Permanent magnet synchronous motor phase current reconstruction method based on discrete vector modulation model prediction

Abstract

The invention discloses a permanent magnet synchronous motor phase current reconstruction method based on discrete vector modulation model prediction. The method comprises the following steps: firstly, analyzing and designing a candidate voltage vector supplement scheme based on a control period equal division principle, designing a voltage vector optimization scheme in combination with a dead-beat model prediction control method, and giving a vector synthesis sequence improvement scheme based on a fixed sampling moment principle; and the whole permanent magnet synchronous motor phase current reconstruction system based on discrete vector modulation model prediction is completed. According to the invention, the cost and the size of the control system are effectively reduced, the bandwidth of a control closed loop is improved, the current reconstruction error caused by a single current sensor is reduced, and the stability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motor drive control. Specifically, it relates to a method for reconstructing the phase current of a permanent magnet synchronous motor based on a discrete vector modulation model prediction. Background Art

[0002] Permanent Magnet Synchronous Motor (PMSM) has the advantages of high power density, high efficiency, and small volume due to the adoption of high-performance rare earth permanent magnet materials. It has been widely used in industrial fields such as aerospace and new energy electric vehicles. The simplification of its system structure and the improvement of control strategies have been the research hotspots in recent years.

[0003] Regarding the control system architecture of PMSM, traditional three-phase current information usually needs to be detected by configuring two or more current sensors, which leads to an increase in system cost and complexity of the structure. Adopting a single current sensor control scheme can significantly reduce the hardware cost and volume, and at the same time avoid the synchronization deviation problem caused by multiple sensors. However, as high-precision current sampling is a prerequisite for high-performance control, the single current sensor control scheme faces two major challenges while simplifying the system: First, the single current sensor control can only collect current signals during the active vector action period. If the active vector action time is shorter than the sampling window requirement, it will lead to current sampling failure (reconstruction dead zone problem); Second, within the same control cycle, it is necessary to collect two-phase current information in different active vector periods respectively, while the ideal three-phase current information should be collected at the same moment. Time-sharing sampling will inevitably lead to errors in the three-phase current reconstruction information (time-sharing sampling error problem).

[0004] Model Predictive Control (MPC) has been widely used due to its good dynamic response, easy multi-variable control, and easy handling of non-linear constraints. However, traditional MPC only acts on one voltage vector within a control cycle, which cannot ensure complete voltage tracking, resulting in an increase in current ripple. The prediction accuracy of MPC depends on the reconstruction accuracy of the current and the accuracy of the prediction model. The reconstruction accuracy of the three-phase current depends on the successful sampling of the current and the precise compensation of the reconstruction error. If traditional MPC is directly adopted, it will be difficult to obtain a high-performance PMSM single current control system. Therefore, solving the key problems of single current control under MPC and improving the MPC prediction accuracy and control robustness while improving the three-phase current reconstruction accuracy is of great significance for realizing a PMSM system with high dynamic response under low complexity. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method for reconstructing the phase current of a permanent magnet synchronous motor based on discrete vector modulation model prediction, which can effectively improve the bandwidth of the control system, simplify the hardware circuit of the permanent magnet synchronous motor control system, reduce the cost of the entire control system, and enhance the control fault tolerance ability of the entire system.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention proposes a method for reconstructing the phase current of a permanent magnet synchronous motor based on discrete vector modulation model prediction, and the method includes the following steps:

[0008] S1: Analyze the candidate voltage vector supplement scheme based on the principle of equal division of the control period.

[0009] S2: According to the candidate voltage supplement scheme in step S1, combine the deadbeat model predictive control to design a voltage vector optimization scheme.

[0010] S3: Analyze and design an improved scheme for the vector synthesis sequence based on the principle of fixed sampling time.

