Permanent magnet motor model predictive control method and system based on triangular discrete space

Abstract

The present invention discloses a model predictive control method and system for a permanent magnet motor based on a triangular discrete space. The method proposes a voltage vector discretization strategy based on an equilateral triangle, and generates up to 6×4 n‑1 A virtual voltage vector is generated, and the current at the next moment is predicted using the virtual voltage vector. Finally, the optimal voltage vector is selected with the minimum cost function value as the goal for motor control. In addition, based on the virtual voltage vector constructed based on the midpoint of the equilateral triangle, the present invention explores the specific geometric relationship between the virtual voltage vector constructed in each layer of iterative segmentation process and the optimal voltage vector, thereby realizing efficient optimal voltage vector search and effectively reducing the number of voltage vectors evaluated online, thereby greatly reducing the computational burden. In summary, the technical solution of the present invention can not only ensure the precise control of the permanent magnet motor, but also effectively reduce the current harmonics of the motor and reduce thrust fluctuations.

Description

Technical Field

[0001] The present invention belongs to the technical field of motor control, and in particular relates to a permanent magnet motor model predictive control method and system based on triangular discrete space. Background Art

[0002] The permanent magnet motor (PMM)—inspired by radially splitting and flattening a rotating permanent magnet motor—generates thrust directly in a specific direction without requiring an intermediate mechanical transmission. Its advantages include compact size, simple structure, and easy maintenance. It has become a green and efficient drive option in a wide range of fields, including industrial servos, rail transportation, and aerospace.

[0003] In terms of control strategies, permanent magnet motors commonly employ field-oriented control, direct torque control, and model predictive control (MPC). Among these, MPC is particularly noteworthy. It eliminates the traditional inner current loop and its parameter tuning process, eliminating the need for a hysteresis regulator. Instead, it directly determines the inverter drive signal through optimization of a value function, thus eliminating the complex steps of pulse-width modulation. Furthermore, this method offers excellent system constraint handling capabilities and control target scalability, along with a simple structure and rapid dynamic response.

[0004] However, traditional model predictive control methods have limitations. They use only one base voltage vector in each control cycle, which can deviate significantly from the actual optimal vector, leading to significant voltage errors, large current harmonics, and thrust fluctuations. In recent years, domestic and international scholars have conducted in-depth research and exploration to address this prominent issue, proposing improved model predictive control methods based on the modulation of multiple voltage vectors (e.g., two or three), such as new MPC strategies based on discrete space vector modulation (DSVM). These strategies use three base voltage vectors VV to generate the optimal voltage vector VV, which helps reduce current ripple and thrust fluctuations. These DSVM-based methods involve a large number of virtual voltage vectors VV, which increases the computational load during online calculations and poses certain challenges to practical applications.

[0005] The conventional discrete space vector method is to use the inscribed circle of a regular hexagon as the basis, and evenly divide the voltage vector plane with the help of the two dimensions of amplitude and phase angle, or CN116800144A adopts a mixed division, and its inverter output voltage vector plane can be regarded as divided into 6 triangles and discretized on the midline of each triangle. However, the division of this technology is uneven, and the maximum error in different areas is different, which makes the motor control performance unstable; such as CN116317795A, a permanent magnet motor current prediction control method based on discrete space vector modulation, performs voltage vector division on the hexagonal contour and diagonal line. It is similar to the method of CN116800144A, which can only select fixed hexagonal contours and diagonals, resulting in large control errors.

[0006] Therefore, to further enhance the performance of model predictive control, it is necessary to optimize the voltage vector plane segmentation technique to generate virtual voltage vectors that better meet the requirements. Furthermore, the optimal voltage vector acquisition method should be optimized to improve computational efficiency, thereby effectively suppressing current harmonics and thrust fluctuations. This will help achieve high-performance control goals for permanent magnet motors and promote their application and development in more fields. Summary of the Invention

[0007] The present invention aims to optimize the method for obtaining the optimal voltage vector and address the problems of high current harmonics and large thrust fluctuations in existing predictive current control methods for permanent magnet motors. To this end, the present invention provides a model predictive control method and system for a permanent magnet motor based on a triangular discrete space. This method divides the voltage vector plane based on the principle of equal division of an equilateral triangle and determines the midpoint of the equilateral triangle as a virtual voltage vector. Using the center of the equilateral triangle as the dimension, the voltage vector plane is evenly divided. This not only provides the system with a richer selection space for voltage vectors and makes the divided voltage vector plane more uniform and discrete, but also allows the voltage vector to be theoretically selected anywhere within the entire inverter output voltage vector plane (the hexagonal region, the feasible region). Furthermore, precisely because the voltage vectors are constructed entirely using the midpoints of the equilateral triangle and the iterative segmentation method is employed, the iterative segmentation at each level possesses unique geometric properties, laying the foundation for search optimization, namely, constructing the improved layer-by-layer narrowing optimization technique of the present invention and facilitating the calculation of the virtual voltage vector values ​​for the next level.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A model predictive control method for a permanent magnet motor based on a triangular discrete space comprises the following steps:

[0010] Construct a virtual voltage vector. Based on the two-level voltage vector plane structure of the inverter, perform iterative segmentation of the inverter's output voltage vector plane based on the equilateral triangle discrete space to generate a virtual voltage vector.

