the speed factor, which determines the tracking speed of nonlinear TD; h is the filter factor, which determines the filtering effect of nonlinear TD; fst() is the optimal comprehensive function of discrete fast control; nfal() is the improved error nonlinear function; α determines the linearity of nfal function, and its value is 0<α<1; δ is the filter factor, which is generally 5T s <δ<10T s ; T e * is the torque reference value; z 1 (k) is the estimated value of the rotation speed; 1 (k+1) is the k+1 moment z 1 (k) predicted value; z 2 (k) is z 1 The differential of (k) is the total disturbance of the system; z 2 (k+1) is the k+1 moment z 2 The predicted value of (k), β 1 ,β 2 is the control parameter; e is the error, u 0 (k) is the control function, T e * (k) is the output reference torque;

[0012] Step S04: By using the reference torque T e * (k), and the reference flux ψ is generated by formula (15) s * (k)

[0013]

[0014] Where: s * (k) is the motor flux amplitude at time k+1; ψ f is the stator flux; L qis the inductance q; P n is the number of motor pole pairs;

[0015] Step S05: The motor is running, feedback current is obtained, and the current i based on the dq axis is obtained through transformation. d ,i q , obtain the current motor running torque and motor angle;

[0016] Step S06: According to the feedback parameters obtained in step S05, the predicted values ​​of the torque and flux at the next moment are calculated by using the prediction equations of the torque and flux;

[0017] The prediction equations for the torque and flux are as follows:

[0018]

[0019] Where: i d 、i q , L d , L q are d-axis current and q-axis current and inductance respectively; i d (k+1), i q (k+1) is the d and q axis current at time k+1; R s is the stator resistance; ω e is the rotor electrical angular velocity; ψ f is the stator flux; T e is the electromagnetic torque; P n is the number of motor pole pairs; ψ s (k+1) is the motor flux amplitude at time k+1; ψ d (k+1),ψ q (k+1) is the d- and q-axis magnetic flux at time k+1;

[0020] At the same time, the cost function of the improved fixed weight factor is used to obtain the value of the cost function corresponding to each predicted torque, flux and reference torque and flux. The cost function of the improved fixed weight factor is as follows:

[0021] g 1 =|T e * -T e i (k+1)|+A|ψ s * -ψ s i (k+1)| (24)

[0022] g 2 =(T e * -T e i(k+1)) 2 +C 1 (25)

[0023] g 3 =(ψ q * -ψ q i (k+1)) 2 +C 2 (28)

[0024] Where: A = T N / ψ sN is the weight factor, where T N represents the rated torque, ψ sN is the stator flux amplitude under rated condition; T e * is the reference torque; T e i (k+1) is the torque value under the action of the i-th voltage vector at time k+1; ψ s * is the reference flux value; ψ q * is the reference flux value of q axis; ψ q i (k+1) is the q-axis flux value under the action of the i-th voltage vector at time k+1; ψ s i (k+1) is the flux value under the action of the i-th voltage vector at time k+1; C 1 , C 2 are flux linkage and torque stability factors respectively;

[0025] Substitute the voltage vector into formula (21) to obtain the prediction result, substitute the prediction result into the improved cost function of the fixed weight factor, obtain the value of the predicted torque that minimizes the cost function, and obtain the voltage vector combination that produces the predicted torque;

[0026] Step S07: Apply the optimal voltage vector combination obtained in step S06 to the motor.

[0027] Furthermore, the step S03 further includes the following steps:

[0028] Assume that the set speed is ω ref , the sensor obtains the rotation speed as ω * (k) Design a discrete speed loop nonlinear tracking differentiator as follows:

[0029]

[0030] In the formula, e 1D (k) is the speed error at time k; ωref is the speed given signal; x 1D (k) Tracking input signal ω * (k), x 1D (k+1) is the value of x at time k+1 1D (k) predicted value; x 2D (k) is x 1D The differential of (k), x 2D (k+1) is the value of x at time k+1 2D (k) predicted value; T s is the discrete control period; r is the speed factor, which determines the tracking speed of the nonlinear TD; h is the filtering factor, which determines the filtering effect of the nonlinear TD; fst() is the optimal comprehensive function of discrete fast control, which is expressed as:

[0031]

[0032]

[0033] Where sgn() is the sign function;

[0034] According to the speed outer loop of the control system, equation (1) is rewritten as:

[0035]

