Five-phase permanent magnet synchronous motor dead-beat fault-tolerant direct torque control method based on harmonic current closed-loop suppression

Abstract

The invention discloses a five-phase permanent magnet synchronous motor dead-beat fault-tolerant direct torque control method based on harmonic current closed-loop suppression, and the method comprises the steps: firstly constructing a subspace and a virtual voltage vector in the subspace based on a principle that the harmonic current of the subspace is suppressed to be zero in a normal operation state of a motor; and then, analyzing the distribution characteristics of a double-sub-space virtual voltage vector under any phase open-circuit fault, introducing a simplified dead-beat direct torque control structure, and inhibiting motor open-circuit disturbance by combining a disturbance observer. Furthermore, a virtual voltage vector of the subspace is selected in an open-circuit fault state, and the virtual voltage vector is combined with deadbeat direct torque control to realize closed-loop suppression of harmonic current of the subspace, so that the system can effectively suppress third harmonic current in both normal operation and fault states. According to the method, open-circuit fault diagnosis and system reconstruction are not needed, the torque ripple and the third current harmonic can be suppressed at the same time, and the method has a good engineering application prospect.

Description

Technical Field

[0001] This invention relates to the field of multiphase motor drive and control technology, specifically providing a deadbeat-tolerant direct torque control method for five-phase permanent magnet synchronous motors (PMSMs) based on harmonic current closed-loop suppression. This method aims to solve the problem of stable operation of PMSMs during open-circuit faults. Its applications can cover aerospace, electric vehicles, and ship electric propulsion, among other applications with stringent requirements for system safety and reliability. Background Technology

[0002] Multiphase motors (PMSMs) are renowned for their high torque density, high efficiency, and high reliability, and are widely used in various high-performance drive applications. Compared to three-phase motors, the core value of multiphase motors lies in the redundancy of control degrees of freedom they provide, making them particularly suitable for high-reliability applications. In the event of a fault, they can maintain predetermined performance indicators through fault-tolerant control algorithms.

[0003] Currently, research on fault-tolerant control of five-phase PMSMs mainly focuses on three major directions: vector control, direct torque control (DTC), and model predictive control. Among these, DTC technology is receiving increasing attention in the field of fault-tolerant direct torque control of five-phase PMSMs due to its fast dynamic response, simple system structure, and strong robustness.

[0004] The direct torque control strategy calculates the difference between the current PMSM torque and flux linkage and the given torque and flux linkage, respectively, and inputs them based on the rotor rotating coordinate system. d-q In a deadbeat controller (in a coordinate system), the optimal voltage vector is determined based on the results. However, for five-phase PMSMs, traditional deadbeat direct torque control (DBDTC) suffers from complex flux linkage and load angle calculations, and the determined optimal vector in the fundamental space simultaneously couples to the harmonic space. Therefore, some studies have proposed using a virtual voltage vector to simplify system complexity and improve current quality by making its composite vector in the third harmonic space zero. However, these methods primarily address normal motor operation and essentially employ open-loop control in the harmonic space. When an open-circuit fault occurs in the motor, the system faces severe current distortion, increased torque ripple, and the need for rapid fault diagnosis. Therefore, a fault-tolerant deadbeat direct torque control algorithm with closed-loop harmonic current implementation still requires further research.

[0005] When a single-phase open-circuit fault occurs in a five-phase PMSM (taking phase A as an example), the number of available voltage vectors in the system will decrease from 32 to 16, and the fundamental spatial voltage vectors will exhibit unequal amplitudes and uneven distribution. To suppress torque ripple and reduce current harmonic content, Chinese invention patent application No. 201910669271.9 discloses a fault-tolerant direct torque control method for a five-phase permanent magnet synchronous motor based on a novel fault-tolerant switch table, achieving good fault-tolerant control results. However, this method only designs a unified fault-tolerant switch table for phase A faults and is not universally applicable to fault situations in other phases, and its control prerequisite still requires accurate diagnosis of the faulty phase. To avoid performance degradation caused by misdiagnosis of the faulty phase, IEEE... Transactions on Industrial Electronics The paper "Switching-table-based diagnosis-free fault-tolerant DTC for five-phase PMSM with any phase open-circuit fault" (vol. 71, no. 11, pp. 13790-13800, Nov. 2024) proposes a fault-tolerant control based on virtual vectors for fault-free phase diagnosis of five-phase motors. However, this strategy requires open-loop control of harmonic components. Therefore, further research is needed on a fault-free phase diagnosis-based closed-loop fault-tolerant direct torque control algorithm based on harmonic current to improve its harmonic suppression capability while maintaining dynamic performance. Summary of the Invention

