Dual-objective global optimal model-free predictive control method for SMPMSM drive system based on DSVM

Abstract

The present invention relates to a dual-objective global optimal model-free predictive control method for an SMPMSM drive system based on DSVM, including: establishing a hyperlocal model of the SMPMSM drive system to generate an inverter reference voltage vector; dividing the inverter voltage hexagon into three sub-regions according to the position of the generated inverter reference voltage vector; obtaining candidate voltage vectors with optimal current control performance in each sub-region; designing a dual-objective cost function including stator current error and inverter switching times, taking the inverter switching times of the candidate voltage vector with the best current control performance in each sub-region as a reference, and then selecting candidate voltage vectors to be evaluated online in each sub-region and putting them into a candidate voltage vector set; performing online evaluation on the candidate voltage vectors in the candidate voltage vector set based on the dual-objective cost function to obtain a globally optimal voltage vector. The present invention has the advantages of ensuring that the system obtains a globally optimal inverter voltage vector, having a small calculation load, and strong robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of SMPMSM drive systems, and in particular to a dual-objective global optimal model-free predictive control method for an SMPMSM drive system based on DSVM. Background Art

[0002] Permanent magnet synchronous motors (PMSMs) have the advantages of high efficiency, high power density, and low maintenance, and are widely used in new energy vehicles, elevators, air compressors, etc. In practical applications, in order to achieve high-quality operation of the PMSM drive system, it is urgent to break through the key technologies of dual-objective optimization control of the motor current and inverter switching frequency of the PMSM drive system.

[0003] Model Predictive Control (MPC) has received research attention due to its clear physical concept, flexible control structure, high dynamic response, and easy implementation in PMSM drive systems. MPC can be divided into Continuous Control Set Model Predictive Control (CCS-MPC) and Finite Control Set Model Predictive Control (FCS-MPC). Discrete Space Vector Modulation (DSVM) expands the number of inverter candidate voltage vectors by introducing virtual voltage vectors. For a PMSM drive system based on FCS-MPC of DSVM, single-vector control, dual-vector control, or triple-vector control can be automatically selected according to different operating conditions of the system, increasing the degree of freedom of inverter voltage vector selection. In addition, with the increase of the DSVM time interval of the inverter, the number of candidate voltage vectors in the inverter voltage vector hexagon also increases, further reducing the torque ripple during the steady-state operation of the PMSM drive system.

[0004] For the finite control set model predictive control with dual-objective optimization of the current control performance and inverter switching frequency of the SMPMSM drive system based on DSVM, the inverter voltage vectors obtained by existing methods may only be local optimal solutions. The enumeration method that enumerates all inverter candidate voltage vectors and minimizes the dual-objective cost function can obtain the global optimal inverter voltage vector, but it needs to enumerate all inverter candidate voltage vectors, resulting in an increase in the online calculation burden of the controller. In addition, the system control performance is sensitive to the modeling accuracy of the SMPMSM drive system. Summary of the Invention

[0005] The object of the present invention is to provide a model-free predictive control method for dual-objective global optimization of a SMPMSM drive system based on DSVM, which does not rely on accurate modeling of the SMPMSM drive system, does not require online evaluation of all inverter voltage vectors, and can generate an inverter voltage vector that is globally optimal for both current control performance and inverter switching frequency in real time. To achieve the above object, the present invention adopts the following technical solutions: A model-free predictive control method for dual-objective global optimization of a SMPMSM drive system based on DSVM, the method comprising the following steps in sequence:

[0006] (1) Establish a super-local model of the SMPMSM drive system to generate the inverter reference voltage vector u * αβ (k);

[0007] (2) According to the position of the inverter reference voltage vector u * αβ (k), divide the inverter voltage hexagon into three sub-regions, and then define the sub-region where the inverter reference voltage vector u * αβ (k) is located as The other two sub-regions are defined as and

[0008] (3) Obtain the candidate voltage vectors with the best current control performance in the sub-regions and respectively;

[0009] (4) Generate the candidate voltage vectors to be evaluated online in each sub-region and put them into the candidate voltage vector set;

[0010] (5) Online evaluate the candidate voltage vectors in the candidate voltage vector set to obtain the globally optimal inverter voltage vector

