A Low Switching Frequency Predictive Control Method for Dual Three-Phase Synchronous Motors Based on Dual Boundaries

By constructing error boundaries in the dq and xy subspaces and combining them with an event-triggered mechanism, the switching frequency and computational load of the dual three-phase permanent magnet synchronous motor are reduced. This solves the problems of complex parameter tuning and large computational load in the existing technology, and achieves a balance between current tracking accuracy and harmonic suppression performance, making it suitable for deployment on low-cost embedded platforms.

CN123122334APending Publication Date: 2026-11-17SHANGHAI JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611236886.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies in the control of dual three-phase permanent magnet synchronous motors struggle to effectively reduce switching frequency and computational burden without relying on weight coefficient tuning, while also ensuring current tracking accuracy and harmonic suppression performance. This results in complex parameter tuning and a large computational load, limiting the increase in control frequency and the deployment of algorithms on low-cost embedded platforms.

Method used

A low switching frequency predictive control method for dual three-phase synchronous motors based on dual boundaries is adopted. By constructing elliptical error boundaries in the dq subspace and circular error boundaries in the xy subspace, and combining event triggering mechanism and offline candidate vector lookup table, the average switching frequency of the inverter is reduced and the amount of online calculation is reduced, while balancing current tracking accuracy and harmonic suppression performance is achieved.

Benefits of technology

It achieves a significant reduction in the average switching frequency of the inverter, a significant reduction in online computation, and an effective balance between current tracking accuracy and harmonic suppression performance across the entire operating range, thereby improving the real-time performance of the controller and the quality of the current waveform.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122334A_ABST
    Figure CN123122334A_ABST
Patent Text Reader

Abstract

This invention discloses a low-switching-frequency predictive control method for a dual-three-phase synchronous motor based on dual boundaries, belonging to the field of permanent magnet synchronous motor control technology. The method includes: constructing mathematical models and discrete prediction models of the dual-three-phase permanent magnet synchronous motor in dq and xy subspaces; constructing error boundaries in both dq and xy subspaces to form heterogeneous boundary constraints in both spaces; determining whether to trigger voltage vector optimization based on the current and predicted boundary states; if both the current and predicted states are within the boundaries, maintaining the previous voltage vector; otherwise, initiating optimization; during optimization, rapidly screening candidate vectors based on the reference voltage direction, calculating the dwell time of each vector within the boundaries of each subspace, and performing hierarchical screening and arbitration based on boundary-crossing states to select the optimal voltage vector for output. This invention achieves a significant reduction in the average switching frequency of the inverter, a significant reduction in online computation, and an effective balance between current tracking accuracy and harmonic suppression performance across the entire operating range.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of permanent magnet synchronous motor control technology, and more specifically to a low switching frequency predictive control method for dual three-phase synchronous motors based on dual boundaries. Background Technology

[0002] Dual three-phase permanent magnet synchronous motors (PMSMs) possess advantages such as strong fault tolerance, high power density, and low torque ripple, making them promising candidates for applications in new energy vehicles, electric propulsion, and aerospace. Finite set model predictive control (FMCC), due to its fast response, lack of PWM modulation requirement, and ease of handling multivariate constraints, has been widely adopted in PMSM control systems in recent years.

[0003] To reduce the high switching frequency of traditional finite set model predictive control, existing techniques typically introduce an additional switching action cost term into the cost function, besides the current tracking error term. This switching action cost term can be represented as the number of switching state changes between the current candidate voltage vector and the voltage vector of the previous cycle, multiplied by a weighting coefficient. By increasing the weighting coefficient, the controller considers both current tracking error and the number of switching actions when selecting a voltage vector, thus favoring voltage vectors with fewer or no switching changes, thereby reducing the average switching frequency of the inverter.