[0036] In the formula, is the total torque disturbance, T L is the motor load, B is the reluctance torque, J is the moment of inertia;

[0037] The state equation is obtained as follows:

[0038]

[0039] The discrete ESO is obtained as follows:

[0040]

[0041] Where fal() is the error nonlinear function; α determines the linearity of the fal function, and its value is 0<α<1; δ is the filtering factor, which is generally 5T s <δ<10T s ; T e * is the torque reference value; ω(k) is the actual motor speed, z 1 (k) is the estimated value of the rotation speed; 1 (k+1) is the k+1 moment z 1 (k) predicted value; z 2 (k) is z 1The differential of (k) is the total disturbance of the system; z 2 (k+1) is the k+1 moment z 2 The predicted value of (k), β 1 ,β 2 is the control parameter; according to the traditional function expression, the fal function is improved, and when |e|≤δ, fal is rewritten as:

[0042]

[0043] To ensure continuity and differentiability at |e| = δ, the value of the fal function and its derivative must be the same, that is,

[0044]

[0045] Solving the above equations together gives:

[0046]

[0047] The nfal function is written as

[0048]

[0049] Design of discrete nonlinear state error feedback

[0050]

[0051] Where T e * (k) is the output reference torque, e is the error, u 0 (k) is the control function.

[0052] Furthermore, the step S06 further includes the following steps:

[0053] The mathematical model of the built-in PMSM in the d and q axis coordinate system is expressed as:

[0054]

[0055] In the formula, i d 、i q , L d , L q are d-axis current and q-axis inductance respectively; u d 、u q is the d and q axis voltage; R s is the stator resistance; ω e is the rotor electrical angular velocity; ψ f is the stator flux;

[0056] The torque equation of PMSM is:

[0057]

[0058] Where, T e is the electromagnetic torque; P n is the number of motor pole pairs;

[0059] The stator flux equation is:

[0060]

[0061] In the formula, ψ s is the motor flux amplitude; ψ d , ψ q They are d-axis and q-axis magnetic flux respectively;

[0062] According to the motor speed expression, the forward Euler method is used to discretize equation (16), and the stator current prediction equation at time k+1 can be obtained:

[0063]

[0064] In the formula, i d (k), i q (k),u d (k) and u q (k) represents the d-axis and q-axis current and voltage values ​​at time k respectively; i d (k+1), i q (k+1) represents the predicted current value at time k+1, and the prediction equations for torque and flux are:

[0065]

[0066] In the formula, ψ d (k+1),ψ q (k+1) represents the d-axis and q-axis flux linkage values ​​at time k+1 respectively;

[0067] Introducing torque and flux determination factors, the torque is:

[0068] T i =T e * -T e i (k+1) (22)

[0069] The magnetic linkage determination factor is set to:

[0070] ψ i =ψ q * -ψ q i (k+1) (23)

[0071] Among them, the value range of i is 1 to 8 of the 8 voltage vectors; T i and ψ i are the discriminant factors of torque and flux respectively; T e * , ψ q * are the reference torque and q-axis reference flux respectively; T e i (k+1) and ψ q i (k+1) is the torque and q-axis flux parameter value under the action of the i-th voltage vector at time k+1.

[0072] The beneficial effects of the present invention are as follows: the present invention proposes a novel model predictive torque control strategy with an improved auto-disturbance rejection controller, which improves the overall robustness and control accuracy of the control system. By replacing the speed PI controller in the traditional model predictive torque control with an auto-disturbance rejection controller, and replacing the nonlinear function in the auto-disturbance rejection controller with a new function in order to improve the smoothness of the controller, the robustness of the traditional speed PI controller and the poor smoothness of the traditional auto-disturbance rejection controller are solved.

[0073] The present invention reconstructs the cost function and process in the traditional model prediction torque control strategy. First, the first optimal voltage vector is selected by adopting a fixed coefficient cost function. Secondly, the second voltage vector that makes the torque reach the optimal value is selected by a cost function that only contains torque factors, and the magnetic flux is stabilized by a stabilizing factor. Finally, the vector after the first two voltage vectors act is regarded as a new vector, and the third voltage vector that makes the magnetic flux reach the optimal value is selected by a cost function that only contains q-axis magnetic flux factors, and the torque is stabilized by a stabilizing factor. This solves the problem of difficult adjustment of weight factors in the cost function and low control accuracy caused by improper selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a schematic diagram of the overall flow of the control strategy of the present invention;