[0011] The specific content of step (1) is as follows:

[0012] In the dq synchronous speed rotating coordinate system, establish a mathematical model of the SMPMSM drive system including motor parameter uncertainty, inverter nonlinearity and unknown disturbance, expressed as:

[0013]

[0014] In the formula: i d , i q represent the d-axis and q-axis stator currents, u* d and u* q represent the d-axis and q-axis reference voltages of the inverter; n p is the number of pole pairs; Ω ris the mechanical angular velocity of the motor; R s , L s and represent the nominal parameters of the stator resistance, stator inductance, and permanent magnet flux linkage, respectively; f ds and f qs represent the disturbances caused by the uncertainty of the motor parameters; V d,dead and V q,dead represent the disturbances caused by the inverter non-linearity, d d and d q are unknown disturbances, α d and α q represent the proportionality coefficients of the d, q-axis reference voltages of the inverter;

[0015] Based on the model-free control, a super-local model of the SMPMSM drive system is established:

[0016]

[0017] In the formula:

[0018] The estimations of F d and F q are obtained through the algebraic parameter identification method. The estimation expressions of F d and F q are:

[0019]

[0020] In the formula: T F = n F Ts, n F is the window length, Ts is the sampling time; δ is the independent variable of the integral, u d (δ) and u q (δ) represent the d, q-axis reference voltages of the inverter at the δ moment, respectively. i d (δ) and i q (δ) represent the d, q-axis stator sampled currents at the δ moment, respectively;

[0021] Starting from the k-th sampling moment to the end of the d, q-axis reference voltages of the inverter calculated based on the data at the k-th sampling moment, there is a delay of two control periods; Assume and are respectively equal to and By performing Euler discretization and delay compensation on formula (1), the predicted values of i d and i q at the (k + 2)-th moment are i d (k + 2) and i q(k + 2), expressed as:

[0022]

[0023] According to the deadbeat predictive control, the d - axis and q - axis stator currents reach the reference values at the (k + 2)th moment Generate the reference voltages of the d - axis and q - axis of the inverter at the kth moment, expressed as:

[0024]

[0025] Then, generate the inverter reference voltage vector u * αβ (k).

[0026] The specific content of step (2) is as follows:

[0027] According to the position of the inverter reference voltage vector u * αβ (k), taking the center of the inverter voltage hexagon as the origin, divide the inverter voltage hexagon into three sub - regions according to three angular ranges of [0, 2π / 3], [2π / 3, 4π / 3], and [4π / 3, 2π]. The three sub - regions are defined as Z1, Z2, and Z3 in sequence;

[0028] Since the position of the generated u * αβ (k) is not fixed, define the sub - region where u * αβ (k) is located as Denote the ceiling function, and the other two sub - regions are defined as in the counter - clockwise order as and

[0029] The specific content of step (3) includes the following steps:

[0030] (3a) Obtain the inverter candidate voltage vector with the optimal current control performance in the sub - region :

[0031] Based on the minimum - distance principle, the inverter candidate voltage vector closest to the inverter reference voltage vector u * αβ (k) is the candidate voltage vector with the optimal current control performance. The inverter candidate voltage vectors include the inverter basic voltage vectors and the virtual voltage vectors. The sub - region is divided by a grid with a height of width U dc / 3N, and the two candidate voltage vectors at each grid vertex are defined as j = 1, 2, The expressions of the α-axis and β-axis are:

[0032]

[0033] where: N is the time interval of discrete space vector modulation (DSVM), and U dc is the DC bus voltage of the inverter;

[0034] sub-region The candidate voltage vector with the optimal current control performance in is the voltage vector closest to in and is defined as

[0035]

[0036] (3b) Obtain the candidate voltage vectors with the optimal current control performance in sub-regions and :

[0037] In and , the candidate voltage vectors closest to are respectively on the common side of and and on the common side of and . Project onto these two common sides respectively. The candidate voltage vector closest to the projection point is the candidate voltage vector with the optimal current control performance in and ; when the projection point is not on the common side, the candidate voltage vector with the optimal current control performance is V0; Define the candidate voltage vectors with the optimal current control performance in and as and respectively, which is expressed as:

[0038]

[0039]

[0040] where: If round represents the rounding operation;

[0041] where c and d are auxiliary variables; m1, m2, and m3 are sub-region numbers, and θ is the motor rotor position angle.