[0004] However, the aforementioned existing technologies have the following shortcomings: First, this method requires manually setting a weighting coefficient λ in the cost function to balance current control performance and switching frequency. However, the requirements for both vary under different motor speeds and load conditions. A fixed weighting coefficient makes it difficult to achieve ideal control performance across the entire operating range, and the parameter tuning process is complex, requiring extensive debugging work in practical applications. Second, when the weighting coefficient is increased to reduce the switching frequency, the effect of the switching action term is amplified. The controller may choose a voltage vector with poor current tracking performance to reduce the number of switching operations, leading to decreased current tracking accuracy and increased current harmonics. There is a clear contradiction between reducing the switching frequency and improving current control performance. Third, dual three-phase permanent magnet synchronous motors have a total of 64 basic voltage vectors. Existing methods require traversing all vectors and performing prediction and cost function calculations for each one in each control cycle. This results in a large computational load and places high demands on the real-time processing capabilities of the control chip, limiting the improvement of the control frequency and the deployment of the algorithm on low-cost embedded platforms.

[0005] Therefore, how to effectively reduce the switching frequency and alleviate the computational burden without relying on the tuning of the weighting coefficients, while taking into account both current tracking accuracy and harmonic suppression performance, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a low-switching-frequency predictive control method for dual three-phase synchronous motors based on dual boundaries. By constructing elliptical error boundaries in the dq subspace and circular error boundaries in the xy subspace, and establishing an event triggering mechanism based on the current and predicted boundary states, and combining an offline candidate vector lookup table with a collaborative optimization strategy of the dual-space hierarchical dwell time ratio, this method achieves a significant reduction in the average switching frequency of the inverter, a significant reduction in online computation, and an effective balance between current tracking accuracy and harmonic suppression performance across the entire operating range.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] This invention provides a low-switching-frequency predictive control method for dual three-phase synchronous motors based on dual boundaries, comprising the following steps: Mathematical models and discrete prediction models of dual three-phase permanent magnet synchronous motors in dq and xy subspaces are constructed. Error boundaries are constructed in the dq subspace and xy subspace to form a dual-space heterogeneous boundary constraint. Based on the current boundary state under the current operating state of the dual three-phase permanent magnet synchronous motor and the predicted boundary state under the predicted state obtained from the discrete prediction model, it is determined whether to trigger the voltage vector optimization process. If the current operating state and the predicted state are within the boundary, the voltage vector of the previous control cycle remains unchanged; otherwise, the voltage vector optimization process is started. In the voltage vector optimization process, candidate voltage vectors are quickly screened according to the reference voltage direction, and the dwell time of the screened candidate voltage vectors in different subspace boundaries is calculated. The boundary crossing states of the dq subspace and the xy subspace are combined to perform hierarchical screening and optimization arbitration, and the optimal voltage vector is selected for output.

[0009] Furthermore, the error boundaries are constructed in the dq subspace and the xy subspace, wherein the dq subspace and the xy subspace are each independently selected from one of the following forms: elliptical boundary, circular boundary, hyperelliptical boundary, rectangular boundary, and polygonal boundary.

[0010] Furthermore, the boundary function of the dq subspace is the elliptic error boundary function:

[0011] Among them, e d e q These represent the d-axis and q-axis current errors, respectively, B d B q These represent the maximum allowable errors for the d-axis and q-axis, respectively. The circular error boundary function defined for the xy subspace is:

[0012] Among them, e x e y The current errors along the x and y axes are respectively, B. xy This represents the maximum allowable error for harmonic current.

[0013] Furthermore, the determination of whether the voltage vector optimization process is triggered specifically involves: calculating the dq subspace boundary function value Φ at the current time k, the one-step prediction time k+1, and the two-step prediction time k+2. dq and the boundary function value Φ of the xy subspace xy When Φ at those three moments dq ≤1 and Φ xy If the value is ≤1, it is determined that the current operating state and the predicted state are within the boundary, and the voltage vector of the previous control cycle remains unchanged; otherwise, the voltage vector optimization process is started.

[0014] Furthermore, the method for calculating the reference voltage direction angle is as follows: when the trigger voltage vector is optimized, the reference voltage V in the dq coordinate system is obtained by inverse solving the discrete prediction model based on the predicted current state. d_ref V q_ref The reference voltage V in the αβ stationary coordinate system is obtained by inverse Park transformation. α_ref V β_ref Then, the reference voltage direction angle θ is calculated. ref =atan2(V β_ref V α_ref ).