[0075] Figure 2 It is a steady-state simulation diagram of the ITV-MPTC motor in a specific implementation manner;

[0076] Figure 3 It is a dynamic simulation diagram of the motor with torque control predicted by the ADRC model in a specific implementation mode;

[0077] FIG4( a ) is a steady-state simulation diagram of a DV-MPTC motor in a comparative example;

[0078] FIG4( b ) is a steady-state simulation diagram of the TTV-MPTC motor in the comparative example;

[0079] Figure 5This is the dynamic simulation diagram of the traditional PI-MPTC motor in the comparative example. DETAILED DESCRIPTION

[0080] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0081] The present invention provides a three-vector model predictive torque control method for a permanent magnet synchronous motor. Figure 1 As shown, the following steps are included:

[0082] Step S01, setting the running speed of the motor, and grinding different ores at different speeds;

[0083] Step S02: Obtain the current motor speed through the speed sensor, and subtract it from the set speed to obtain the speed error;

[0084] Step S03, design an improved ADRC, input the speed error into the improved ADRC, and obtain the corresponding reference torque; the ADRC controller consists of three parts: Tracking Differentiator (TD), Extended State Observer (ESO), and Nonlinear State Error Feedback (NLSEF). TD can be used to adjust the speed control quantity transition process, thereby resolving the contradiction between the overshoot and rapidity of the motor system. The state variables and unknown disturbances of the system are observed by ESO. Finally, NLSEF is used to transform the control object into a simple integral series system.

[0085] Assume that the set speed is ω ref , the sensor obtains the rotation speed as ω * (k), and the speed expression is obtained through the permanent magnet synchronous motor:

[0086]

[0087] Where, T e is the electromagnetic torque, ω e is the electromagnetic speed of the motor, T L is the load torque; B is the reluctance torque; J is the moment of inertia.

[0088] According to formula (1), the discrete speed loop nonlinear tracking differentiator is designed as follows:

[0089]

[0090] In the formula, e 1D (k) is the speed error at time k; ω refis the speed given signal; x 1D (k) Tracking input signal ω * (k), x 1D (k+1) is the value of x at time k+1 1D (k) predicted value; x 2D (k) is x 1D The differential of (k), x 2D (k+1) is the value of x at time k+1 2D The predicted value of (k). s is the discrete control period; r is the speed factor, which determines the tracking speed of the nonlinear TD; h is the filtering factor, which determines the filtering effect of the nonlinear TD; fst() is the optimal comprehensive function of discrete fast control, which is expressed as:

[0091]

[0092] Where sgn() is the sign function.

[0093] According to the speed outer loop of the control system, equation (1) can be rewritten as:

[0094]

[0095] In the formula, is the total torque disturbance, T L is the motor load, B is the reluctance torque, J is the moment of inertia;

[0096] From this we can get the state equation:

[0097]

[0098] From this, the discrete ESO can be obtained as:

[0099]

[0100] Where fal() is the error nonlinear function; α determines the linearity of the fal function, and its value is 0<α<1; δ is the filtering factor, which is generally 5T s <δ<10T s ; T e * is the torque reference value; ω(k) is the actual motor speed, z 1 (k) is the estimated value of the rotation speed; 1 (k+1) is the k+1 moment z 1 (k) predicted value; z 2 (k) is z 1 The differential of (k) is the total disturbance of the system. 2 (k+1) is the k+1 moment z2 The predicted value of (k), β 1 ,β 2 is the control parameter.

[0101] The nonlinear function fal() is an important part of the nonlinear error feedback law. The traditional nonlinear function is mainly the fal function, which has a certain filtering effect. The traditional function expression is:

[0102]

[0103] However, for the traditional fal function, when |e|=δ, the function is continuous but not smooth. The sudden change of the derivative may cause the system performance to deteriorate. To solve this problem, and based on the principles followed by the fal function, an improved fal function principle is proposed: 1) continuous and smooth at the origin; 2) differentiable and continuous at |e|=δ. Then when |e|≤δ, fal is rewritten as:

[0104]

[0105] To ensure continuity and differentiability at |e| = δ, the fal function value and its derivative must be the same, that is,

[0106]

[0107] The equations in the above formula can be solved together:

[0108]

[0109] In summary, the nfal function can be written as

[0110]

[0111] Finally, the discrete nonlinear state error feedback is designed

[0112]

[0113] Where T e * (k) is the output reference torque, e is the error, u 0 (k) is the control function.