[0042] The specific content of step (4) is as follows:

[0043] For the SMPMSM drive system with finite control set model-free predictive control, a dual-objective cost function including current control performance and inverter switching frequency is designed, which can be expressed as:

[0044]

[0045] In the formula: n represents the candidate voltage vector of the inverter to be evaluated, E n represents the value of its current error cost function, J n represents the value of its dual-objective cost function, S n (k + 2) represents the corresponding number of inverter switches, and λ is the weight factor;

[0046] In the sub-region the candidate voltage vector with the optimal current control performance generates the minimum current error. If the number of switches of the candidate voltage vector (CVV) of the inverter is greater than or equal to the number of switches, then the value of its dual-objective cost function is greater than Therefore, only and the candidate voltage vectors with the number of inverter switches less than are put into the candidate voltage vector set, and only the candidate voltage vectors in the candidate voltage vector set need to be evaluated online; using the same method, in the sub-regions and in, and the candidate voltage vectors with the number of inverter switches less than and are respectively put into the candidate voltage vector set.

[0047] The specific content of step (5) is as follows: Substitute the voltage vectors in the candidate voltage vector set into the designed dual-objective cost function, and the candidate voltage vector corresponding to the minimum value of the dual-objective cost function is the global optimal voltage vector The inverter adopts a discontinuous minimum modulation strategy to generate the on and off signals of the inverter power switching devices, and controls the operation of the inverter in real time.

[0048] As can be seen from the above technical solutions, the beneficial effects of the present invention are as follows: First, by establishing a super-local model of the SMPMSM drive system, the inverter reference voltage vector u * αβ (k) is generated through deadbeat predictive control, and then according to u * αβThe position of (k) divides the inverter voltage hexagon into three sub-regions, and a method for obtaining candidate voltage vectors with the best current control performance in each sub-region is proposed, overcoming the deficiency that the system current control performance is sensitively dependent on the system modeling accuracy; second, a dual-objective cost function including the stator current error and the number of inverter switchings is designed. Taking the number of inverter switchings of the candidate voltage vector with the best current control performance in each sub-region as a benchmark, the candidate voltage vectors to be evaluated online in each sub-region are determined and put into the candidate voltage vector set, and then the candidate voltage vectors in the candidate voltage vector set are evaluated online based on the designed dual-objective cost function to obtain the globally optimal inverter voltage vector, avoiding the online evaluation of all candidate voltage vectors. As shown by the experimental results, the present invention has the advantage of ensuring that the system obtains the globally optimal inverter voltage vector under different weight factors and different working conditions. Description of the Drawings

[0049] Figure 1 Candidate voltage vectors of DSVM with a time interval of 5;

[0050] Figure 2 Schematic diagram of the proposed sub-region division method;

[0051] Figure 3 For Schematic diagram of the division method;

[0052] Figure 4 For And Schematic diagram of the candidate voltage vector with the best current control performance in

[0053] Figure 5 Flowchart of the method of the present invention;

[0054] Figure 6 Control structure diagram of the proposed SMPMSM drive system;

[0055] Figure 7 Schematic diagram of the dual-objective performance comparison of SMPMSM with a speed of 100 rpm and a q-axis current reference value of 10 A;

[0056] Figure 8 Schematic diagram of the dual-objective performance comparison of SMPMSM with a speed of 100 rpm and a q-axis current reference value of 20 A;

[0057] Figure 9 Schematic diagram of the dual-objective performance comparison of SMPMSM with a speed of 500 rpm and a q-axis current reference value of 10 A;

[0058] Figure 10Schematic diagram for dual - objective performance comparison when the speed of SMPMSM is 500 rpm and the q - axis current reference value is 20 A;

[0059] Figure 11 Schematic diagram for comparison of execution times of three control methods;

[0060] Figure 12 Schematic diagram for comparison between the model - based method and the proposed method under parameter uncertainties. Detailed implementation manners

[0061] As Figure 5 shown, a dual - objective global - optimal model - free predictive control method based on an SMPMSM drive system, the method includes the following steps in sequence:

[0062] (1) Establish a super - local model of the SMPMSM drive system to generate the inverter reference voltage vector u * αβ (k);

[0063] (2) According to the position of the inverter reference voltage vector u * αβ (k), divide the inverter voltage hexagon into three sub - regions, and then define the sub - region where the inverter reference voltage vector u * αβ (k) is located as Define the other two sub - regions as and

[0064] (3) Respectively obtain the candidate voltage vectors with the best current control performance in the sub - regions and ;

[0065] (4) Generate the candidate voltage vectors to be evaluated online in each sub - region and put them into the candidate voltage vector set;

[0066] (5) Conduct online evaluation on the candidate voltage vectors in the candidate voltage vector set to obtain the globally optimal inverter voltage vector

[0067] The specific content of step (1) is:

[0068] In the dq synchronous - speed rotating coordinate system, establish a mathematical model of the SMPMSM drive system including motor parameter uncertainties, inverter nonlinearities, and unknown disturbances, expressed as:

[0069]

[0070] Where: i d 、i qrepresent the d-axis and q-axis stator currents, u* d and u* q represent the d-axis and q-axis reference voltages of the inverter; n p is the number of pole pairs; Ω r is the mechanical angular velocity of the motor; R s , L s and represent the nominal parameters of the stator resistance, stator inductance, and permanent magnet flux linkage respectively; f ds and f qs represent the disturbances caused by the uncertainties of the motor parameters; V d,dead and V q,dead represent the disturbances caused by the non-linearity of the inverter, d d and d q are unknown disturbances, α d and α q represent the proportionality coefficients of the d-axis and q-axis reference voltages of the inverter;

[0071] Based on the model-free control, a super-local model of the SMPMSM drive system is established:

[0072]

[0073] In the formula:

[0074] The estimations of F d and F q are obtained through the algebraic parameter identification method, and the estimation expressions of F d and F q are:

[0075]

[0076] In the formula: T F = n F Ts, n F is the window length, Ts is the sampling time; δ is the independent variable of the integral, u d (δ) and u q (δ) represent the d-axis and q-axis reference voltages of the inverter at the δth moment respectively, and i d (δ) and i q (δ) represent the d-axis and q-axis stator sampled currents at the δth moment respectively;

[0077] Starting from the kth sampling moment to the end of the d-axis and q-axis reference voltages of the inverter calculated based on the data at the kth sampling moment, there is a delay of two control periods; Assume and are respectively equal to and By performing Euler discretization and delay compensation on formula (1), we can obtain i at the (k+2)th moment: d and i q The predicted value of i d (k+2) and i q (k+2), expressed as:

[0078]

[0079] According to the deadbeat predictive control, the d-axis and q-axis stator currents reach the reference value at the (k+2)th moment. Generate the inverter d and q axis reference voltage at time k, expressed as:

[0080]

[0081] Then, the inverter reference voltage vector u is generated by coordinate transformation * αβ (k).

[0082] like Figure 2 As shown, the step (2) specifically refers to:

[0083] According to the inverter reference voltage vector u * αβ (k), taking the center of the inverter voltage hexagon as the origin, the inverter voltage hexagon is divided into three sub-areas according to the three angle ranges of [0, 2π / 3], [2π / 3, 4π / 3], and [4π / 3, 2π]. The three sub-areas are defined as Z1, Z2, and Z3 respectively;

[0084] Since the generated u * αβ The position of (k) is not fixed. * αβ The sub-region where (k) is located is defined as Indicates rounding up, and the other two sub-areas are defined in counterclockwise order as and

[0085] The step (3) specifically comprises the following steps:

[0086] (3a) Get sub-region The candidate inverter voltage vector with the best current control performance in is:

[0087] like Figure 3 As shown, based on the minimum distance principle, the distance inverter reference voltage vector The nearest inverter candidate voltage vector is the candidate voltage vector with the optimal current control performance. The inverter candidate voltage vector includes the basic voltage vector of the inverter and the virtual voltage vector, and the sub-region is divided by a grid with a width of U / 3N. The two candidate voltage vectors at each grid vertex are defined as dc j = 1, 2, The expressions of the α-axis and β-axis are:

[0088]