[0006] The purpose of this invention is to provide a deadbeat-free fault-tolerant direct torque control method that can achieve closed-loop suppression of third harmonic current when any one phase of a five-phase permanent magnet synchronous motor experiences an open-circuit fault, without requiring fault phase detection and diagnosis. This method effectively suppresses torque ripple, significantly reduces system computational complexity, avoids reconfiguring the control system due to faults, and thus ensures excellent dynamic and steady-state performance of the motor under both normal and fault conditions.

[0007] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0008] A deadbeat-tolerant direct torque control method for a five-phase permanent magnet synchronous motor based on harmonic current closed-loop suppression is described below: Step 1: Use the Clarke transformation matrix to transform the normal phase voltage to a two-phase stationary coordinate system to obtain the voltage vector distribution of the fundamental and harmonic spaces under normal conditions.

[0009] Step 2, to suppress third harmonic current i x =0, i yWith 0 as the target, built on x-y A virtual voltage vector with a spatially synthesized vector of zero. Specifically, this is achieved through... In space, large and medium base voltage vectors are synthesized according to the action ratio of 0.618:0.382, generating 10 virtual voltage vectors.

[0010] Step 3, using closed-loop control x-y With subspace harmonic components as the target, in x-y Within the subspace, large and medium base voltage vectors are synthesized at a ratio of 0.618:0.382, generating 10 virtual voltage vectors V. i 3’ (i=1-10) This synthesis allows it to... The voltage vector sum in space is zero, thus avoiding any impact on the fundamental frequency control.

[0011] Step 4: When any one phase of the five-phase PMSM experiences an open-circuit fault, the voltage source inverter still outputs the same base voltage vector as under normal operating conditions. However, because the faulty phase cannot receive excitation, the effective voltage vector actually acting on the motor changes. At this time, in the fundamental and third harmonic space, the base voltage vector received by the faulty PMSM can be obtained by mapping the voltage actually borne by the non-faulty phase to the two-phase stationary coordinate system through a modified transformation matrix constructed after removing fault-related terms from the traditional Clarke transformation matrix. Therefore, the base voltage vector actually received by the motor is no longer consistent with the base voltage vector output by the inverter. In the case of phase A being open-circuited, these vectors lack phase A state components. Compared with the base voltage vector set of the normal system, the base voltage vectors obtained under fault conditions not only have a spatial shift but also a reduced total number.

[0012] Step 5: Fault-Tolerant Selection of Virtual Voltage Vector. When an open-circuit fault occurs in any phase, the position shift of the base voltage vector will directly cause a shift in the synthesized virtual voltage vector. To compensate for this shift, the selection strategy for the virtual voltage vector must be reconstructed. The core of this is to establish a new mapping relationship so that the actual virtual voltage vector synthesized at the motor end after the inverter output base voltage vector passes through the fault topology can achieve a unified spatial position and effect with the expected optimal virtual voltage vector calculated by the control algorithm.

[0013] Step 6: To achieve accurate observation and real-time compensation of lumped disturbances caused by unmodeled system dynamics, external load disturbances, and open-circuit faults, and to ensure the motor obtains the optimal virtual voltage vector, this step designs and introduces a disturbance observer. This observer performs online estimation of the aforementioned disturbances and integrates the observed values ​​as feedforward compensation quantities into the control loop. Simultaneously, to simplify the control system structure and maintain the consistency of the design framework, the mathematical model of the five-phase PMSM and the disturbance observer are uniformly established in a rotating coordinate system oriented towards stator flux linkage (…).M-T In a coordinate system, thus in M-T The control algorithm and the observer are integrated and operated in a coordinate system.

[0014] Step 7, based on traditional d-q Deadbeat direct torque control (DBDTC) in coordinate system is converted to stator flux orientation. M-T The coordinate system is rotated and discretized. By setting the stator flux and torque setpoints of the current control cycle to their corresponding predicted values ​​for the next time step, the stator voltage reference value is directly solved, thus achieving complete decoupling control of stator flux and torque in the MT coordinate system. This method effectively avoids the complex calculation of load angle in traditional strategies, significantly reducing the computational load on the system. Furthermore, the disturbance components observed by the disturbance observer (DOB) are... d 1. As a feedforward quantity, it is directly compensated to the torque reference channel, which can effectively suppress torque pulsation caused by open circuit faults and the third harmonic component of back EMF, and improve the steady-state control accuracy of the system.