[0015] Furthermore, the rapid screening of candidate voltage vectors can be achieved using one of the following methods: using an offline fixed sector lookup table, using an online sector partitioning method, or using an adaptive candidate vector generation method.

[0016] Furthermore, the method of using an offline fixed sector lookup table is as follows: the voltage vector plane is evenly divided into 12 sectors at 30° intervals. Each sector pre-stores all voltage vector numbers whose direction angle is within ±45° of the center angle of the sector, forming a fixed-width lookup table. During online control, the candidate voltage vector set is obtained by directly looking up the table according to the sector where the reference voltage direction angle is located.

[0017] Furthermore, the tiered filtering strategy executes different filtering processes according to different operating modes, including: Mode A: Only the dq subspace needs to be controlled, i.e. Or prediction ,and

[0018] Mode B: Only the xy subspace needs to be controlled, i.e. ; Mode C: Both the dq and xy subspaces need to be controlled simultaneously, meaning both are either out of bounds or predicted to be out of bounds.

[0019] Furthermore, the tiered filtering strategy is specifically characterized by: In Mode A, firstly, vectors that can reduce the dq boundary function and have an effective dq dwell time are selected; secondly, vectors that can reduce the xy boundary function are selected from these vectors; and finally, the vector with the largest ratio of dq dwell time to the number of switching actions is selected from the results. In Mode B, firstly, vectors that can reduce the xy boundary function and have an effective xy dwell time are selected; secondly, vectors that can reduce the dq boundary function are selected from these vectors; and finally, the vector with the largest ratio of xy dwell time to the number of switching actions is selected from the results. In Mode C, first, voltage vectors that cause the dq boundary function to decrease are selected. Then, voltage vectors that cause the xy boundary function to decrease are selected from among them. Next, voltage vectors whose ratio of dq dwell time to number of switching actions reaches 90% or more of the maximum value of the ratio are selected from among them. Finally, the voltage vector with the largest ratio of xy dwell time to number of switching actions is selected as the optimal voltage vector.

[0020] Furthermore, for each candidate voltage vector, the number of time steps or time length required for the current state of the dual three-phase permanent magnet synchronous motor under the action of that vector to exceed the boundary of the dq subspace or the xy subspace is calculated, and this is taken as the corresponding dq dwell time T. dq and the dwell time T of xy xy .

[0021] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: Reducing Switching Frequency: This invention utilizes an event-triggered mechanism to maintain the voltage vector from the previous control cycle unchanged when the system state is within the boundary and predicted not to exceed it. Voltage vector optimization is only initiated when the boundary is actually exceeded or predicted to exceed it, thereby reducing the number of control updates. During the candidate voltage vector selection process, the ratio of dwell time to the number of switching actions is used as an evaluation index. While ensuring control performance, vectors with fewer switching actions are prioritized, further reducing the average switching frequency of the inverter and improving converter operating efficiency.

[0022] To reduce computational complexity and improve real-time performance, this invention establishes a fixed-width candidate voltage vector lookup table offline. During online control, only a set of candidate vectors, typically containing about 25 vectors, is obtained by looking up the table based on the reference voltage direction angle, eliminating the need to traverse all 64 voltage vectors. This method reduces the number of vectors requiring dwell time calculation and filtering in each control cycle, lowers the online computation load, and improves the real-time execution capability of the control algorithm. It is suitable for implementation on embedded control platforms such as DSPs and FPGAs.

[0023] To reduce current ripple and improve current waveform quality, this invention employs elliptical boundaries in the dq subspace and circular boundaries in the xy subspace, respectively constraining the current error in both subspaces. When the error is within the boundaries, the controller does not forcibly correct the current; adjustment is only made when the error exceeds the limits. This control method makes the current change process smoother, which helps reduce current ripple.

[0024] Reducing current harmonic content: In this invention, the hierarchical screening strategy prioritizes the control performance of the dq subspace while also considering harmonic suppression in the xy subspace. By constraining the harmonic currents in the xy subspace, the harmonic currents can be limited, reducing the total harmonic distortion rate of the current, which is beneficial for improving the current waveform quality and enhancing the smoothness of motor operation. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0026] Figure 1 This is a flowchart of a low switching frequency predictive control method for dual three-phase synchronous motors based on dual boundaries, provided in an embodiment of the present invention.