[0011] The iterative segmentation based on the equilateral triangle discrete space refers to segmenting the output voltage vector plane into six congruent equilateral triangles, and then iteratively segmenting the equilateral triangles as the segmentation object of the next layer, that is, each iteration splits the equilateral triangle into four equilateral triangles; the virtual voltage vector is a vector from the midpoint of the output voltage vector plane to the center of the equilateral triangle;

[0012] Filter the optimal voltage vector, use the virtual voltage vector to predict the current at the next moment, and then use the cost function based on the current prediction value to filter the optimal voltage vector from the constructed virtual voltage vectors with the goal of minimizing the cost function value;

[0013] Motor control generates bridge arm switching pulse signals for each phase based on the optimal voltage vector to control the motor.

[0014] Furthermore, the process of constructing the virtual voltage vector is:

[0015] First, the output voltage vector plane is divided into 6 congruent equilateral triangles, and the midpoint of each equilateral triangle is taken to construct a virtual voltage vector;

[0016] Each equilateral triangle is further divided into four small equilateral triangles, and the midpoint of each small equilateral triangle is taken to construct a virtual voltage vector;

[0017] Using this equal division rule, repeat the iterative division n times to generate a virtual voltage vector set, and obtain A virtual voltage vector.

[0018] Furthermore, the process of selecting the optimal voltage vector is as follows:

[0019] Construct a virtual voltage vector set, then introduce a comparative optimization strategy or adopt a traversal screening method to screen out the optimal voltage vector from the constructed virtual voltage vector set;

[0020] Alternatively, a comparative optimization strategy is introduced to screen the optimal voltage vector by combining iterative regional segmentation of the output voltage vector plane with optimization. That is, only the equilateral triangle where the optimal voltage vector after the previous layer of regional segmentation is located is segmented in the current layer based on the equilateral triangle discrete space to generate a new virtual voltage vector to participate in the optimization process of the optimal voltage vector in the current layer.

[0021] Furthermore, the comparative optimization strategy is:

[0022] 1) For the first layer of area segmentation, set the virtual voltage vector u(1,i) = 2U m / 3∠(π / 6+i*π / 3), where the equilateral triangle markers i=0, 1, 2, 3, 4, 5 represent the six equilateral triangles that divide the output voltage vector plane when the first layer of region segmentation is performed; then based on the cost function, the six virtual voltage vectors u (1,i) Determine the optimal voltage vector u in the current area with the minimum cost function g opt1 ;

[0023] in, , U m is the amplitude of the circle inscribed in the output voltage vector plane, u dc is the DC bus voltage, ∠ is the angle symbol;

[0024] 2) For the second layer of region segmentation, according to the previous layer u opt1 The equilateral triangle i is further divided into 4 sub-equilateral triangles, and the calculation of u is performed. opt1 The three new virtual voltage vectors u outside the sub-equilateral triangle (2,j) , then based on the cost function, determine the optimal voltage vector u of the current region with the minimum cost function g from the current virtual voltage vector opt2 ;

[0025] If i∈{0, 2, 4}, then set the direction identifier X2 to 1 and add a virtual voltage vector u (2,j) = u opt1 +U m / 3∠θ j If i∈{1, 3, 5}, set the direction identifier X2 to -1 and add a virtual voltage vector u (2,j) = u opt1 + U m / 3∠-θ j , angle θ j =-5π / 6+ (j-1)*2π / 3, and the new virtual voltage vector marker j=1, 2, 3; θ j is the angle between the newly added virtual voltage vector and the positive horizontal axis of the output voltage vector plane;

[0026] For example: m / 3∠-θ j is a loss, U m / 3 represents the amplitude, -θ j Indicates the specific angle, θ j =-5π / 6+ (j-1)*2π / 3 is the angle value.

[0027] 3) For the third layer of region segmentation, obtain the previous layer u opt2The equilateral triangle i is further divided into 4 sub-equilateral triangles, and the calculation of u is performed. opt2 The three new virtual voltage vectors u outside the sub-equilateral triangle (3,j) , then based on the cost function, determine the optimal voltage vector u of the current region with the minimum cost function g from the current virtual voltage vector opt3 ;

[0028] Among them, if u opt2 =u opt1 , direction identifier X3=-X2,u (3,j) = u opt2 + U m / (3*2 n-2 )∠θ j ,j=1, 2, 3;if u opt2 ≠ u opt1 , direction identifier X3=X2,u (3,j) = u opt2 + U m / (3*2 n-2 )∠-θ j ;

[0029] According to the third layer's regional segmentation and optimal voltage vector optimization rules, the subsequent layers' regional segmentation and optimal voltage vector update are iteratively performed until the nth layer outputs the optimal voltage vector u optn , as the final screened optimal voltage vector.