[0015] Step 8, based on Ten pre-constructed fundamental spatial virtual voltage vectors (V) in the subspace 3 The spatial distribution of V, with adjacent V 3 Using the angle bisector as the boundary, The subspace is uniformly divided into 10 sectors. By determining the sector position of the stator voltage reference vector, the corresponding V for that sector is selected. 3 The optimal virtual voltage vector is used as the agent. Further, the optimal V is calculated using the geometric projection method. 3 The duty cycle is specifically determined by projecting the stator voltage reference vector onto the selected V. 3 The direction is determined based on the projection relationship to establish the ratio of its action time, so as to achieve accurate synthesis of the reference voltage.

[0016] Step 9, based on x‑y Subspace virtual voltage vector (V 3’ Based on the spatial distribution characteristics of the motor under normal operating conditions and any one-phase open-circuit fault conditions, and combined with the principle of harmonic current suppression—that is, weakening harmonic components by applying a voltage vector with an angle greater than 90° to the harmonic current vector—a virtual voltage vector selection table for the harmonic subspace is pre-set. In actual operation, by detecting the real-time position of the harmonic current vector and consulting this table, the virtual voltage vector to be applied can be quickly determined, thereby achieving [the desired effect]. x‑y Directional closed-loop suppression of subspace harmonic currents.

[0017] Step 10: First, compare the given rotational speed n with the actual rotational speed. n The deviation signal is sent to the speed loop PI regulator, and the output is used as the torque reference value. Te Simultaneously, the stator flux linkage is acquired in real time and the current torque is calculated using a stator flux linkage observer. T e Then, the torque reference value... T e The disturbance component output by the disturbance observer d The sum of 1 and the measured torque T e By comparison, the torque deviation Δ is obtained. T e ； Simultaneously calculate the given magnetic flux. With observation magnetic flux deviation Based on torque deviation Δ T e Magnetic flux deviation The quadrant information of the harmonic current components, combined with a simplified deadbeat-free direct torque control algorithm, is used to synchronously select... Subspace and x-y The optimal virtual voltage vector in the subspace is determined, and the corresponding duty cycle is calculated. Finally, the modulation signal is input to the PWM module to generate the drive waveform, which controls the switching state of the power devices in the voltage source inverter.

[0018] The present invention has the following beneficial effects: 1) This invention synthesizes virtual voltage vectors in the α-β and xy subspaces using a fixed ratio (0.618:0.382), and combines this with a preset virtual voltage vector selection table for the harmonic subspace, achieving continuous closed-loop suppression of the third harmonic current before and after an open-circuit fault in any phase. This effectively overcomes the limitation of traditional methods in effectively controlling harmonic currents under fault conditions without fault detection means, significantly improving the current quality and steady-state operation performance of the motor under fault conditions.

[0019] 2) By introducing based on M-T The coordinate system disturbance observer accurately observes the lumped disturbances caused by open-circuit faults, third harmonic back EMF, and external interference, and uses these disturbances as feedforwards to compensate the torque loop, thereby effectively suppressing torque pulsations caused by open circuits in any single phase. This method does not rely on the diagnosis of the faulty phase, thus avoiding the risk of misdiagnosis and system performance degradation during the diagnosis delay, and improving the operational reliability and dynamic control quality of the five-phase PMSM under fault conditions.

[0020] 3) By converting traditional deadbeat direct torque control to stator flux orientation... M-TThe coordinate system is discretized, achieving complete decoupling control of torque and flux linkage. This design retains the advantage of fast torque response in traditional DBDTC systems while effectively avoiding the complex calculation of load angles in conventional strategies. This significantly reduces system computational complexity, saves CPU resources, and provides a simple and efficient overall solution for real-time system control.

[0021] 4) This invention not only effectively suppresses torque pulsation caused by open-circuit faults in any phase, but also achieves closed-loop suppression of harmonic currents, improving the steady-state performance of the motor under fault conditions and making its dynamic performance comparable to that under normal conditions.