[0114] Step S04: By using the reference torque T e * (k), and the reference flux ψ is generated by formula (15) s * (k)

[0115]

[0116] Where: s* (k) is the motor flux amplitude at time k+1; ψ f is the stator flux; L q is the inductance q; P n is the number of motor pole pairs;

[0117] Step S05: The motor is running, feedback current is obtained, and the current i based on the d and q axes is obtained by transformation. d ,i q , obtain the current motor running torque and motor angle;

[0118] Step S06: According to the feedback parameters obtained in step S05, the predicted values ​​of the torque and flux at the next moment are calculated through the prediction equations of the torque and flux; the cost function of each predicted torque, flux and reference torque and flux is calculated through the improved cost function of the fixed weight factor. Finally, the predicted torque value that minimizes the cost function is found, and the voltage vector combination that produces this predicted torque is found.

[0119] First, the mathematical model of the built-in PMSM in the d and q axis coordinate system can be expressed as:

[0120]

[0121] In the formula, i d 、i q , L d , L q are d-axis current and q-axis current and inductance respectively; u d 、u q is the d and q axis voltage; R s is the stator resistance; ω e is the rotor electrical angular velocity; ψ f is the stator flux.

[0122] The torque equation of PMSM is:

[0123]

[0124] Where, T e is the electromagnetic torque; P n is the number of motor pole pairs.

[0125] The stator flux equation is:

[0126]

[0127] In the formula, ψ s is the motor flux amplitude; ψ d , ψ q is the d and q axis magnetic flux.

[0128] The speed expression of the motor is:

[0129]

[0130] Where, T L is the load torque; B is the reluctance torque; J is the moment of inertia.

[0131] The forward Euler method is used to discretize equation (16), and the stator current prediction equation at time k+1 can be obtained:

[0132]

[0133] In the formula, i d (k), i q (k),u d (k) and u q (k) represents the d-axis and q-axis current and voltage values ​​at time k respectively; i d (k+1), i q (k+1) represents the predicted current value at time k+1.

[0134] Substituting equation (20) into equation (17) and equation (18), the prediction equations of torque and flux linkage are obtained as follows:

[0135]

[0136] In the formula, ψ d (k+1),ψ q (k+1) represents the d-axis and q-axis magnetic flux values ​​at time k+1 respectively.

[0137] Secondly, the torque and flux determination factors are set. In order to simplify the model predictive control process of the three-vector and ensure the prediction accuracy in the selection process of the second and third voltage vectors, the torque and flux determination factors are introduced:

[0138] T i =T e * -T e i (k+1) (22)

[0139] When selecting the flux determination factor, considering that the q axis is mainly responsible for generating torque in the motor structure, in order to simplify the flux prediction process, a vector control strategy is introduced when selecting the second and third voltage vectors. d = 0, at this time, the d-axis flux in equation (17) is only related to the permanent magnet flux; and the permanent magnet flux is a constant, at this time the amplitude of the flux in equation (17) is ψ s and q-axis flux ψ q is proportional to, so we can s The prediction is changed to ψ qFor prediction, the magnetic linkage discrimination factor is set as:

[0140] ψ i =ψ q * -ψ q i (k+1) (23)

[0141] In equations (22) and (23), the value range of i is 1 to 8 of the 8 voltage vectors; T i and ψ i are the determining factors of torque and flux respectively; T e * , ψ q * are the reference torque and q-axis reference flux respectively; T e i (k+1) and ψ q i (k+1) is the torque and q-axis flux parameter value under the action of the i-th voltage vector at time k+1.

[0142] When selecting the vector voltage, assume that the first voltage vector selected is u 1 In order to reduce the amount of calculation, the second voltage vector is selected only from the vectors that have the opposite effect to the second voltage vector. 1 The torque vector T generated by the action e 1 >T e * , if at this time T 1 >0, when the next voltage vector is selected, only the torque smaller than T e * The voltage vector can make the final torque value closer to the reference torque, that is, to choose T 1 *T i The voltage vector is less than 0. Similarly, the effect of the flux determination factor can be analyzed.