[0011] Further, in step S1, design the candidate voltage vector supplement scheme based on the principle of equal division of the control period:

[0012] The design of the candidate voltage vector includes the following steps:

[0013] First, in order to improve the accuracy of the model predictive current control, consider expanding the candidate vectors. The common methods include "double vector synthesis", "bisecting the phase angle in the hexagonal sector", "bisecting according to the amplitude and phase angle", and "bisecting the control period". Considering the improvement of the current prediction accuracy and the limitation of the minimum sampling time based on the single current sensor control, the present invention adopts the method of bisecting the period. The expanded vectors are essentially discrete virtual voltage vectors (DVVs). The number of DVV is closely related to the number of equal divisions of the control period. Define the number of equal divisions of the control period as N, then the total number of newly added DVV is

[0014] N V = 3N 2 + 3N - 6

[0015] The larger the number of equal divisions N of the control period, the more candidate vectors can be obtained, which can effectively reduce the error between the reference voltage vector and the actual synthesized vector. However, if N is too large, it will also pose a test to the computational performance of the control system. Therefore, in the present invention, the number of equal divisions of the control period is set to 4.

[0016] Further, in step S2, combine the model predictive control to design a voltage vector optimization scheme:

[0017] The voltage vector optimization scheme includes the following steps:

[0018] First, for the permanent magnet synchronous motor control system, MPC generally selects the state space equation according to the relationship of each variable to design the prediction model of the discrete-time system, which is expressed as follows:

[0019]

[0020] According to the discrete state model and the forward Euler method, the following can be obtained:

[0021]

[0022] Where T sp is the switching period time, k and k + 1 respectively represent the relevant measurement values of the kth switching period and the predicted values of the (k + 1)th switching period. and are the measured dq-axis current values, ω e is the motor speed feedback value, and are the dq-axis voltages applied to the motor in the kth switching period, which are the input quantities of the controlled system. Among them, the reference currents and are used to replace the predicted currents and of the (k + 1)th switching period, and the prediction model is obtained as:

[0023]

[0024] Where and are the decomposition values of the reference voltage vector on the direct and quadrature axes in the rotating coordinate system.

[0025] Secondly, taking the first sector as an example, when N = 4, the number of small sectors is 16. Each small sector is numbered, and the numbering naming rule is as follows: First, define l 1 , l 2 , l 3 as three reference edges, which coincide with the three sides of the sector I triangle respectively. Then, the reference voltage calculated by the model prediction is decomposed into and on the αβ axes. Then The distances from the vector end point to the reference edges l 1 , l 2 , l 3 can be calculated respectively as:

[0026]

[0027] Subsequently, divide d 1 , d 2 , d 3 by respectively and take the integer to obtain h 1 , h 2 , h 3 , which can be expressed as follows:

[0028]

[0029] where floor() is to take the integer downward. Then the small sector number can be calculated as h = 100h 1 + 10h 2 + h 3 , and the corresponding number of each small sector can be obtained accordingly. The three vertices of each small sector triangle respectively correspond to three candidate vectors, and each candidate vector is represented by the serial numbers of four voltage vectors. When N = 4, the corresponding relationship between different small sectors and the serial numbers of candidate vectors is shown in Table 1.

[0030] Table 1 Corresponding candidate vector serial numbers for small sectors

[0031]

[0032] After determining the small sector where is located, only need to find the minimum distance between and the three discrete vectors in the corresponding small sector to obtain the optimal vector Compared with the traditional method that needs to traverse all eight vectors, the calculation burden is effectively reduced.

[0033] It should be noted that under conditions such as sudden load or sudden acceleration of the motor, the reference voltage vector may fall into the overmodulation region outside the hexagonal sector. Considering that the voltage vector with the smallest amplitude error from these vectors is at the sector boundary, the reference vector needs to be overmodulated, scaled down proportionally to the linear hexagonal region, and then the candidate vectors are screened.

[0034] Furthermore, in step S3, an improved scheme for the vector synthesis sequence is designed based on the principle of fixed sampling time:

[0035] The improved steps of the vector synthesis sequence are as follows:

[0036] First, in traditional single-vector model predictive control, only one voltage vector acts in each period, so only one-phase current can be sampled. DSVM-MPC divides the voltage vectors evenly in a period, so a period can contain multi-phase current information for the reconstruction of three-phase currents. Meanwhile, by adjusting the number of period divisions to control the action time of a single voltage vector in a period, it can ensure that the current sampling meets the minimum sampling time limit to avoid the problem of reconstruction dead zone. Based on the above characteristics, it is considered to combine DSVM-MPC with single-current control to improve the accuracy of current reconstruction.