[0027] Figure 2 This is a flowchart illustrating the low switching frequency predictive control method for dual three-phase synchronous motors based on dual boundaries provided in this embodiment of the invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1: like Figure 1 As shown, this embodiment of the invention discloses a low switching frequency predictive control method for a dual three-phase synchronous motor based on dual boundaries, comprising the following steps: Mathematical and discrete prediction models of dual three-phase permanent magnet synchronous motors in dq and xy subspaces are constructed to provide a computational basis for subsequent current state prediction and boundary judgment.

[0030] Error boundaries are constructed in the dq subspace and xy subspace to form a dual-space heterogeneous boundary constraint; the dq subspace adopts an elliptical boundary to match the torque control requirements, and the xy subspace adopts a circular boundary to limit harmonic current, thereby achieving decoupling constraints for control objectives of different subspaces.

[0031] Based on the current boundary state of the dual three-phase permanent magnet synchronous motor under the current operating state and the predicted boundary state under the predicted state obtained from the discrete prediction model, it is determined whether to trigger the voltage vector optimization process. If the current operating state and the predicted state are within the boundary, the voltage vector of the previous control cycle remains unchanged; otherwise, the voltage vector optimization process is started. This event triggering mechanism reduces unnecessary control updates and switching actions, thereby reducing the inverter switching frequency.

[0032] In the voltage vector optimization process, candidate voltage vectors are quickly screened according to the reference voltage direction, and the dwell time of the screened candidate voltage vectors in different subspace boundaries is calculated. The out-of-bounds states of the dq subspace and the xy subspace are combined to perform hierarchical screening and optimization arbitration. The ratio of dwell time to number of switching actions is used as the evaluation index to select the optimal voltage vector output, thereby achieving a coordinated balance between control performance and switching losses.

[0033] This invention establishes an event-triggered mechanism based on current and predicted boundary states. Voltage vector optimization is triggered only when the system state goes out of bounds or is predicted to go out of bounds; otherwise, the voltage vector of the previous control cycle remains unchanged. During the optimization process, a large-angle-coverage fixed-width lookup table is used to reduce the number of candidate vectors from 64 to approximately 25. The ratio of boundary dwell time to the number of switching actions is used as an evaluation index for hierarchical screening and arbitration. Ultimately, this invention achieves a significant reduction in the average switching frequency of the inverter and a significant reduction in online computation. Furthermore, the fundamental current error and harmonic current error are independently constrained in the dq subspace and xy subspace, respectively, effectively suppressing current ripple and significantly reducing the total harmonic distortion rate of the current. It also ensures excellent current waveform quality under low switching frequency operating conditions. This invention provides a complete technical solution to the problems of prominent contradiction between switching losses and current quality, difficult parameter tuning, and heavy computational burden in traditional methods.

[0034] The following is a detailed description of the low switching frequency predictive control method for dual three-phase synchronous motors based on dual boundaries of the present invention: This patent proposes a low-switching-frequency predictive control method for a dual three-phase permanent magnet synchronous motor based on elliptical and circular boundaries. This method is based on a mathematical model of the dual three-phase permanent magnet synchronous motor, constructing elliptical error boundaries in the dq subspace and circular error boundaries in the xy subspace. An event-triggered mechanism is established based on the positional relationship between the system state and the boundaries. When the system state is within the boundaries and the predicted state is also within the boundaries, the previous control voltage vector remains unchanged; when the system state exceeds the boundaries or the prediction is about to exceed the boundaries, a voltage vector optimization process is initiated. The control flow of this method is as follows: Figure 2 As shown.

[0035] In the voltage vector optimization process, candidate voltage vectors are first quickly screened using an offline candidate vector lookup table based on the reference voltage direction. Then, the boundary dwell time corresponding to each candidate voltage vector is calculated. In combination with the control requirements of the dq subspace and xy subspace, hierarchical screening and ratio optimization arbitration are performed. Finally, the optimal voltage vector is selected and output to the inverter to achieve the goals of reducing switching frequency, reducing online calculation, and taking current control performance into account.