[0030] Furthermore, the number of virtual voltage vectors participating in the optimization calculation is 3n+3, where n is the number of segmentation layers of the output voltage vector plane that is iteratively segmented based on the equilateral triangle discrete space.

[0031] It should be understood that the larger the value of n, the smaller the control error, but the greater the amount of calculation, so a compromise needs to be considered. Its value is related to the following factors:

[0032] 1. Related to the computing power of digital signal processing chips;

[0033] 2. Related to search methods, algorithm redundancy, etc.;

[0034] 3. Related to the control target, if the requirement is high, the value of n will be larger, otherwise it will be smaller.

[0035] Therefore, the value of n needs to be set according to the actual situation and can be debugged through trial and error. For example, n=3 can meet most requirements.

[0036] Furthermore, the comparative optimization strategy is constructed based on the geometric regularity between the virtual voltage vector constructed by iterative segmentation at each layer and the optimal voltage vector;

[0037] The geometric rule is: when the actual optimal voltage vector is located within a certain equilateral triangle z of the current layer of area segmentation, for the virtual voltage vector generated after the current layer of area segmentation, the optimal voltage vector searched by the cost function is the virtual voltage vector formed by the midpoint of the equilateral triangle z.

[0038] The calculation formula for the predicted current at time k+1 is:

[0039]

[0040] Among them, u αβ =[u α , u β ] T is the output voltage vector of the inverter, u α ,u β are the α and β components of the output voltage vector, i αβ =[i α , i β ] T is the primary current vector, i α ,i β are the α and β components of the primary current vector, e αβ =[e α , e β ] T is the back electromotive force vector, e α , e β are the α and β components of the back electromotive force vector respectively; and there is e α = -ω r ψ f cosθ r , e β =ω r ψ f sinθ r ,ω r is the angular velocity; ψ f is the permanent magnet flux; θ r is the rotor angle of the motor, R s is the primary resistance, T s To control the cycle, is the stator inductance, k represents the time, T is the matrix transpose symbol, is the predicted current at time k+1, is the current sampling value at time k.

[0041] In addition, the technical solution of the present invention further provides a control system, which applies the above-mentioned permanent magnet motor model predictive control method, and the control system includes:

[0042] A virtual voltage vector construction module is used to iteratively segment the output voltage vector plane of the inverter based on the equilateral triangle discrete space according to the two-level voltage vector plane structure to generate a virtual voltage vector;

[0043] The iterative segmentation based on the equilateral triangle discrete space refers to segmenting the output voltage vector plane into six congruent equilateral triangles, and then iteratively segmenting the equilateral triangles as the segmentation object of the next layer, that is, each iteration splits the equilateral triangle into four equilateral triangles; the virtual voltage vector is a vector from the midpoint of the output voltage vector plane to the center of the equilateral triangle;

[0044] A current prediction module is used to predict the current at the next moment using a virtual voltage vector;

[0045] An optimal voltage vector screening module is used to screen out the optimal voltage vector from the constructed virtual voltage vectors using a cost function based on the current prediction value with the goal of minimizing the cost function value;

[0046] The motor control module is used to generate bridge arm switch pulse signals of each phase according to the optimal voltage vector to control the motor.

[0047] The technical solution of the present invention further provides a computer-readable storage medium storing a computer program, wherein the computer program is called by a processor to implement:

[0048] Steps of the permanent magnet motor model predictive control method based on triangular discrete space.

[0049] Compared with the prior art, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0050] (1) The voltage vector plane is a regular hexagon. The conventional discrete space vector method is to use the inscribed circle of the regular hexagon as the basis and evenly divide the voltage vector plane with the two dimensions of amplitude and phase angle, or to discretize on the midline of each triangle of the regular hexagon in CN116800144A and to generate virtual voltage vectors on the fixed hexagon outline and diagonal lines in CN116317795A.

[0051] The technical solution of the present invention adopts the principle based on discrete space vectors, that is, the voltage vector plane is evenly divided based on the principle of equal division of the equilateral triangle, and the midpoint of the equilateral triangle is used as the voltage vector. Compared with the traditional model prediction current control method, this design not only provides the system with a richer voltage vector selection space but also makes the divided voltage vector plane more uniform. Theoretically, its voltage vector can be selected to any position in the entire inverter output voltage vector plane, and the control error is small. In addition, it is precisely because all voltage vectors are constructed using the midpoints of the equilateral triangle and the iterative segmentation method is adopted that it has unique geometric characteristics in the iterative segmentation of each layer, laying the foundation for search optimization, that is, constructing the improved optimization technology of narrowing the range layer by layer of the present invention. It is theoretically proved that if the actual optimal voltage vector is within a certain equilateral triangle area, the virtual voltage vector formed based on the midpoint of the equilateral triangle must be the optimal voltage vector of the current area determined by the technical solution of the present invention, rather than other virtual voltage vectors.