[0022] 5) This invention avoids the fault diagnosis work of the faulty phase, thereby avoiding the possibility of diagnostic errors and the deterioration of motor performance caused by the fault during the fault diagnosis time, and improving the reliability and performance of the five-phase PMSM operation; at the same time, it realizes the use of the same direct torque control strategy before and after the fault, avoiding the need to re-architect the control system after the fault occurs, and ensuring similar steady-state and dynamic performance before and after the fault. Attached Figure Description

[0023] Figure 1 Block diagram of deadbeat-tolerant direct torque control for a five-phase permanent magnet synchronous motor based on harmonic current closed-loop suppression; Figure 2 The virtual voltage vector distribution of the fundamental and harmonic space under normal conditions of a five-phase PMSM is shown in (a) for the a-β subspace and (b) for the xy subspace. Figure 3 The virtual voltage vector distribution of a five-phase PMSM under the condition of open circuit in any phase; Figure 4 For five-phase PMSM in M-T Motor model in coordinate system; Figure 5 Block diagram of the disturbance observer; Figure 6 For optimal V 3 Duty cycle modulation principle diagram; Figure 7 for x-y Subspace harmonic vector distribution; (a) permanent magnet flux linkage has only fundamental component; (b) permanent magnet flux linkage contains third harmonic component; Figure 8 The torque and current waveforms of the five-phase PMSM are sequentially switched from normal operation to phase AE open circuit fault state; (a) from normal to phase B fault operation; (b) from phase C fault operation to phase E fault operation; Figure 9The torque step and corresponding speed and current waveforms under the open circuit fault state of the five-phase PMSM: (a) load step increase; (b) load step decrease. Detailed Implementation

[0024] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0025] The control block diagram designed in this invention is as follows: Figure 1 As shown, this invention consists of a dual closed-loop system composed of a deadbeat-based direct torque control inner loop and a PI control speed outer loop. The speed is given... n With real-time feedback speed n The difference is calculated, and after passing through a PI controller, the output is the given torque. T e The DOB method is used to estimate the torque ripple caused by an open-circuit fault in any phase, as well as external disturbances and disturbances caused by unmodeled disturbances. The results are then analyzed. d 1 as torque reference T e Feedforward compensation can effectively suppress torque ripple. The torque is compared with the output of the flux observation. T e With magnetic linkage With the given torque respectively T e and given magnetic flux Difference, input based on M-T The target voltage vector is obtained from the simplified DBDTC coordinate system. Sector determination is performed on the target voltage vector, and the optimal V is selected. 3 And its duty cycle, and select the corresponding V in the diagonal quadrant according to the quadrant where the harmonic current component is located. 3' Finally, after PWM switching sequence optimization, the switching signals of each bridge arm are sent to the voltage source inverter to achieve direct torque control of the motor.

[0026] Under the condition that any phase of a five-phase permanent magnet synchronous motor experiences an open-circuit fault, the deadbeat-tolerant direct torque control method based on harmonic current closed-loop suppression is implemented according to the following steps: First, the standard Clarke transformation matrix is ​​used to transform the phase voltages output by the upper bridge arm power devices in the voltage source inverter driving the five-phase PMSM under different switching states to a two-phase stationary coordinate system, thereby obtaining the set of base voltage vectors in the fundamental subspace and harmonic subspace when the motor is running normally.

[0027] To suppress third harmonic current (i.e., satisfy) i x =0, i y=0) is the control target, at the fundamental frequency Two types of base voltage vectors, large and medium, with the same direction, are selected in the subspace and synthesized into a virtual voltage vector (V) with an action ratio of 0.618:0.382. 3 ), so that this set of vectors is in the third harmonic ( x‑y The synthesized voltage amplitude in the subspace is zero, thus effectively suppressing the third harmonic current. The constructed virtual voltage vector V1 3 ~V2 3 In The magnitude of each subspace is 0.5528. U dc As shown in Table 1, Table 1 represents the virtual voltage vector: ; In the table above, V i 3 (i=1-10) Represented as Name of the virtual voltage vector in space; V i(i=0-31) Represented as Name of the space base voltage vector; V l for Large spatial base voltage vector; V m for Space-type base voltage vector.

[0028] Following the method in Table 1, by... Within the subspace, the large and medium vectors with the same direction are coordinated and their interaction ratio is strictly limited to 0.618:0.382, so that the large and medium base voltage vectors are in harmony with the third harmonic (…). x-y The voltage vector components generated in the subspace are superimposed in a canceling manner, so that the equivalent output voltage of the virtual voltage vector in the xy subspace is zero. Thus, feedforward and high-efficiency suppression of the third harmonic current is achieved without additional closed-loop control of the harmonic current.