[0089] In the formula: N is the time interval of Discrete Space Vector Modulation (DSVM), and U dc is the DC bus voltage of the inverter;

[0090] The candidate voltage vector with the optimal current control performance in the sub-region is the voltage vector in that is closest to Figure 3 As shown in it is defined as

[0091]

[0092] (3b) Obtain the candidate voltage vectors with the optimal current control performance in the sub-regions and :

[0093] As shown in Figure 4 in and the candidate voltage vectors closest to are on the common side of and and the common side of and respectively. Project onto these two common sides respectively. The candidate voltage vector closest to the projection point is the candidate voltage vector with the optimal current control performance in and ; when the projection point is not on the common side, the candidate voltage vector with the optimal current control performance is V0; define the candidate voltage vectors with the optimal current control performance in and as and respectively. As shown in Figure 4 it is expressed as:

[0094] ​

[0095]

[0096] Wherein: If round represents the rounding operation;

[0097] Wherein, c and d are auxiliary variables; m1, m2, and m3 are sub-region numbers, and θ is the motor rotor position angle.

[0098] The specific step (4) refers to:

[0099] For the SMPMSM drive system with finite control set model-free predictive control, design a dual-objective cost function including current control performance and inverter switching frequency, which can be expressed as:

[0100]

[0101] Wherein: n represents the candidate voltage vector of the inverter to be evaluated, E n represents the value of its current error cost function, and J n represents the value of its dual-objective cost function, and S n (k + 2) represents the corresponding number of inverter switches, and λ is a weighting factor;

[0102] In the sub-region the candidate voltage vector with the optimal current control performance generates the minimum current error. If the number of switches of the candidate voltage vector (CVV) of the inverter is greater than or equal to the number of switches, then the value of its dual-objective cost function is greater than Therefore, only and the candidate voltage vectors with the number of inverter switches less than are put into the candidate voltage vector set, and only the candidate voltage vectors in the candidate voltage vector set need to be evaluated online; using the same method, in the sub-regions and the and the candidate voltage vectors with the number of inverter switches less than and are respectively put into the candidate voltage vector set.

[0103] The specific step (5) refers to: substituting the voltage vectors in the candidate voltage vector set into the designed dual-objective cost function, and the candidate voltage vector corresponding to the minimum value of the dual-objective cost function is the global optimal inverter voltage vector The inverter adopts a discontinuous minimum value modulation strategy to generate on and off signals of the inverter power switching devices, and controls the operation of the inverter in real time.

[0104] The following is a further description of the present invention in conjunction with Figures 1 to 12 this.

[0105] For DSVM, the virtual voltage vector can be synthesized by applying several voltage vectors within the control period. For a three-phase two-level voltage source inverter, when the time interval of DSVM is 5, the candidate voltage vectors are as Figure 1 shown by the dots.

[0106] In the inverter voltage hexagon, there are inverter voltage vectors with the same number of switching times, as Figure 1 shown by the gray dots in. If these voltage vectors are substituted into Equation (10) for evaluation, only the voltage vector with the optimal current control performance can be obtained, and the dual-objective optimization control degenerates into a single-current-objective optimization control. Therefore, the area surrounded by the gray dots is the dual-objective invalid optimization area of the stator current and the inverter switching frequency, as Figure 1 shown by the gray area in. The traditional dual-objective FCS-MPC based on DSVM only evaluates the three candidate voltage vectors around u * αβ (k), which is called the three-candidate voltage vector method. If u * αβ (k) is in the invalid optimization area, as Figure 1 shown by the lower triangle in, the cost function can only select the candidate voltage vector with the optimal current control performance, and the switching times of the inverter cannot be reduced, that is, the switching frequency of the inverter cannot be reduced. In addition, the traditional dual-objective FCS-MPC based on DSVM generates u * αβ (k) based on the mathematical model of the SMPMSM drive system, and then selects the candidate voltage vectors around u * αβ (k) for evaluation. However, the motor parameters in the actual system are uncertain, the inverter is nonlinear, and the unknown disturbances affect the accurate generation of u * αβ (k) and the selection of the candidate voltage vectors around u * αβ (k). Therefore, the traditional dual-objective FCS-MPC based on DSVM cannot ensure obtaining the globally optimal inverter voltage vector for dual-objective optimization, and there is a deficiency that the system control performance is sensitive to the system modeling accuracy.