[0036] The specific implementation steps are as follows: 1. Establish mathematical models of dual three-phase permanent magnet synchronous motors in dq and xy subspaces. Use the forward Euler discretization method to discretize the continuous model in the mathematical model to obtain the predicted expressions for the current in the dq subspace and the current in the xy subspace, providing a predictive basis for subsequent event triggering judgment and voltage vector optimization.

[0037] The prediction expression for the dq subspace current is:

[0038] Among them, T s For the control period, f(·) is the sum of the voltage vector, current, and electric angular velocity ω. e Related functions.

[0039] 2. Based on the control characteristics of different subspaces of the dual three-phase permanent magnet synchronous motor, an elliptic error boundary function is defined in the dq subspace:

[0040] Among them, e d e q These represent the d-axis and q-axis current errors, respectively, B d B q These represent the maximum allowable errors for the d-axis and q-axis, respectively. Define a circular error boundary function in the xy subspace:

[0041] Among them, e x e y The current errors along the x and y axes are respectively, B. xy This represents the maximum allowable error for harmonic currents. A boundary function value less than or equal to 1 indicates that the system state is within the boundary.

[0042] 3. Using the discrete prediction model established in step 1, calculate the boundary function values ​​of the dq subspace and xy subspace at the current time, the one-step prediction time, and the two-step prediction time, respectively. If the current state and both prediction states satisfy Φ... dq ≤1 and Φ xy If the value is ≤1, the voltage vector output from the previous control cycle remains unchanged, and voltage vector optimization is not performed; otherwise, the voltage vector optimization process is initiated. This event-triggered mechanism reduces unnecessary calculations and switching actions by the controller.

[0043] 4. When optimizing the trigger voltage vector, the reference voltage V in the dq coordinate system is obtained by inverse solving the Euler discrete model based on the predicted current state. d_ref V q_ref Then, the reference voltage V in the αβ stationary coordinate system is obtained by the inverse Park transformation. α_ref V β_ref Then, the reference voltage direction angle θ is calculated. ref =atan2(V β_ref V α_ref This direction angle serves as the basis for rapid selection of candidate voltage vectors.

[0044] 5. Offline calculation of the spatial position and orientation angle of all 64 voltage vectors in the αβ plane. The voltage vector plane is evenly divided into 12 sectors at 30° intervals, with the central angles of each sector being 0°, 30°, ..., 330°. For each sector, the voltage vector numbers whose orientation angles fall within ±45° of the sector's central angle are pre-stored, forming a fixed-width lookup table of dimension 12×N, where N is the maximum number of candidate vectors in each sector. Since the candidate voltage vector ranges covered by adjacent sectors overlap significantly, when the reference voltage orientation angle fluctuates slightly near the sector boundary, the change in the candidate vector set is small, thus reducing voltage vector jumps caused by sector switching and helping to maintain the stability of the control system.

[0045] 6. During the control process, the reference voltage direction angle θ obtained in step 4 is used. ref The sector number of the sector is determined, and the candidate voltage vector set corresponding to the sector is obtained directly by looking up the lookup table established in step 5. Only the vectors in the set are calculated later, thus avoiding traversing all 64 voltage vectors.

[0046] 7. For each candidate voltage vector selected, calculate the dwell time T within the elliptical boundary of the dq subspace under its influence. dq The dwell time T within the circular boundary of the xy subspace xy Boundary dwell time reflects the ability of the voltage vector to maintain the system state within the allowable error range. The longer the dwell time, the more beneficial the vector is for extending the boundary holding time and reducing the number of switching operations.

[0047] 8. Based on whether out-of-bounds errors occur in the dq and xy subspaces, the system operating state is divided into three operating modes: Mode A: Only the dq subspace needs to be controlled, i.e. Or prediction ,and And prediction ; Mode B: Only the xy subspace needs to be controlled, i.e. Or prediction ,and And prediction ; Mode C: Both the dq and xy subspaces need to be controlled simultaneously, meaning both are either out of bounds or predicted to be out of bounds.