[0143] New voltage vector selection

[0144] For a three-phase two-level voltage source inverter, there are eight switching states, which can provide seven different basic voltage vectors, namely six effective voltage vectors and two zero vectors. The three-vector MPTC with fixed coefficients acts on three voltage vectors in one sampling cycle, and the fixed coefficients are only included in the cost function of the first voltage vector. The selection of the first voltage vector is still through the traditional traversal method, but unlike the traditional method, since the torque and flux linkage are predicted and adjusted respectively when the second and third voltage vectors are selected, the adjustment factor of the cost function in the selection of the first voltage vector is a fixed value, and its expression is:

[0145] g 1 =|T e * -T e i (k+1)|+A|ψ s * -ψ s i (k+1)| (24)

[0146] Where A=T N / ψ sN is the weight factor, where T N represents the rated torque, ψ sN is the stator flux amplitude under rated condition. At the same time, let the first optimal voltage vector u opt1 T i =T 1 , ψ i =ψ 1 .

[0147] When selecting the second voltage vector, unlike the traditional voltage vector selection, the second voltage vector of the present invention only selects the voltage vector that makes the torque reach the optimal value, and the improved torque cost function expression is:

[0148] g 2 =(T e * -T e i (k+1)) 2 +C 1 (25)

[0149] In the formula, the torque cost function consists of two parts, namely (T e * -T e i (k+1)) 2 and C 1 ;(T e * -T e i (k+1)) 2 In order to select the voltage vector that achieves the optimal torque, C 1 is the stabilizing factor, ensuring that the remaining factors are stabilized while selecting the torque optimal voltage vector, C 1 The expression is:

[0150]

[0151] In the formula, ω e ref -ω e(k+1) is the speed judgment factor added to consider the stability of the speed outer loop. e ref is the set speed, ω e (k+1) is the predicted torque at time k+1. e The calculation formula for (k+1) is:

[0152]

[0153] After the second voltage vector is selected, the first two voltage vectors are calculated to obtain the corresponding duty cycle d 1 After that, the voltage u under the combined action of the two voltage vectors can be calculated. 1-2 , and regard it as an equivalent voltage vector.

[0154] The q-axis magnetic flux ψ after the action is calculated by the equivalent voltage vector q 1-2 (k+1) and torque T e 1-2 (k+1); calculate T at the same time i =T 2 , ψ i =ψ 2 .

[0155] Similarly, the third voltage vector only selects the voltage vector that makes the flux linkage reach the optimal value, and its cost function expression is:

[0156] g 3 =(ψ q * -ψ q i (k+1)) 2 +C 2 (28)

[0157] In the formula, g 3 It also consists of two parts, namely ψ q * -ψ q i (k+1) and C 2 , ψ q * -ψ q i (k+1) is to select the voltage vector that makes the flux linkage reach the optimal value, C 2 is a stabilizing factor, ensuring that the remaining factors are stable while selecting the optimal voltage vector for flux linkage. Its expression is:

[0158]

[0159] Where ω e(k+1) can be obtained by formula (21). At this time, only the cost function g is selected. 3 The minimum voltage vector u opt3 .

[0160] Action time calculation

[0161] After selecting the voltage vector combination through the improved model prediction torque strategy, three effective vectors are used to control the PMSM in one control cycle, and the duty cycle of the first effective voltage vector is defined as d 1 , the duty cycle of the second effective voltage vector is d 2 , the duty cycle of the third vector is 1-d 1 -d 2 In order to achieve the dual goals of torque and flux linkage control, the voltage vector action time is calculated using the torque and flux linkage deadbeat method. opt1 ,u opt2 ,u opt3 are the first, second and third optimal voltage vectors respectively; u opt1-2 is the voltage vector after the first optimal voltage vector and the second optimal voltage vector act on each other; S T1 , S T2 are the torque slopes under the first optimal voltage vector and the second optimal voltage vector respectively; S ψq1-2 , S ψq3 They are respectively the q-axis flux slope after the first and second optimal voltage vectors act together, and the q-axis flux slope under the action of the third optimal voltage vector.

[0162] First, calculate S by the formula T1 , S T2 , and its calculation formula is as follows:

[0163]

[0164] In the formula, u dopt1 ,u dopt2 ,u qopt1 ,u qopt2 They are the d-axis component and q-axis component under the action of the first and second voltage vectors respectively.