[0037] Among them, there are mainly two types of current reconstruction errors. One is the error between the sampled current i a_ t s1 of phase A and the current i a_ T sp at the end of the period. The other is the error between the sampled current i c_ t s2 of phase C and the current i c_ T sp at the end of the period. Current reconstruction needs to calculate and compensate both errors simultaneously, resulting in a large computational burden. It is considered to reduce the current error to one type by improving the vector synthesis sequence to achieve one-step compensation of the error.

[0038] By adjusting T sp and the number of period divisions N to control the action time of each voltage vector to be greater than the minimum sampling time T min , it can ensure successful current sampling during the vector action, that is, the problem of current reconstruction dead zone can be avoided.

[0039] In addition, if the candidate vectors contain zero vectors and only one type of active voltage vector, such as "1ZZZ" or "ZZZZ", only one-phase current information can still be sampled in a period and two-phase currents cannot be obtained. Since the zero vector can be composed of the synthesis of two complementary non-zero vectors, therefore, it is considered to use the "zero-vector replacement principle" to solve this problem. Fix the sampling moments as t s1 and t s2 , and by using complementary vectors to replace the zero vector and changing the vector synthesis sequence simultaneously, current reconstruction at the fixed sampling moments can be achieved.

[0040] In addition, for the candidate vectors "1111" and "2222", since they only contain one type of active vector and no zero vectors, the zero-vector replacement method cannot be used. It is considered to use the "measurement vector insertion method" to ensure that two-phase current information can be obtained in a period.

[0041] As can be seen from the above analysis, by adopting the "zero vector replacement" or "measurement vector insertion" method for different types of candidate vectors and combining with the adjustment of the vector synthesis sequence, the fixed-time sampling of current can be achieved under the condition of meeting the minimum sampling time. Different candidate vectors adopt different vector synthesis sequence improvement methods. In Sector I, the improved sequence is adopted to replace the original sequence, as shown in Table 2, where the two vector numbers within || respectively represent the two vectors acting within T sp / 4 time.

[0042] Table 2 Candidate vector sequence in Sector I

[0043]

[0044] Each sector has two adjacent vectors. For the convenience of representation, R represents the first vector in the counterclockwise direction of the sector, and L represents the second vector. The complementary vectors corresponding to these two vectors are represented as H and M respectively. The vector numbers corresponding to the sectors are shown in Table 3.

[0045] Table 3 Vector numbers corresponding to sectors

[0046]

[0047] Combining Table 2 and Table 3, the improved sequence of candidate vectors in the entire sector is shown in Table 4.

[0048] Table 4 Candidate vector synthesis sequence

[0049]

[0050] Then, a permanent magnet synchronous motor phase current reconstruction control system based on discrete vector modulation model prediction can be constructed.

[0051] The beneficial effects of the present invention are as follows:

[0052] 1. The discrete vector model predictive control proposed by the present invention simplifies the structure of the system and improves the dynamic performance of the system.

[0053] 2. The present invention proposes to equally divide the control period, expand the number of candidate vectors, reduce the error between the actual voltage and the reference voltage, improve the current reconstruction accuracy, and reduce the system fluctuation caused by the reconstructed phase current error.

[0054] 3. The voltage vector optimization scheme based on deadbeat model prediction proposed by the present invention realizes the screening of the optimal vector by judging the position of the reference voltage vector in the small sector and combining with the cost function.

[0055] 4. The improved scheme of the vector synthesis sequence proposed by the present invention ensures the acquisition of two-phase current information within the same period, realizes the current reconstruction at a fixed sampling moment, and simplifies the computational complexity of phase current reconstruction. Description of the Drawings

[0056] Figure 1 It is a flowchart of a method for reconstructing the phase current of a permanent magnet synchronous motor based on discrete vector modulation model prediction according to an embodiment of the present invention.