[0107] The control structure diagram of the proposed SMPMSM drive system is as Figure 6 shown. The three-phase current sensors and the rotary encoder collect the current i in real timea (k), i b (k), i c (k) and the rotor position angle θ, the dq-axis sampled current i is obtained through coordinate transformation dq (k) and the reference voltage u * dq(k); Based on the sampled current i dq (k) and the reference voltage u * dq(k) estimates the parameter uncertainty, inverter nonlinearity and position disturbance to obtain and performs delay compensation on the sampled current to obtain the dq-axis current i at time k+2 dq (k+2); Substitute i dq (k+2), i dq (k) into the proposed double-objective global optimal model-free predictive control to obtain the global optimal inverter voltage vector The inverter adopts a discontinuous minimum PWM modulation strategy to generate the on and off signals of the inverter power switching devices, and real-time controls the operation of the inverter.

[0108] Online evaluating all candidate voltage vector methods can obtain the global optimal solution and the minimum cost function value. Therefore, taking the cost function value of the online evaluating all candidate voltage vector method as the reference value, if the cost function values obtained by the proposed method and the three-candidate voltage vector method are greater than the reference value, the obtained is a sub-optimal solution. On the contrary, if the cost function values obtained by the proposed method and the three-candidate voltage vector method are equal to the reference value, the global optimal solution is obtained.

[0109] When setting the speed of the SMPMSM drive system to 100 rpm, the d-axis reference current to 0 A, and the q-axis reference currents to 10 A and 20 A respectively, the comparison of the double-objective control performances of the online evaluating all candidate voltage vector method and the three-candidate voltage vector method is as Figure 7 and Figure 8 shown.

[0110] When the weight factor is 0.05, the three-candidate voltage vector method and the proposed method have similar dq-axis current ripples, phase current THD and inverter switching frequency, but the three-candidate voltage vector method selects a small number of sub-optimal voltage vectors. In addition, as the weight factor increases, the proportion of sub-optimal voltage vectors of the three-candidate voltage vector method also increases significantly. In addition, the inverter switching frequency cannot be reduced as the weight factor increases. In contrast, the method proposed in the present invention can obtain the double-objective global optimal voltage vector under different weight factor conditions, at the cost of a slight increase in the A-phase current THD, significantly reducing the inverter switching frequency. In addition, the method proposed in the present invention retains the double-objective compromise effect of the weight factor, allowing the designer to adjust the current control performance and the inverter switching frequency according to actual needs.

[0111] When the rotational speed of the SMPMSM drive system is set to 100 rpm, the d-axis reference current is 0 A, and the q-axis reference currents are 10 A and 20 A respectively, the comparison of the dual-objective control performance of all candidate voltage vector methods and the three-candidate voltage vector method is evaluated online as follows Figure 9 and Figure 10 shown

[0112] When the weight factor is 0.05, both methods obtain the global optimal voltage vector. This is because at high speeds, the trajectory of the reference voltage vector u*(k) is close to the voltage hexagon boundary, and the three voltage vectors evaluated by the three-candidate voltage vector method are not in the invalid optimization region. However, as the weight factor increases, the inverter switching frequency of the three-candidate voltage vector method cannot be further reduced, and the dual-objective global optimal voltage vector cannot be obtained. The method proposed in the present invention can obtain the global optimal voltage vector under different weight factors and ensures the dual-objective compromise function of the weight factor αβ (k).

[0113] Figure 11 The comparison of the execution times of three control methods is shown. Compared with the three-candidate voltage vector method, the execution time of the method proposed in the present invention increases slightly, but is significantly lower than that of the method of evaluating all candidate voltage vectors

[0114] To verify the robustness of the proposed method, a comparative study on the current control performance of the model-based method and the method proposed in the present invention was carried out when the motor parameters changed. For the model-based method, the mathematical model of the SMPMSM drive system was used to generate the inverter reference voltage vector and predict the current, and the rest was the same as the method proposed in the present invention. For both methods, the motor parameters were set to R = 1.4R s , L = 1.2L s , and the remaining experimental conditions were the same as those of the previous experiment