[0165] Using torque deadbeat control, let T e * =T e (k+1), we can get the vector u opt1 Duty cycle. Then d 1 The calculation formula is:

[0166]

[0167] The voltage vector under the combined action of the first and second voltage vectors can be obtained from the above formula:

[0168] u opt1-2 =d 1 u opt1 +(1-d 1 ) opt2 (32)

[0169] S is obtained by the following formula ψq1-2 With S ψq3 The expression is:

[0170]

[0171] Where i dopt1-2 (k+1), i qopt1-2 (k+1) is the value of u opt1-2 Substitute into equation (4) to obtain the voltage vector u opt1-2 d,q axis current under the action of .

[0172] Then, by using flux linkage deadbeat control, the synthetic voltage vector u can be obtained. opt1-2 Duty cycle. 2 The calculation formula is:

[0173]

[0174] Where d 2 is the action time of the synthetic vector.

[0175] The final resultant voltage vector is:

[0176] u s =(u opt1 d 1 T s +u opt2 (1-d 1 )T s )d 2 T s +(1-d 2 ) opt3 T s (36)

[0177] Step S07: Apply the selected optimal voltage vector combination to the motor.

[0178] In order to verify the effectiveness of the proposed control strategy (ITV-MPTC) in improving the steady-state performance and robustness of the motor, the proposed strategy is simulated and analyzed by Matlab / Simulink. The parameters of the PMSM in the simulation are shown in Table 1.

[0179] Table 1 PMSM parameter table

[0180]

[0181] In the experiment, the parameters of ADRC are: α = 0.15, δ = 0.1, β 1 =900,β 2 =202500, compensation factor b 0 =250, adjustment factor k=1500, speed factor r=150000, sampling time 100KHz. The operation result of this control strategy at 5N·m load and 100r / min of the motor, that is, the steady-state simulation diagram of the ITV-MPTC motor, is as follows: Figure 2 From top to bottom in the figure are the simulation diagrams of speed, electromagnetic torque and q-axis current. From the figure, we can get that the pulsation of speed is ±0.02 rpm, while the oscillation amplitude of electromagnetic torque is 0.15N·M, and the fluctuation of q-axis current is only 0.25A.

[0182] In order to verify the dynamic performance and robustness of the proposed torque prediction strategy with improved ADRC model, the dynamic performance of the torque prediction control with improved ADRC model was simulated. The simulation conditions are that the motor is accelerated from static to 200r / min under no-load condition, and a load of 5N·m is suddenly added from no-load at 0.2s, and the load torque is removed at 0.4s to return to no-load condition. The simulation results are shown in Figure 2. Figure 3 As shown, this is the ADRC model predictive torque control motor dynamic simulation diagram.

[0183] When the speed suddenly increases, the strategy proposed in this invention has small overshoot in speed, torque and current; there is almost no overshoot in speed, the overshoot in torque reaches 9N·M, and the q-axis current peak reaches 12A. When the load is suddenly increased, the speed of the strategy proposed in this article only drops to 195r / min; and the speed recovery time and torque arrival time are relatively short. Overall, the control strategy proposed in this article has relatively good improvements in control accuracy and robustness.

[0184] Comparative Example

[0185] Under the same motor working conditions, motor 5N·M load, 100r / min working conditions, the existing traditional dual-vector model predictive control (DV-MPTC) and traditional three-vector model torque control (TTV-MPTC) are compared and simulated. The parameters of the PMSM in the simulation are shown in Table 1. The speed outer loop PI parameters used in the experiment are: K P =0.36,K i=99. In the simulation program, the motor runs at a low speed of 100r / min under a load of 5N·M, and the sampling time is 100KHz. The two existing control strategies, DV-MPTC and TTV-MPTC, are shown in Figure 4(a) and Figure 4(b) from top to bottom, which are the simulation diagrams of speed, electromagnetic torque and q-axis current.

[0186] From Figure 4 (a), it can be seen that the pulsation of the speed of the traditional two-vector model predictive torque control strategy is ±0.8 rpm, while the oscillation amplitude of the electromagnetic torque is 0.47 N·M, and the fluctuation of the q-axis current is 0.8 A. From Figure 4 (b), it can be seen that the pulsation of the speed of the traditional three-vector model predictive torque control strategy is ±0.6 rpm, while the oscillation amplitude of the electromagnetic torque is 0.19 N·M, and the fluctuation of the q-axis current is 0.55 A. It can be seen that the existing traditional control prediction control strategy and the strategy proposed in the present invention are inferior to the control accuracy of the present invention.