[0057] Figure 2 It is a structural diagram of a method for reconstructing the phase current of a permanent magnet synchronous motor based on discrete vector modulation model prediction according to an embodiment of the present invention.

[0058] Figure 3 It is a schematic diagram of a candidate vector expansion scheme according to an embodiment of the present invention.

[0059] Figure 4 It is a schematic diagram of the principle of the voltage vector optimization method according to an embodiment of the present invention.

[0060] Figure 5 It is a schematic diagram of the vector synthesis sequence of "12ZZ" according to an embodiment of the present invention.

[0061] Figure 6 It is a schematic diagram of the improved vector synthesis sequence of "12ZZ" according to an embodiment of the present invention.

[0062] Figure 7 It is a schematic diagram of the improved vector synthesis sequences of "1ZZZ" and "111Z" according to an embodiment of the present invention.

[0063] Figure 8 It is a schematic diagram of the improved vector synthesis sequence of "1111" according to an embodiment of the present invention.

[0064] Figure 9 It is a schematic diagram of the reconstructed phase current effect of the method for reconstructing the phase current of a permanent magnet synchronous motor using discrete vector modulation model prediction according to an embodiment of the present invention.

[0065] Figure 10 It is a schematic diagram of the d-q axis current effect of the method for reconstructing the phase current of a permanent magnet synchronous motor using discrete vector modulation model prediction according to an embodiment of the present invention. Specific Implementation Modes

[0066] The present invention will be further described in detail below with reference to the drawings and specific implementation modes.

[0067] Embodiment 1

[0068] Figure 1 It is a flowchart of a method for reconstructing the phase current of a permanent magnet synchronous motor based on discrete vector modulation model prediction. Figure 2It is the structure diagram of the permanent magnet synchronous motor phase current reconstruction method based on discrete vector modulation model prediction in the embodiment of the present invention. This embodiment proposes a permanent magnet synchronous motor phase current reconstruction method based on discrete vector modulation model prediction, and the method includes the following steps:

[0069] S1: Analyze the candidate voltage vector supplement scheme based on the principle of equal division of the control period.

[0070] S2: According to the candidate voltage supplement scheme in step S1, combine the deadbeat model predictive control to design a voltage vector optimization scheme.

[0071] S3: Analyze and design an improved scheme for the vector synthesis sequence based on the principle of fixed sampling time.

[0072] S4: Propose a permanent magnet synchronous motor phase current reconstruction method based on discrete vector modulation model prediction.

[0073] I. Candidate voltage vector supplement method

[0074] As Figure 3 shown, (a) is the double-vector synthesis scheme. By dividing the hexagonal sector into regions and adopting the method of synthesizing two vectors in different regions, the range of vector selection is expanded; (b) is the method of bisecting the phase angle of the sector, trisecting the phase angle to obtain additional sector boundary vectors as candidate vectors, as shown by the candidate vector V s composed of V 1 and 0.75V 3 synthesized together; (c) on the basis of bisecting the phase angle, bisect the vector amplitude, and expand the candidate vector as shown by the black dots in the figure; (d) the scheme bisects the time of each period and expands the candidate vector as shown by the blue dots in the figure.

[0075] As Figure 3 (d) shown, the control period is equally divided into 4 parts, and 54 new DVV are added, greatly expanding the vector selection range. The increase in the number of optional vectors effectively reduces the error between the reference voltage vector and the actual synthesized vector.

[0076] II. Voltage vector optimization method based on deadbeat model predictive control

[0077] Taking sector I as an example, the candidate vectors in the figure are synthesized by V 1 (100) and V 2 (100), and sector I can be divided into multiple small triangular sectors, and the number of small sectors is

[0078] N s = N 2

[0079] As Figure 4 shown, falls within the blue shaded area in the figure. Below the αβ axis, there is According to the small sector judgment method described above, substituting N = 4, h can be calculated as 1 = 2, h 2 = 1, h 3 = 3. Then the number of the small sector is h = 213.