[0052] (2) Based on the equilateral triangle discrete space constructed by the present invention and the construction of the virtual voltage vector with the center of the equilateral triangle, the screening method of the optimal voltage vector is further optimized. The technical solution of the present invention constructs a technical idea of ​​narrowing the range layer by layer by relying on the optimal voltage vector of the previous layer to narrow the optimization range of the next layer. By reconstructing the effective area of ​​the voltage vector, the repeated search of the voltage vector action area is avoided, and the search efficiency is high. Compared with other search methods, under the same computational burden, the present invention can obtain more voltage vectors, thereby improving the steady-state control performance of the permanent magnet motor. That is, the technical solution of the present invention constructs a search technology that narrows the range layer by layer based on the unique geometric characteristics of the virtual voltage vector constructed with the midpoint of the equilateral triangle, taking into account the accuracy of the optimal voltage vector and the screening efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention provides a flow chart of predictive control of a permanent magnet motor model based on a triangular discrete space.

[0054] Figure 2 This is a theoretical analysis of the optimal voltage vector search method provided by an embodiment of the present invention.

[0055] Figure 3 It is a decomposition diagram of the execution steps of the optimal voltage vector search method provided by an embodiment of the present invention.

[0056] Figure 4 This is an overall control block diagram of a permanent magnet motor model predictive control method based on triangular discrete space provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. The present invention will be further described below in conjunction with embodiments, taking a permanent magnet motor as an example, but the technical concept of the present invention is not limited to a type of permanent magnet motor.

[0058] One of the core points of this invention is the proposed discrete space method for the inverter output voltage vector plane, which discretizes the hexagonal plane layer by layer into several triangular regions, with the vector selected as the center of the triangle. Compared with other discretization methods, the discrete space constructed by the technical solution of the present invention is uniformly discrete. Therefore, the maximum error between the final selected voltage vector and the theoretical optimal voltage vector is fixed, and the control performance is relatively stable. The second core point is the efficient search method coordinated with the discrete space. By reconstructing the effective area of ​​the voltage vector, it avoids repeated searches of the voltage vector's action area, and has high search efficiency. Compared with other search methods, the technical solution of the present invention effectively reduces the computational burden.

[0059] The embodiment of the present invention provides a model predictive control method for a permanent magnet motor based on a triangular discrete space, which uses a current sensor and a speed sensor to sample the motor phase current and speed signals respectively. The embodiment of the present invention establishes a state equation with the primary current vector as the state variable:

[0060] (1)

[0061] Among them, u αβ =[u α , u β ] T is the output voltage vector of the inverter, u α ,u β are the α and β components of the output voltage vector, i αβ =[i α , i β ] T is the primary current vector, i α ,i β are the α and β components of the primary current vector, e αβ =[e α , e β ] T is the back electromotive force vector, e α , e β are the α and β components of the back electromotive force vector respectively; and there is e α = -ω r ψ fcosθ r , e β =ω r ψ f sinθ r ,ω r is the angular velocity; ψ f is the permanent magnet flux; θ r is the rotor angle of the motor, R s is the primary resistance, T s To control the cycle, is the stator inductance, k represents the time, and T is the matrix transpose symbol.

[0062] It should be understood that formula (1) is a current prediction model, which is used to predict the current value at the next moment and the output voltage vector u of the inverter required for the calculation process. αβ This is related to the voltage vector that needs to be optimized in the present invention. To this end, the goal of the optimization process is defined: the sum of the squares of the differences between the current reference values ​​and the predicted values ​​corresponding to the α and β directions is defined as the cost function g, and the voltage vector with the smallest cost function is selected as the optimal voltage vector and applied to the inverter at the next moment. In other feasible embodiments, defining other cost functions also meets the technical concept of the present invention and falls within the scope of protection of the present invention. The formula of the cost function g is:

[0063] (2)

[0064] Among them, g is the cost function value, and are the α and β components of the primary current reference.

[0065] Based on the above formula (1) and formula (2), the permanent magnet motor model predictive control method based on triangular discrete space provided by the embodiment of the present invention is as follows: Figure 1 as well as Figure 4 As shown, the following steps are included:

[0066] S1: Construct a virtual voltage vector, that is, iteratively divide the output voltage vector plane of the inverter to generate a total of 6×4 n-1 virtual voltage vectors, and n is the number of segmentation layers of the output voltage vector plane based on the iterative segmentation of the equilateral triangle discrete space.

[0067] In order to increase the number of voltage vectors available for selection, the technical solution of the present invention proposes a voltage vector discretization method based on an equilateral triangle. That is, by dividing the voltage vector plane into equal parts using the shape of an equilateral triangle, the selection space of the voltage vector is refined and a more uniform discrete space is constructed. The specific division method is as follows:

[0068] First, the entire regular hexagonal voltage vector plane is divided into 6 congruent equilateral triangles, and the midpoint of each equilateral triangle is taken as the voltage vector. Then, each equilateral triangle is further divided into 4 small equilateral triangles, and the midpoint of each small equilateral triangle is the voltage vector. Repeat this rule (equal division) for n times to complete the division of the regular hexagonal voltage vector. The total number of voltage vectors that can be obtained is:

[0069] (3)

[0070] Among them, n total is the total number of discrete voltage vectors, and n is the number of segmentation layers of the output voltage vector plane based on the iterative segmentation of the equilateral triangle discrete space.