[0115] Let increase from 0 A to 10 A at 0.02 s, and the dq-axis currents of the two methods are as follows Figure 12 shown Figure 12Among them, both methods have fast dynamic responses at 100 rpm and 400 rpm. However, under steady-state conditions, the dq-axis current ripples of the model-based method are significantly higher than those of the proposed method. The reason is that when the motor parameters change, the inverter reference voltage vectors and predicted currents generated in the model-based method are inaccurate, and the cost function cannot select appropriate candidate voltage vectors, resulting in a significant increase in dq-axis current ripples. The method proposed in the present invention does not rely on the accurate modeling of the SMPMSM drive system. Therefore, even when the motor parameters are uncertain, it still has good dynamic and steady-state control performances, which verifies the robustness of the proposed method.

[0116] In summary, the present invention establishes a super-local model of the SMPMSM drive system, getting rid of the dependence on the accurate modeling of the SMPMSM drive system; based on the established super-local model of the SMPMSM drive system, generates the reference voltage vector u* αβ (k) of the inverter, and then divides the inverter voltage hexagon into three sub-regions according to the position of u* αβ (k), and proposes a method to obtain candidate voltage vectors with the best current control performance in each sub-region; designs a dual-objective cost function including stator current error and inverter switching times, takes the inverter switching times of the candidate voltage vector with the best current control performance in each sub-region as a benchmark, determines the candidate voltage vectors to be evaluated online in each sub-region and puts them into the candidate voltage vector set, and then conducts online evaluation on the candidate voltage vectors in the candidate voltage vector set based on the designed dual-objective cost function to obtain the global optimal voltage vector, avoiding online evaluation of all candidate voltage vectors. As shown by the experimental results, the present invention has the technical advantages of ensuring the acquisition of the global optimal voltage vector, small computational load, and strong robustness.

Claims

1. A dual-objective global optimal model-free predictive control method for a SMPMSM drive system based on DSVM, characterized in that: The method includes the following steps in sequence: (1) Establish the super-local model of the SMPMSM drive system to generate the inverter reference voltage vector u * αβ (k); (2) According to the position of the inverter reference voltage vector u * αβ (k), divide the inverter voltage hexagon into three sub-regions, and then define the sub-region where the inverter reference voltage vector u * αβ (k) is located as Define the other two sub-regions in counterclockwise order as and (3) Obtain the candidate voltage vectors with the best current control performance in the sub-region and respectively. (4) Generating candidate voltage vectors to be online evaluated in each sub-region and putting them into a candidate voltage vector set; (5) Online evaluate the candidate voltage vectors in the candidate voltage vector set to obtain the globally optimal inverter voltage vector The specific steps of step (3) include the following steps: (3a) Obtain the inverter candidate voltage vector with the optimal current control performance in the sub-region : Based on the minimum distance principle, the distance to the inverter reference voltage vector The nearest inverter candidate voltage vector is the candidate voltage vector with the optimal current control performance. The inverter candidate voltage vectors include the inverter basic voltage vectors and the virtual voltage vectors. The sub-region is divided by a grid with a width of U / 3N. The two candidate voltage vectors at each grid vertex are defined as dc The expressions of the α-axis and β-axis of are: In the formula: N is the time interval of discrete space vector modulation, and U dc is the DC bus voltage of the inverter; Sub-region The candidate voltage vector with the optimal current control performance in In and The voltage vector closest to And there is: (3b) Obtain the candidate voltage vector with the optimal current control performance in and : In and , the candidate voltage vectors closest to the distance are respectively on the common side of and and on the common side of and . Project onto these two common sides respectively. The candidate voltage vector closest to the projection point is the candidate voltage vector with the optimal current control performance in and ; when the projection point is not on the common side, the candidate voltage vector with the optimal current control performance is V0; Define the candidate voltage vectors with the optimal current control performance in and as and respectively, which is expressed as: In the formula: If round represents the rounding operation; Wherein, c and d are auxiliary variables; m1, m2, m3 are sub-region numbers, and θ is the motor rotor position angle.