[0080] By judging the number of the small sector, the three candidate vectors corresponding to the small sector can also be determined. According to Table 1, they are "12ZZ", "22ZZ", and "122Z". Substituting them into the cost formula for calculation, the optimal voltage vector with the smallest error between can be screened out The serial number is "12ZZ", and the synthesis sequence of the vector is as Figure 5 shown.

[0081] III. Method for improving the vector synthesis sequence based on the principle of fixed sampling time

[0082] As Figure 6 shown, place the V 1 , V 2 vectors in the synthesis sequence of "12ZZ" in the second half cycle, and place the sampling time at the end of the action time of the two active vectors V 1 , V 2 , where t s1 and t s2 . Among them, t s1 = 3T sp / 4, t s2 = T sp . Then the error compensation only needs to consider a reconstruction error between i a_ t s1 and i a_ T sp . To a certain extent, the calculation burden is reduced, and the improved sequence number is "ZZ12".

[0083] By adjusting T sp and the number of cycle equal division N to control the action time of each voltage vector to be greater than the minimum sampling time T min , it can ensure successful current acquisition during the vector action, that is, the current reconstruction dead zone problem can be avoided.

[0084] For the case where the candidate vector contains a zero vector and only one active voltage vector, such as "1ZZZ" or "ZZZZ", only one-phase current information can still be acquired within one cycle and two-phase current cannot be obtained. It is solved by using the "zero vector replacement principle". As Figure 7 shown, in (a), taking "1ZZZ" as an example, replace the two zero vectors with complementary vectors V 5and V 2 Instead, the two vectors act respectively for T sp / 4 of the time, and place V 2 at the end of the vector synthesis sequence, then the vector sequence number can be changed to "Z512"; in (b), taking "111Z" as an example, replace the zero vector with the complementary vectors V 5 and V 2 Instead, the two vectors act respectively for T sp / 8 of the time, and at the same time place V 2 at the end of the vector synthesis sequence, then i s1 can be collected at t a , and i s2 can be collected at t c , and the vector sequence number can be changed to "111|52|", where |52| means that there are V sp and V 5 and V 2 acting respectively within T

[0085] In addition, for the candidate vectors "1111" and "2222", since they only contain one active vector and no zero vector, the zero vector replacement method cannot be used. Consider using the "measurement vector insertion method" to ensure that two-phase current information can be obtained within one cycle. As Figure 8 shown, taking "1111" as an example, insert the measurement vector V 2 at the end of the vector synthesis sequence, and set the action time to T min to ensure that the current sampling meets the minimum sampling time, then the current reconstruction with a fixed sampling time can be realized. The improved vector synthesis sequence can be expressed as 111|12|, where |12| means that there are V sp / 4 of the time, and there are V 1 and V 2 vectors acting respectively.

[0086] From the above analysis, by adopting the "zero vector replacement" or "measurement vector insertion" method for different types of candidate vectors and combining with the adjustment of the vector synthesis sequence, the fixed-time sampling of the current under the condition of meeting the minimum sampling time can be realized.

[0087] IV. Permanent Magnet Synchronous Phase Current Reconstruction System Based on Discrete Vector Modulation Model Prediction