[0071] S2: At the current moment k, the phase current and position information of the motor are collected in real time, and the speed signal is calculated using the position data. , and convert the phase current to the αβ coordinate system to obtain It should be understood that this step is a conventional technical means in this field and can be implemented in many ways. The present invention does not impose any specific limitations or restrictions on this.

[0072] S3: Perform current prediction based on the d-axis current reference value, the q-axis current reference value, the angular velocity, and the virtual voltage vector to obtain the predicted current at time k+1; select the optimal voltage vector with the minimum cost function value based on the predicted current.

[0073] In some embodiments, step S1 generates The virtual voltage vector set is constructed by virtual voltage vectors, and step S3 can use a traversal method to find the optimal voltage vector. However, this embodiment preferably introduces the following comparison optimization strategy to select the optimal voltage vector from the virtual voltage vector set.

[0074] S4: Generate bridge arm switch pulse signals of each phase according to the optimal voltage vector to realize the control of the permanent magnet motor.

[0075] About comparative optimization strategy:

[0076] In traditional model-based predictive current control, directly applying the numerous discretely generated voltage vectors sequentially to the current prediction model and cost function for rolling optimization results in a dramatic increase in computational complexity, severely impacting the real-time performance and efficiency of control. To address this issue, the present invention designs an efficient optimal voltage vector search method. Figure 2 The principle of this search method for searching the optimal voltage vector in the discrete voltage vector plane is described in detail:

[0077] Based on the equilateral triangle discrete space proposed in this invention, u optIs the optimal voltage vector to be found. If the optimal voltage vector u opt is u OK ,correspond Figure 2 K point in the, then u OA 、u OB 、u OC and u OD These are the four voltage vectors that need to be evaluated online. OB If it is the optimal voltage vector found, then its cost function value should be the smallest, that is, |KB| should be less than |KA|, |KC|, and |KD|. However, in the right triangle RTΔANK and RTΔBNK, it is easy to find that |AN|<|BN|. Therefore, it can be deduced that |AN|²+|NK|²<|BN|²+|NK|², that is, |KA|<|KB|. This means that u OB cannot be the optimal voltage vector. In other words, the optimal voltage vector must be u OA , that is, if the optimal voltage vector u opt Located in u OA The optimal voltage vector must be u OA After several iterations, the optimal voltage vector can be found more accurately.

[0078] like Figure 3 As shown in the figure, the specific implementation process of comparison optimization is as follows:

[0079] 1) For the first layer of area segmentation, set the virtual voltage vector u (1,i) = 2U m / 3∠(π / 6+i*π / 3), where the equilateral triangle markers i=0, 1, 2, 3, 4, 5 represent the six equilateral triangles that divide the output voltage vector plane when the first layer or the previous layer is divided; then based on the cost function, the six virtual voltage vectors u (1,i) Determine the optimal voltage vector u in the current area with the minimum cost function g opt1 , , U m is the amplitude of the circle inscribed in the output voltage vector plane, u dc is the DC bus voltage;

[0080] 2) For the second layer of region segmentation, obtain the previous layer u opt1 The three new virtual voltage vectors u corresponding to the four equilateral triangles divided by the equilateral triangle i (2,j) , then based on the cost function, determine the optimal voltage vector u of the current region with the minimum cost function g from the current virtual voltage vector opt2 ;

[0081] If i∈{0, 2, 4}, then set the direction identifier X2 to 1 and add a virtual voltage vector u (2,j) = u opt1 +U m / 3∠θ j If i∈{1, 3, 5}, set the direction identifier X2 to -1 and add a virtual voltage vector u (2,j) = u opt1 + U m / 3∠-θ j , where θ j =-5π / 6+ (j-1)*2π / 3, and the new virtual voltage vector marker j=1, 2, 3; θ j is the angle between the newly added virtual voltage vector and the positive horizontal axis of the output voltage vector plane;

[0082] 3) For the third layer of region segmentation, obtain the previous layer u opt2 The three new virtual voltage vectors u corresponding to the four equilateral triangles divided by the equilateral triangle i (3,j) , then based on the cost function, determine the optimal voltage vector u of the current region with the minimum cost function g from the current virtual voltage vector opt3 ;

[0083] Among them, if u opt2 =u opt1 , identifier X3=-X2,u (3,j) = u opt2 + U m / (3*2 n-2 )∠θ j ,j=1, 2, 3;if u opt2 ≠ u opt1 , identifier X3=X2,u (3,j) = u opt2 + U m / (3*2 n-2 )∠-θ j ;

[0084] According to the third-layer region segmentation and optimal voltage vector optimization rules, the region segmentation of subsequent layers and the update of the optimal voltage vector are iteratively performed until the nth layer, as follows:

[0085] n) if u optn-2 = u optn-1 , identifier X n =-X n-1 In this case, u (n,j) = u optn-1 + U m / (3*2 n-2 )∠θj , and substitute these three new voltage vectors into formulas (1)-(2) to obtain the optimal voltage vector u optn . If u optn-2 ≠u optn-1 , then the identifier X n =X n-1 In this case, u (n,j) = u optn-1 + U m / (3*2 n-2 )∠-θ j , and substitute these three new voltage vectors into formulas (1)-(2) to obtain the optimal voltage vector u optn .