[0048] Different tiered filtering strategies are used for different modes: Mode A first filters vectors that can reduce the dq boundary function and have an effective dq dwell time. Then, it filters vectors that can reduce the xy boundary function. Finally, it selects the vector with the largest ratio of dq dwell time to the number of switching actions from the filtered results. Mode B first filters vectors that can reduce the xy boundary function and have an effective xy dwell time. Then, it filters vectors that can reduce the dq boundary function. Finally, it selects the vector with the largest ratio of xy dwell time to the number of switching actions from the filtered results. Mode C first filters out voltage vectors that cause the dq boundary function to decrease, then selects voltage vectors that cause the xy boundary function to decrease, then filters out voltage vectors whose ratio of dq dwell time to number of switching actions reaches 90% or more of the maximum value of the ratio, and finally selects the voltage vector with the largest ratio of xy dwell time to number of switching actions as the optimal voltage vector.

[0049] 9. Based on the optimization results, output the inverter switching signal corresponding to the optimal voltage vector to drive the dual three-phase inverter to complete motor control. Simultaneously, update the voltage vector and switching state of the previous control cycle, and count the number of switching actions to provide basic data for predictive control in the next control cycle.

[0050] It should be noted that in the above embodiments of the present invention, the example given is that the dq subspace uses an elliptical boundary and the xy subspace uses a circular boundary. However, the present invention is not limited to this. Those skilled in the art can use other closed geometric boundary shapes to constrain the errors of the two subspaces without departing from the technical concept of the present invention. For example, hyperelliptical boundaries, rectangular boundaries, polygonal boundaries, or other continuous or piecewise continuous closed boundary forms, as long as the boundary can limit the current error of the dq subspace and the xy subspace respectively, and can complete the screening and evaluation of the voltage vector based on the dwell time within the boundary, it belongs to the alternative embodiments of the present invention. All of the above alternative methods can achieve the purpose of the present invention and obtain the corresponding technical effects.

[0051] Furthermore, the boundary parameters such as the major and minor axes of the ellipse and the radius of the circular boundary in this invention are not fixed and can be adjusted according to different motor parameters, control objectives, or operating conditions. For example, smaller boundary parameters can be used under low-speed, low-load conditions to improve current control accuracy, while larger boundary parameters can be used under high-speed, high-load conditions to prioritize ensuring a low switching frequency. Boundary parameters can also be adjusted online adaptively, calculated in real time based on current error or speed to dynamically match current operating requirements. All of the above-mentioned methods of adjusting boundary parameters do not change the basic control concept of this invention and are also alternative implementations of this invention.

[0052] In the above embodiments of the present invention, the boundary state at the current moment, combining one-step and two-step prediction, is used as the basis for determining event triggering. However, the present invention is not limited to this. In practical applications, those skilled in the art can choose to use one-step, three-step, or more-step prediction to achieve boundary prediction based on the computing power and performance requirements of the control system. As long as the prediction step size enables the controller to determine whether the system state is about to exceed the boundary based on the prediction result, and decides whether to update the voltage vector accordingly, it belongs to an alternative implementation of the present invention, and can also achieve the technical effects of reducing switching frequency and reducing unnecessary calculations.

[0053] It should be noted that in the above embodiments of the present invention, the method of obtaining the candidate voltage vector set by establishing a fixed sector lookup table offline is used as an example. However, the present invention is not limited to this. Those skilled in the art can also use other methods to pre-screen the candidate voltage vectors without departing from the technical concept of the present invention, such as online sector partitioning, adaptive candidate vector generation, or preliminary screening of 64 voltage vectors based on other rules to reduce the size of the candidate set. Regardless of the screening method used, as long as its core purpose is to reduce the number of voltage vectors traversed online, reduce the real-time computation burden, and complete the optimized selection of voltage vectors while ensuring control performance, it belongs to the alternative implementation of the present invention and can also achieve the technical effect of reducing the amount of online computation.