[0088] In the embodiment of the present invention, the phase current reconstruction of the permanent magnet synchronous motor using the discrete vector modulation model prediction reduces the current reconstruction error and enhances the system stability by using the proposed method. Figure 9 is the simulation current reconstruction performance diagram of two algorithms under a given rotational speed of 300 rpm and a load of 2.5 N·m, including the actual value, the reconstructed value of the A-phase current, and the error between the two. Figure 10Simulation comparison diagram of two algorithms for the motor speed to switch from -300 rpm to 300 rpm in 0.3 s under a 2.5 N·m load. Observe Figure 9 It can be seen that the reconstructed three-phase current has better sinusoidality and lower total harmonic distortion rate under the DSVM-MPC algorithm, and the reconstruction error is also smaller, only within 0.035 A. Observe Figure 10 It can be seen that in terms of dynamic performance, the q-axis current of DSVM-MPC converges more rapidly, proving better dynamic performance. In terms of steady-state performance, compared with DL-MPC, the current ripple of DSVM-MPC is smaller and the current tracking performance is better. In summary, compared with DL-MPC, the proposed algorithm has better tracking performance both statically and dynamically. This is because the proposed algorithm divides the period equally, greatly expanding the number of candidate vectors, making the selected optimal voltage vector closer to the reference voltage vector than the optimal and sub-optimal vectors selected by DL-MPC, that is, the adjacent vectors in the same sector, resulting in smaller voltage tracking error, better current tracking performance, and smaller current ripple. In addition, under DL-MPC, the sampling moments and sampling intervals of the DC bus current for different phase currents are not fixed, so the reconstruction error also has strong uncertainty. In the proposed DSVM-MPC single current control, the current sampling moment and sampling interval are both fixed. The fixed sampling interval avoids the generation of reconstruction dead zones, and the fixed two current acquisitions within one period achieve continuous reconstruction of the three-phase current per period, greatly reducing the reconstruction error. (DL-MPC is the traditional finite set model predictive control method, and DSVM-MPC is the discrete vector modulation model predictive method).

Claims

1. A method for reconstructing phase current of a permanent magnet synchronous motor based on discrete vector modulation model prediction, comprising the following steps: S1: The candidate voltage vector supplement schemes based on the control cycle equal division principle are analyzed. S2: Based on the candidate voltage supplement schemes of step S1, a voltage vector optimization scheme is designed in combination with the deadbeat model predictive control. S3: An improved scheme for vector synthesis sequence is designed based on the principle of fixed sampling time.

2. The control method according to claim 1, characterized in that: In step S1, a candidate voltage vector supplement scheme is designed based on the control cycle equal division principle: The design of candidate voltage vectors includes the following steps: First, in order to improve the accuracy of model prediction of current control, the candidate vectors are considered to be expanded. Common methods include "double vector synthesis", "hexagonal sector bisection according to phase angle", "bisection according to amplitude and phase angle", and "bisection of control cycle". Considering the improvement of current prediction accuracy and the limitation of the minimum sampling time of current sensor control, the present invention adopts the method of bisection of the cycle. The expanded vector is essentially a discrete virtual voltage vector (Discrete VirtualVector, DVV). The number of DVVs is closely related to the control cycle fraction. The control cycle fraction is defined as N, and the total number of newly added DVVs is <h2 style=";text-align:left;direction:ltr">N<h2 style=";text-align:left;direction:ltr"> V <h2 style=";text-align:left;direction:ltr"> <3N<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +3N-6.

3. The control method according to claim 1, characterized in that: In step S2, the voltage vector optimization scheme is designed by analyzing and combining the deadbeat model predictive control: First, for permanent magnet synchronous motor control systems, MPC generally selects state space equations based on the relationship between variables to design a prediction model for discrete time systems, as shown below: According to the discrete state model and forward Euler method, we can get: Where T sp is the switching cycle time, k and k+1 represent the relevant measured value of the kth switching cycle and the predicted value of the k+1th switching cycle respectively. and is the measured value of dq axis current, ω e is the motor speed feedback value, and is the dq axis voltage applied to the motor in the kth switching cycle, and is the input quantity of the controlled system. Among them, the reference current is used and Replace the predicted current of the k+1th switching cycle and The prediction model is obtained as: in and is the reference voltage vector Orthogonal axis decomposition value in the rotated coordinate system. Secondly, taking the first sector as an example, when N=4, the number of small sectors is 16, and each small sector is numbered. The numbering rules are as follows: First, define l1, l2, and l3 as three reference edges, which coincide with the three edges of the triangle of sector I respectively. Then, the reference voltage calculated by the deadbeat prediction is Under the αβ axis, it is decomposed into and but The distances between the vector endpoint and the reference edges l1, l2, and l3 can be calculated as: Then, divide d1, d2, and d3 by Taking integers, we get h1, h2, and h3, which can be expressed as: Among them, floor() is to round down to an integer. Then the small sector number can be calculated as h=100h1+10h2+h3, so the number corresponding to each small sector can be obtained. The three vertices of each small sector triangle correspond to three candidate vectors, and each candidate vector is represented by the serial number of four voltage vectors. When N=4, the corresponding relationship between different small sectors and candidate vector serial numbers is shown in Table 1. Table 1 Candidate vector sequence numbers corresponding to small sectors In judgment After the small sector is located, just The optimal vector can be found by minimizing the distance between the three discrete vectors of the corresponding small sector. Compared with the traditional method that requires traversing all eight vectors, the computational burden is effectively reduced. It should be noted that under conditions such as sudden loading or sudden acceleration of the motor, the reference voltage vector may fall in the overmodulation area outside the hexagonal sector. Considering that the voltage vector with the smallest amplitude error with these vectors is at the sector boundary, the reference vector needs to be overmodulated and proportionally reduced to the linear hexagonal area before the candidate vectors are screened.