[0187] Under the same motor working conditions, the existing MPTC with PI controller (PI-MPTC) was simulated. The simulation conditions were still that the motor accelerated from rest to 200r / min under no-load condition, and a 5N·M load was suddenly added from no-load at 0.2s, and the load torque was removed at 0.4s to return to no-load condition. The traditional PI-MPTC motor dynamic simulation diagram is shown in the figure. Figure 5 As shown. When the speed suddenly increases, the traditional PI model predicts that the torque control strategy has large overshoots in speed, torque and current; the speed overshoot can reach 5 revolutions, the torque overshoot reaches 33N·M, and the q-axis current peak reaches 23A. When the load is suddenly increased, the traditional PI model predicts that the torque control strategy speed drops to 187r / min; and the speed recovery time and torque arrival time are longer than the strategy proposed in this invention.

[0188] In summary, the existing control strategies are inferior to the control strategy proposed in the invention in terms of control accuracy and robustness.

[0189] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. Alterations, modifications, substitutions and variations of the above embodiments by a person skilled in the art are all within the scope of the present invention.

Claims

1. A three-vector model predictive torque control method for a permanent magnet synchronous motor, characterized in that: The steps include: Step S01, setting the running speed of the motor, and grinding different ores at different speeds; Step S02: Obtain the current motor speed through the speed sensor, and subtract it from the set speed to obtain the speed error; Step S03, designing an improved active disturbance rejection controller, inputting the speed error into the improved active disturbance rejection controller, thereby obtaining a corresponding reference torque; The improved active disturbance rejection controller is: Where: e 1D (k) is the speed error at time k; ω ref is the speed given signal; x 1D (k) Tracking input signal ω * (k) value, x 1D (k+1) is the value of x at time k+1 1D (k) predicted value; x 2D (k) is x 1D The differential of (k), x 2D (k+1) is the value of x at time k+1 2D (k) predicted value; T s is the discrete control cycle; r is the speed factor, which determines the tracking speed of the nonlinear TD; h is the filter factor, which determines the filtering effect of the nonlinear TD; fst() is the optimal comprehensive function of discrete fast control; nfal() is the improved error nonlinear function; α determines the linearity of the nfal function, and its value is 0<α<1; δ is the filtering factor, which is generally taken as 5T s <δ<10T s ; T e * is the torque reference value; z1(k) is the estimated value of the speed; z1(k+1) is the predicted value of z1(k) at time k+1; z2(k) is the differential of z1(k), which is the total disturbance of the system; z2(k+1) is the predicted value of z2(k) at time k+1, β1, β2 are control parameters; e is the error, u0(k) is the control function, T e * (k) is the output reference torque; The improved error nonlinear function nfal() is: in, Step S04: Using the reference torque T at time k e * (k), and the reference flux ψ is generated by formula (15) s * (k) Where: s * (k) is the motor flux amplitude at time k; ψ f is the stator flux; L q is the inductance q; P n is the number of motor pole pairs; Step S05: The motor is running, feedback current is obtained, and the current i based on the d and q axes is obtained by transformation. d ,i q , obtain the current motor running torque and motor angle; Step S06: According to the feedback parameters obtained in step S05, the predicted values ​​of the torque and flux at the next moment are calculated by using the prediction equations of the torque and flux; The prediction equations for the torque and flux are as follows: Where: i d 、i q , L d , L q are d-axis current and q-axis current and inductance respectively; i d (k+1), i q (k+1) is the d and q axis current at time k+1; R s is the stator resistance; ω e is the rotor electrical angular velocity; ψ f is the stator flux; T e is the electromagnetic torque; P n is the number of motor pole pairs; ψ s (k+1) is the motor flux amplitude at time k+1; ψ d (k+1),ψ q (k+1) is the d-axis and q-axis magnetic flux at time k+1; T e (k+1) is the torque predicted at time k+1; At the same time, the cost function of the improved fixed weight factor is used to obtain the value of the cost function corresponding to each predicted torque, flux and reference torque and flux. The cost function of the improved fixed weight factor is as follows: g1=|T e * -T e i (k+1)|+A|ψ s * -ψ s i (k+1)| (24) g2=(T e * -T e i (k+1)) 2 +C1 (25) g3=(ψ q * -ψ q i (k+1)) 2 +C2 (28) Where: A = T N / ψ sN is the weight factor, where T N represents the rated torque, ψ sN is the stator flux amplitude under rated condition; T e * is the reference torque; T e i (k+1) is the torque value under the action of the i-th voltage vector at time k+1; ψ s * is the reference flux value; ψ q * is the reference flux value of q axis; ψ q i (k+1) is the q-axis flux value under the action of the i-th voltage vector at time k+1; ψ s i (k+1) is the flux value under the action of the i-th voltage vector at time k+1; C1 and C2 are the flux and torque stability factors respectively; Substitute the voltage vector into formula (21) to obtain the prediction result, substitute the prediction result into the improved cost function of the fixed weight factor, obtain the value of the predicted torque that minimizes the cost function, and obtain the voltage vector combination that produces the predicted torque; Step S07: Apply the optimal voltage vector combination obtained in step S06 to the motor.