[0054] In the above embodiments of the present invention, the dwell time of the candidate voltage vector within the boundary is used as the main evaluation index, and the optimal selection is achieved by combining the number of switching actions. However, the present invention is not limited to this. Those skilled in the art can, based on actual control requirements, further introduce other performance indicators reflecting the system control quality, such as current tracking error, voltage utilization rate, and current harmonic content, on the basis of the boundary dwell time. The dwell time can be combined with one or more of the above indicators to construct a diversified comprehensive evaluation function to complete the final selection of the voltage vector. As long as the evaluation method retains the core idea of ​​measuring the ability of the voltage vector to maintain the system state by using the boundary dwell time, that is, without departing from the technical concept of the present invention, it belongs to the alternative implementation methods of the present invention.

[0055] In summary, this invention addresses the unique characteristic of dual three-phase permanent magnet synchronous motors, which simultaneously possess torque control subspaces (dq subspace) and harmonic control subspaces (xy subspace), by proposing a low-switching-frequency predictive control method based on dual boundaries. This method first constructs an elliptical error boundary in the dq subspace to jointly constrain the direct-axis and quadrature-axis current errors, and then constructs a circular error boundary in the xy subspace to limit the harmonic current errors. By establishing matching boundary forms based on the control objectives of different subspaces, the electromagnetic torque control requirements and harmonic suppression requirements are effectively constrained, avoiding the mutual influence between different control objectives under traditional unified error evaluation methods.

[0056] Based on this, the present invention determines whether to update the voltage vector according to the boundary states of the dq subspace and the xy subspace at the current time and the predicted time. The optimization process is triggered only when either subspace is currently out of bounds or is predicted to be out of bounds. When the predicted state is always within the boundary, the previous control vector is maintained, thereby reducing unnecessary control updates and switching actions.

[0057] During voltage vector optimization, this invention establishes a fixed-width lookup table offline for each sector, containing voltage vectors within a ±45° range of the center angle. During online control, the candidate vector set is obtained only by looking up the table based on the reference voltage direction angle, eliminating the need to traverse all 64 voltage vectors. Furthermore, the overlapping design of candidate vector ranges in adjacent sectors reduces voltage vector fluctuations near sector boundaries, thus reducing online computation while ensuring the optimization range. When both the dq and xy subspaces need to be controlled simultaneously, this invention employs a hierarchical collaborative optimization strategy: First, candidate voltage vectors that can decrease the boundary function of the dq subspace are selected. Then, vectors that can simultaneously decrease the boundary function of the xy subspace are selected, completing the first layer of selection. Next, the ratio of the dwell time of the vectors that passed the first layer of selection within the dq boundary to the number of switching actions is calculated, and vectors with a ratio exceeding 90% are retained, completing the second layer of selection. Finally, the vector with the largest ratio of dwell time within the xy boundary to the number of switching actions is selected from the vectors that passed the second layer of selection as the optimal voltage vector. This method establishes a clear hierarchical cooperative relationship between the dq subspace and the xy subspace, prioritizing the control performance of the dq subspace while taking into account the harmonic suppression effect of the xy subspace. It also uses the ratio of dwell time to the number of switching actions as a unified evaluation index, thus achieving an effective balance between control performance and switching losses.

[0058] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0059] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A low switching frequency predictive control method for a dual three-phase synchronous machine based on double boundaries, characterized in that, Includes the following steps: Mathematical models and discrete prediction models of dual three-phase permanent magnet synchronous motors in dq and xy subspaces are constructed. Error boundaries are constructed in the dq subspace and xy subspace to form a dual-space heterogeneous boundary constraint. Based on the current boundary state under the current operating state of the dual three-phase permanent magnet synchronous motor and the predicted boundary state under the predicted state obtained from the discrete prediction model, it is determined whether to trigger the voltage vector optimization process. If the current operating state and the predicted state are within the boundary, the voltage vector of the previous control cycle remains unchanged; otherwise, the voltage vector optimization process is started. In the voltage vector optimization process, candidate voltage vectors are quickly screened according to the reference voltage direction, and the dwell time of the screened candidate voltage vectors in different subspace boundaries is calculated. The boundary crossing states of the dq subspace and the xy subspace are combined to perform hierarchical screening and optimization arbitration, and the optimal voltage vector is selected for output.