4. The control method according to claim 1, characterized in that: In step S3, an improved scheme for vector synthesis sequence based on the principle of fixed sampling time is analyzed and designed: The steps to improve the vector synthesis sequence are as follows: First, in the traditional single-vector model predictive control, only one voltage vector acts in each cycle, so only one-phase current can be collected. DSVM-MPC divides a cycle into two voltage vectors, so a cycle can contain multi-phase current information for three-phase current reconstruction. At the same time, by adjusting the cycle bisection to control the action time of a single voltage vector in a cycle, it can ensure that the current sampling meets the minimum sampling time limit to avoid the reconstruction dead zone problem. Based on the above characteristics, it is considered to combine DSVM-MPC with single current control to improve the accuracy of current reconstruction. Among them, there are two main types of current reconstruction errors. One is the A phase current i a_ t s1 The end-of-cycle current i a_ T sp The second is the error between the C phase current i c_ t s2 and the end-of-cycle current i c_ T sp The current reconstruction needs to calculate and compensate for the two errors at the same time, which is a heavy computational burden. It is considered to reduce the current error to one by improving the vector synthesis sequence to achieve one-step error compensation. By adjusting T sp The number N is divided equally by the cycle to control the action time of each voltage vector to be greater than the minimum sampling time T min , it can ensure that the current is successfully collected during the vector action period, thus avoiding the current reconstruction dead zone problem. In addition, if the candidate vector contains a zero vector and only contains one active voltage vector, such as "1ZZZ" or "ZZZZ", only one-phase current information can be collected in one cycle, and two-phase current cannot be obtained. Since the zero vector can be composed of two complementary non-zero vectors, the "zero vector replacement principle" is considered to solve this problem. The fixed sampling time is t s1 and t s2 ,By replacing the zero vector with the complementary vector and changing the vector synthesis sequence, the current reconstruction at a fixed sampling time can be achieved. In addition, for the candidate vectors "1111" and "2222", since they only contain one active vector and no zero vector, the zero vector replacement method cannot be used. Consider using the "measurement vector insertion method" to ensure that the two-phase current information can be obtained within one cycle. From the above analysis, it can be seen that the "zero vector replacement" or "measurement vector insertion" method is used for different types of candidate vectors, combined with the adjustment of the vector synthesis sequence, the fixed-time sampling of the current under the minimum sampling time condition can be achieved. Different candidate vectors use different vector synthesis sequence improvement methods. In sector I, the improved sequence is used to replace the original sequence, as shown in Table 2, where the two vector numbers in || represent T sp Two vectors acting within / 4 time. Table 2 Candidate vector sequences in sector Ⅰ Each sector has two adjacent vectors. For ease of representation, R represents the first vector in the counterclockwise direction of the sector, L represents the second vector, and the complementary vectors corresponding to these two vectors are H and M respectively. The vector numbers corresponding to the sectors are shown in Table 3. Table 3 Sector corresponding vector number Combining Table 2 and Table 3, the improved sequence of candidate vectors in the entire sector is shown in Table 4. Table 4 Candidate vector synthesis sequence Then a permanent magnet synchronous motor phase current reconstruction control system based on discrete vector modulation model prediction can be constructed.

Images