[0086] Through the above method, the present invention can quickly and accurately find the voltage vector that is closest to the actual optimum without traversing all voltage vectors. This not only greatly reduces the amount of calculation and improves the real-time performance of the control, but also ensures that the current prediction model and cost function can more effectively guide the control process of the motor, thereby achieving more accurate and stable current control and thrust output. The design of this optimal voltage vector search method is one of the important improvements of the present invention to the model prediction current control strategy. For example, after n iterative searches in this embodiment, the number of virtual voltage vectors evaluated online is only 3n+3, which greatly reduces the number of voltage vectors that need to be evaluated compared to the conventional method (i.e., calculating all possible voltage vectors). Finally, when the optimal voltage vector is found, it will be immediately sent to the PWM modulator to generate the corresponding PWM signal. After these n times of careful searches, the selected optimal voltage vector u optn The actual optimal voltage vector u opt Very close.

[0087] Secondly, in other feasible embodiments, the optimal voltage vector can be screened by combining iterative regional segmentation of the output voltage vector plane with optimization, that is, only the equilateral triangle where the optimal voltage vector after the previous layer of regional segmentation is located is segmented, and the current layer region segmentation based on the equilateral triangle discrete space is performed to generate a new virtual voltage vector to participate in the optimal voltage vector optimization process of the current layer region.

[0088] To verify the feasibility of the above method, the present invention is compared with existing discrete space segmentation methods based on the vertices of an equilateral triangle. Specifically, method 1 is the non-optimal model predictive torque control technology based on discrete space vector modulation proposed by I. Osman et al., and method 2 is the low switching frequency model predictive flux control (for induction motor drive) technology based on discrete space vector modulation proposed by I. Osman et al. This proves that the proposed optimal voltage vector VV search method is superior to the traditional method. The results shown in Table 1 include N fA value indicating the number of VVs that need to be evaluated.

[0089] Table 1 Comparison results

[0090] .

[0091] As shown in Table 1, when the number of VVs N evaluated online f When the number of generated VVs is similar, the proposed method generates more VVs. This shows that when the computational burden is the same, the proposed method achieves higher accuracy. Thereafter, the current ripple and thrust fluctuation of the permanent magnet motor drive system can be effectively suppressed. When the number of generated VVs n total When the control accuracy is the same, the proposed method has a smaller number of VVs to be evaluated online. This indicates that the proposed method achieves less computational burden when the control accuracy is the same.

[0092] It will be easily understood by those skilled in the art that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0093] In some embodiments, the technical solution of the present invention also provides a control system based on a permanent magnet motor model predictive control method, including: a virtual voltage vector construction module, a current prediction module, an optimal voltage vector screening module and a motor control module.

[0094] The virtual voltage vector construction module is used to iteratively segment the output voltage vector plane of the inverter based on the equilateral triangle discrete space to generate a virtual voltage vector; wherein, the iterative segmentation based on the equilateral triangle discrete space refers to segmenting the output voltage vector plane into 6 congruent equilateral triangles, and then iteratively segmenting with the equilateral triangle as the segmentation object of the next layer, that is, each layer of iteration segments the equilateral triangle into 4 equilateral triangles; the virtual voltage vector is the vector from the midpoint of the output voltage vector plane to the center of the equilateral triangle.

[0095] The current prediction module is used to use the virtual voltage vector to predict the current at the next moment. The optimal voltage vector screening module is used to use a cost function based on the current prediction value to screen the optimal voltage vector from the constructed virtual voltage vector with the goal of minimizing the cost function value. The motor control module is used to generate the bridge arm switching pulse signal of each phase according to the optimal voltage vector to control the motor.

[0096] It should also be understood that the specific implementation process of each module please refer to the above method content, the present invention will not go into details here, and the division of the above functional modules is only for illustration. In some embodiments, some functional modules can be merged, and some functional modules can be split. Each functional module can be implemented in software or hardware or a combination of software and hardware. Among them, the hardware and software devices include but are not limited to general-purpose computer devices, programmable gate arrays, digital signal processors, microprocessors and their corresponding programming or burning software.