2. The method of claim 1, wherein, The error boundaries are constructed in the dq subspace and the xy subspace, wherein the dq subspace and the xy subspace are each independently selected from one of the following forms: elliptical boundary, circular boundary, hyperelliptical boundary, rectangular boundary, and polygonal boundary.

3. The method of claim 2, wherein, The boundary function of the dq subspace is the elliptic error boundary function: wherein e d , e q are the d-axis and q-axis current errors, respectively, B d , B q are the maximum allowed errors for the d-axis and q-axis, respectively; The circular error boundary function defined for the xy subspace is: where e x , e y are the x-axis and y-axis current errors, respectively, and B xy is the maximum error allowed for the harmonic current.

4. The method of claim 1, wherein, The judging whether to trigger the voltage vector optimization process specifically comprises: calculating dq subspace boundary function values Φ of a current moment k, one-step prediction moment k+1 and two-step prediction moment k+2 dq and xy subspace boundary function values Φ xy When the Φ of the three moments dq ≤1 and Φ xy ≤1, it is judged that the current operation state and the predicted state are located in the boundary, and the voltage vector of the last control period is kept unchanged; otherwise, the voltage vector optimization process is started.

5. The method of claim 1, wherein, The method for calculating the reference voltage direction angle is as follows: When the trigger voltage vector is optimized, the reference voltage V in the dq coordinate system is obtained by inverse solving the discrete prediction model based on the predicted current state. d_ref V q_ref The reference voltage V in the αβ stationary coordinate system is obtained by inverse Park transformation. α_ref V β_ref Then, the reference voltage direction angle θ is calculated. ref =atan2(V β_ref V α_ref ).

6. The method of claim 1, wherein, The rapid screening of candidate voltage vectors can be achieved using one of the following methods: using an offline fixed sector lookup table, using an online sector partitioning method, or using an adaptive candidate vector generation method.

7. The method of claim 6, wherein, The method of using an offline fixed sector lookup table is as follows: the voltage vector plane is evenly divided into 12 sectors at 30° intervals. Each sector pre-stores all voltage vector numbers whose direction angle is within ±45° of the center angle of the sector, forming a fixed-width lookup table. During online control, the candidate voltage vector set is obtained by directly looking up the table according to the sector where the reference voltage direction angle is located.

8. The method of claim 1, wherein, The hierarchical filtering strategy executes different filtering processes according to different operating modes, including: Mode A: only dq subspace needs to be controlled, i.e. or predicted , and Mode B: Only the xy subspace needs to be controlled, i.e. ; Mode C: Both the dq and xy subspaces need to be controlled simultaneously, meaning both are either out of bounds or predicted to be out of bounds.

9. The method of claim 8, wherein, The tiered filtering strategy is specifically as follows: In Mode A, firstly, vectors that can reduce the dq boundary function and have an effective dq dwell time are selected; secondly, vectors that can reduce the xy boundary function are selected from these vectors; and finally, the vector with the largest ratio of dq dwell time to the number of switching actions is selected from the results. In Mode B, firstly, vectors that can reduce the xy boundary function and have an effective xy dwell time are selected; secondly, vectors that can reduce the dq boundary function are selected from these vectors; and finally, the vector with the largest ratio of xy dwell time to the number of switching actions is selected from the results. In Mode C, first, voltage vectors that cause the dq boundary function to decrease are selected. Then, voltage vectors that cause the xy boundary function to decrease are selected from among them. Next, voltage vectors whose ratio of dq dwell time to number of switching actions reaches 90% or more of the maximum value of the ratio are selected from among them. Finally, the voltage vector with the largest ratio of xy dwell time to number of switching actions is selected as the optimal voltage vector.

10. The method of claim 1, wherein, for each candidate voltage vector, the time steps or time lengths required for the current state of the dual three-phase permanent magnet synchronous motor under the action of the vector to exceed the dq subspace boundary or the xy subspace boundary from the current time are calculated respectively as the corresponding dq dwell time T dq and the xy dwell time T xy .