2. The dual-objective global optimal model-free predictive control method for the DSVM-based SMPMSM drive system according to claim 1, characterized in that: The specific meaning of step (1) is: In the dq synchronous speed rotating coordinate system, establishing a mathematical model of the SMPMSM drive system including motor parameter uncertainties, inverter nonlinearities, and unknown disturbances, expressed as: Where: i d and i q represent the d - axis and q - axis stator currents, u* d and u* q represent the d - axis and q - axis reference voltages of the inverter; n p is the number of pole pairs; Ω r is the mechanical angular velocity of the motor; R s , L s and represent the nominal parameters of the stator resistance, stator inductance, and permanent - magnet flux linkage respectively; f ds and f qs represent the disturbances caused by the uncertainties of the motor parameters; V d,dead and V q,dead represent the disturbances caused by the nonlinearity of the inverter, d d and d q are unknown disturbances, α d and α q represent the proportionality coefficients of the d - axis and q - axis reference voltages of the inverter; Based on model-free control, establishing a super-local model of the SMPMSM drive system: In the formula: F d and F q are estimated by an algebraic parameter identification method. For F d and F q the estimated expression is: Where: T F = n F Ts,n F is the window length, Ts is the sampling time; δ is the independent variable of integration, u d (δ) and u q (δ) respectively represent the reference voltages of the d and q axes of the inverter at the δ moment, i d (δ) and i q (δ) respectively represent the sampled stator currents of the d and q axes at the δ moment; There is a delay of two control cycles from the k-th sampling moment to the end of the inverter d-axis and q-axis reference voltages calculated based on the data at the k-th sampling moment; assume and are respectively equal to and By performing Euler discretization and delay compensation on Equation (1), the predicted values of i d and i q at the (k + 2)-th moment, i d (k + 2) and i q (k + 2), are expressed as: According to deadbeat predictive control, the stator currents on the d- and q-axes reach the reference values at the (k + 2)-th moment Generate the reference voltages on the d- and q-axes of the inverter at the k-th moment, expressed as: Then, the reference voltage vector u of the inverter is generated through coordinate transformation * αβ (k).

3. The dual-objective global optimal model-free predictive control method for the DSVM-based SMPMSM drive system according to claim 1, wherein: The specific meaning of step (2) is: According to the position of the inverter reference voltage vector u * αβ (k), with the center of the inverter voltage hexagon as the origin, the inverter voltage hexagon is divided into three sub-regions according to three angular ranges of [0, 2π / 3], [2π / 3, 4π / 3], and [4π / 3, 2π]. The three sub-regions are defined as Z1, Z2, and Z3 in sequence; Since the position of the generated u * αβ (k) is not fixed, the sub-region where u * αβ (k) is located is defined as denotes rounding up, and the other two sub-regions are defined in counterclockwise order as and 4. The dual-objective global optimal model-free predictive control method for the DSVM-based SMPMSM drive system according to claim 1, wherein: The specific meaning of step (4) is: For the SMPMSM drive system with finite control set model-free predictive control, designing a dual-objective cost function including current control performance and inverter switching frequency, which is expressed as: where: n represents the candidate voltage vector of the inverter to be evaluated, E n represents the value of its current error cost function, J n represents the value of its bi-objective cost function, S n (k + 2) represents the corresponding number of inverter switches, and λ is the weight factor; In the sub-region the candidate voltage vector with the optimal current control performance generates the minimum current error. If the switching times of the candidate voltage vector of the inverter are greater than or equal to the number of switchings, then its double-objective cost function value is greater than Therefore, only and the candidate voltage vectors with the switching times of the inverter less than are put into the candidate voltage vector set, and only the candidate voltage vectors in the candidate voltage vector set need to be evaluated online; Using the same method, in the sub-regions and where and the number of inverter switching times is less than and the candidate voltage vectors are respectively put into the candidate voltage vector set.

5. The dual-objective global optimal model-free predictive control method for the SMPMSM drive system based on DSVM according to claim 1, characterized in that: The specific content of step (5) is as follows: Substitute the voltage vectors in the candidate voltage vector set into the bi-objective cost function, and the candidate voltage vector corresponding to the minimum value of the bi-objective cost function is the globally optimal inverter voltage vector. The inverter adopts a discontinuous minimum modulation strategy to generate the on and off signals of the inverter power switching devices, and controls the operation of the inverter in real time.

Images