[0097] In some embodiments, the technical solution of the present invention further provides a computer-readable storage medium storing a computer program, which is called by a processor to implement the steps of a permanent magnet motor model predictive control method based on a triangular discrete space. Specific implementation:

[0098] Constructing a virtual voltage vector, performing iterative segmentation of the inverter output voltage vector plane based on the equilateral triangle discrete space to generate a virtual voltage vector;

[0099] The iterative segmentation based on the equilateral triangle discrete space refers to dividing the output voltage vector plane into six congruent equilateral triangles, and then iteratively segmenting the equilateral triangles as the segmentation object of the next layer, that is, each iteration divides the equilateral triangle into four equilateral triangles. The virtual voltage vector is the vector from the midpoint of the output voltage vector plane to the center of the equilateral triangle.

[0100] Filter the optimal voltage vector, use the virtual voltage vector to predict the current at the next moment, and then use the cost function based on the current prediction value to filter the optimal voltage vector from the constructed virtual voltage vectors with the goal of minimizing the cost function value;

[0101] Motor control generates bridge arm switching pulse signals for each phase based on the optimal voltage vector to control the motor.

[0102] For the specific implementation process of each step, please refer to the description of the above method.

[0103] The readable storage medium is a computer-readable storage medium, which can be an internal storage unit of the software and hardware device described in any of the aforementioned embodiments, such as a hard disk or memory of a controller. The readable storage medium can also be an external storage device of the controller, such as a plug-in hard disk equipped on the controller, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Furthermore, the readable storage medium can also include both an internal storage unit of the controller and an external storage device. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium can also be used to temporarily store data that has been output or is to be output.

[0104] Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for causing a computer device (such as a personal computer, server, or network device) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned readable storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0105] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is a flow chart according to the method, device (system), and computer program product of the embodiment of the present application and / or the instructions executed by the processor to generate a device for realizing the function specified in one flow chart or multiple flows and / or one box or multiple boxes of the block diagram. These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a product comprising an instruction device, which realizes the function specified in one flow chart or multiple flows and / or one box or multiple boxes of the block diagram. These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0106] It should be emphasized that the examples described in the present invention are illustrative rather than restrictive. Therefore, the present invention is not limited to the examples described in the specific embodiments. Any other embodiments derived by those skilled in the art based on the technical solution of the present invention that do not depart from the purpose and scope of the present invention, whether modified or replaced, also fall within the scope of protection of the present invention.

Claims

1. A model predictive control method for a permanent magnet motor based on a triangular discrete space, characterized by: include: Construct a virtual voltage vector. Based on the two-level voltage vector plane structure of the inverter, perform iterative segmentation of the inverter's output voltage vector plane based on the equilateral triangle discrete space to generate a virtual voltage vector. The iterative segmentation based on the equilateral triangle discrete space refers to segmenting the output voltage vector plane into six congruent equilateral triangles, and then iteratively segmenting the equilateral triangles as the segmentation object of the next layer, that is, each iteration splits the equilateral triangle into four equilateral triangles; the virtual voltage vector is a vector from the midpoint of the output voltage vector plane to the center of the equilateral triangle; Filter the optimal voltage vector, use the virtual voltage vector to predict the current at the next moment, and then use the cost function based on the current prediction value to filter the optimal voltage vector from the constructed virtual voltage vectors with the goal of minimizing the cost function value; The calculation formula for the predicted current at time k+1 is: ； Among them, u αβ =[u α , u β ] T is the output voltage vector of the inverter, u α ,u β are the α and β components of the output voltage vector, i αβ =[i α , i β ] T is the primary current vector, i α ,i β are the α and β components of the primary current vector, e αβ =[e α , e β ] T is the back electromotive force vector, e α , e β are the α and β components of the back electromotive force vector respectively; and there is e α = -ω r ψ f cosθ r , e β =ω r ψ f sinθ r ,ω r is the angular velocity; ψ f is the permanent magnet flux; θ r is the rotor angle of the motor, R s is the primary resistance, T s To control the cycle, is the stator inductance, k represents the time, T is the matrix transpose symbol, is the predicted current at time k+1, is the current sampling value at time k; The formula of the cost function g is: ； Among them, g is the cost function value, and are the α and β components of the primary current reference; Motor control generates bridge arm switching pulse signals for each phase based on the optimal voltage vector to control the motor.

2. The permanent magnet motor model predictive control method according to claim 1, characterized in that: The process of constructing a virtual voltage vector is: First, the output voltage vector plane is divided into 6 congruent equilateral triangles, and the midpoint of each equilateral triangle is taken to construct a virtual voltage vector; Each equilateral triangle is further divided into four small equilateral triangles, and the midpoint of each small equilateral triangle is taken to construct a virtual voltage vector; Using this equal division rule, repeat the iterative division n times to generate a virtual voltage vector set, and obtain A virtual voltage vector.