2. A three-vector model predictive torque control method for a permanent magnet synchronous motor according to claim 1, characterized in that: The step S03 also includes the following steps: Assume that the set speed is ω ref , the sensor obtains the rotation speed as ω * (k) Design a discrete speed loop nonlinear tracking differentiator as follows: In the formula, e 1D (k) is the speed error at time k; ω ref is the speed given signal; x 1D (k) Tracking input signal ω * (k), x 1D (k+1) is the value of x at time k+1 1D (k) predicted value; x 2D (k) is x 1D The differential of (k), x 2D (k+1) is the value of x at time k+1 2D (k) predicted value; T s is the discrete control period; r is the speed factor, which determines the tracking speed of the nonlinear TD; h is the filtering factor, which determines the filtering effect of the nonlinear TD; fst() is the optimal comprehensive function of discrete fast control, which is expressed as: Where sgn() is the sign function; According to the speed outer loop of the control system, equation (1) is rewritten as: In the formula, is the total torque disturbance, T L is the motor load, B is the reluctance torque, J is the moment of inertia; The state equation is obtained as follows: The discrete ESO is obtained as follows: Where fal() is the error nonlinear function; α determines the linearity of the fal function, and its value is 0<α<1; δ is the filtering factor, which is generally taken as 5T s <δ<10T s ; T e * is the torque reference value; ω(k) is the actual motor speed, z1(k) is the estimated value of the speed; z1(k+1) is the predicted value of z1(k) at time k+1; z2(k) is the differential of z1(k), which is the total disturbance of the system; z2(k+1) is the k+ x The predicted value at time z2(k), T s is the discrete control period, β1, β2 are control parameters; According to the traditional function expression, the fal function is improved. When |e|≤δ, fal is rewritten as: To ensure continuity and differentiability at |e| = δ, the fal function value and its derivative must be the same, that is, Solving the above equations together gives: The nfal function is written as Design of discrete nonlinear state error feedback Where T e * (k) is the output reference torque, e is the error, and u0(k) is the control function.

3. A three-vector model predictive torque control method for a permanent magnet synchronous motor according to claim 1, characterized in that: The step S06 further comprises the following steps: In d. q The mathematical model of the built-in PMSM in the axis coordinate system is expressed as: In the formula, i d 、i q , L d , L q are d-axis current and q-axis current and inductance respectively; u d 、u q is the d and q axis voltage; R s is the stator resistance; ω e is the rotor electrical angular velocity; ψ f is the stator flux; The torque equation of PMSM is: Where, T e is the electromagnetic torque; P n is the number of motor pole pairs; The stator flux equation is: In the formula, ψ s is the motor flux amplitude; ψ d , ψ q They are d-axis and q-axis magnetic flux respectively; According to the motor speed expression, the forward Euler method is used to discretize equation (16), and the stator current prediction equation at time k+1 can be obtained: In the formula, i d (k), i q (k),u d (k) and u q (k) represents the d-axis and q-axis current and voltage values ​​at time k, respectively; i d (k+1), i q (k+1) represents the predicted current value at time k+1, and the prediction equations for torque and flux are: In the formula, ψ d (k+1),ψ q (k+1) represents the d-axis and q-axis flux linkage values ​​at time k+1 respectively; Introducing torque and flux determination factors, the torque is: The magnetic linkage determination factor is set to: Among them, the value range of i is 1 to 8 of the 8 voltage vectors; T i and ψ i are the discriminant factors of torque and flux respectively; T e * , ψ q * are the reference torque and q-axis reference flux respectively; T e i (k+1) and ψ q i (k+1) is the torque and q-axis flux parameter value under the action of the i-th voltage vector at time k+1.

Images