3. The permanent magnet motor model predictive control method according to claim 1, characterized in that: The process of selecting the optimal voltage vector is: Construct a virtual voltage vector set, then introduce a comparative optimization strategy or adopt a traversal screening method to screen out the optimal voltage vector from the constructed virtual voltage vector set; Alternatively, a comparative optimization strategy is introduced to screen the optimal voltage vector by combining iterative regional segmentation of the output voltage vector plane with optimization. That is, only the equilateral triangle where the optimal voltage vector after the previous layer of regional segmentation is located is segmented in the current layer based on the equilateral triangle discrete space to generate a new virtual voltage vector to participate in the optimization process of the optimal voltage vector in the current layer.

4. The permanent magnet motor model predictive control method according to claim 3, characterized in that: The comparative optimization strategy is: 1) For the first layer of area segmentation, set the virtual voltage vector u (1,i) = 2U m / 3∠(π / 6+i*π / 3), where the equilateral triangle markers i=0, 1, 2, 3, 4, 5 represent the six equilateral triangles that divide the output voltage vector plane when the first layer of region segmentation is performed; then based on the cost function, the six virtual voltage vectors u (1,i) Determine the optimal voltage vector u in the current area with the minimum cost function g opt1 ; in, ,U m is the amplitude of the circle inscribed in the output voltage vector plane, u dc is the DC bus voltage, ∠ is the angle symbol; 2) For the second layer of region segmentation, according to the previous layer u opt1 The equilateral triangle i is further divided into 4 sub-equilateral triangles, and the calculation of u is performed. opt1 The three new virtual voltage vectors u outside the sub-equilateral triangle (2,j) , then based on the cost function, determine the optimal voltage vector u of the current region with the minimum cost function g from the current virtual voltage vector opt2 ; If i∈{0, 2, 4}, then set the direction identifier X2 to 1 and add a virtual voltage vector u (2,j) = u opt1 + U m / 3∠θ j If i∈{1, 3, 5}, set the direction identifier X2 to -1 and add a virtual voltage vector u (2,j) = u opt1 + U m / 3∠-θ j , angle θ j =-5π / 6+ (j-1)*2π / 3, and the new virtual voltage vector marker j=1, 2, 3,θ j is the angle between the newly added virtual voltage vector and the positive horizontal axis of the output voltage vector plane; 3) For the third layer of region segmentation, obtain the previous layer u opt2 The equilateral triangle i is further divided into 4 sub-equilateral triangles, and the calculation of u is performed. opt2 The three new virtual voltage vectors u outside the sub-equilateral triangle (3,j) , then based on the cost function, determine the optimal voltage vector u of the current region with the minimum cost function g from the current virtual voltage vector opt3 ; Among them, if u opt2 =u opt1 , direction identifier X3=-X2,u (3,j) = u opt2 + U m / (3*2 n-2 )∠θ j ,j=1, 2, 3;if u opt2 ≠ u opt1 , direction identifier X3=X2,u (3,j) = u opt2 + U m / (3*2 n-2 )∠-θ j ; According to the third layer's regional segmentation and optimal voltage vector optimization rules, the subsequent layers' regional segmentation and optimal voltage vector update are iteratively performed until the nth layer outputs the optimal voltage vector u optn , as the final screened optimal voltage vector.

5. The permanent magnet motor model predictive control method according to claim 4, characterized in that: The number of virtual voltage vectors involved in the optimization calculation is 3n+3, where n is the number of segmentation layers of the output voltage vector plane based on the iterative segmentation of the equilateral triangle discrete space.

6. The permanent magnet motor model predictive control method according to claim 4, characterized in that: The comparative optimization strategy is constructed based on the geometric law between the virtual voltage vector constructed by iterative segmentation at each layer and the optimal voltage vector; The geometric rule is: when the actual optimal voltage vector is located within a certain equilateral triangle z of the current layer of area segmentation, for the virtual voltage vector generated after the current layer of area segmentation, the optimal voltage vector searched by the cost function is the virtual voltage vector formed by the midpoint of the equilateral triangle z.

7. A control system, applying the permanent magnet motor model predictive control method according to any one of claims 1 to 6, characterized in that: The control system includes: A virtual voltage vector construction module is used to iteratively segment the output voltage vector plane of the inverter based on the equilateral triangle discrete space according to the two-level voltage vector plane structure to generate a virtual voltage vector; The iterative segmentation based on the equilateral triangle discrete space refers to segmenting the output voltage vector plane into six congruent equilateral triangles, and then iteratively segmenting the equilateral triangles as the segmentation object of the next layer, that is, each iteration splits the equilateral triangle into four equilateral triangles; the virtual voltage vector is a vector from the midpoint of the output voltage vector plane to the center of the equilateral triangle; A current prediction module is used to predict the current at the next moment using a virtual voltage vector; An optimal voltage vector screening module is used to screen out the optimal voltage vector from the constructed virtual voltage vectors using a cost function based on the current prediction value with the goal of minimizing the cost function value; The motor control module is used to generate bridge arm switch pulse signals of each phase according to the optimal voltage vector to control the motor.

8. A computer-readable storage medium, characterized in that: A computer program is stored, which is called by a processor to implement: The steps of the permanent magnet motor model predictive control method according to any one of claims 1 to 6.

Images