[0029] Based on the above implementation method, a dual design concept is adopted for closed-loop control. x-y With subspace harmonic components as the target, in x-y Two types of base voltage vectors, large and medium, are selected in the subspace and synthesized into a virtual voltage vector V with an action ratio of 0.618:0.382. 3' This set of vectors in The synthesized voltage amplitude in the subspace is zero, thus avoiding coupling effects on the fundamental wave control. The constructed virtual voltage vector V1 3' ~V2 3' exist x-y The subspace has a uniform amplitude of 0.5528. U dc.

[0030] Table 2 is... x-y Virtual voltage vector in space: ; In the table above, V i 3’ (i=1-10) Represented as x-y Name of the virtual voltage vector in space; V i(i=0-31) Represented as x-y Name of the space base voltage vector; V l for x-y Large spatial base voltage vector; V m for x-y Space-type base voltage vector.

[0031] Following the method in Table 2, x-y Within the fundamental subspace, two types of base voltage vectors, large and medium, are selected and synthesized into a virtual voltage vector at a fixed ratio of 0.618:0.382. This precise design of the specific ratio ensures that the voltage vector components generated by this set of vectors in the fundamental subspace cancel each other out, meaning their net impact on the fundamental space is zero. This inherent decoupling characteristic allows the controller to... x-y The subspace can be independently controlled in a closed loop without worrying about interference with... Torque and flux control in the subspace generate crosstalk. Simultaneously, this set of vectors... x-y The subspace has a uniform and constant amplitude of 0.5528. U dc This makes it possible for it to serve as an ideal voltage reference for harmonic control.

[0032] Therefore, the relationship between each synthesized virtual voltage vector and the fundamental voltage vector is shown in Table 3, and the distribution of the virtual voltage vectors in the fundamental and harmonic spaces is shown in Table 3. Figure 2 As shown.

[0033] Table 3 Under normal circumstances and x-y Composition table of subspace virtual vectors: ; When any phase of the motor experiences an open-circuit fault, without loss of generality, the control method will be explained using an open-circuit fault in phase A as an example. When an open-circuit fault occurs in phase A of a five-phase PMSM, the voltage source inverter still outputs the original set of base voltage vectors according to normal operating conditions. However, since the faulty phase cannot form a current path, the effective voltage vector actually received by the motor changes. At this time, the base voltage vector actually acting on the faulty motor can be obtained by performing coordinate transformation on the voltages of the non-faulty phases (phases B, C, D, and E) using a modified transformation matrix obtained by removing the relevant terms of phase A from the traditional Clarke transformation matrix.

[0034] It should be noted that under normal operating conditions, the base voltage vector received by the motor is exactly the same as the base voltage vector output by the inverter; however, under a phase A fault, the two are no longer the same. To maintain consistency in the analysis, this implementation method still uses the voltage vector naming rules under normal operating conditions to identify the voltage vector received by the motor under fault conditions. Compared to normal operating conditions, the base voltage vector received by the motor under fault conditions not only has a shifted spatial distribution, but its usable number is also reduced accordingly.

[0035] When an open-circuit fault occurs in any of phases B, C, D, or E, the analysis method for phase A faults can be referenced. By modifying the corresponding Clarke transform matrix terms, the corresponding fundamental and third harmonic subspace voltage vector distribution characteristics can be obtained. It should be noted that the modified Clarke transform matrix is ​​only used to analyze the voltage vector offset under fault conditions. In the actual controlled voltage source inverter and the control strategy described in this invention, the traditional Clarke transform matrix is ​​still used.

[0036] When any one phase of the five-phase PMSM experiences an open-circuit fault, the base voltage vectors constituting the virtual voltage vector maintain the same composition relationship and duration as under normal operating conditions. Taking an open-circuit fault in phase A as an example, the voltage source inverter continuously outputs the same medium and large base voltage vectors as under normal operating conditions to synthesize the virtual voltage vector. Therefore, under phase A fault conditions, the correspondence between the virtual voltage vector actually received by the motor and its constituent base voltage vectors still follows the synthesis rules shown in Table 4.

[0037] Table 4 is the synthesis table of virtual vectors under the open-circuit fault condition of phase A: ; Under the open-circuit fault condition of phase A, the position of the composite virtual voltage vector received by the PMSM is also shifted due to the offset of the base voltage vector received by the PMSM.

[0038] The method for constructing a virtual voltage vector under an open-circuit fault in phase A can be further extended to the case where any one of phases B, C, D, or E experiences an open-circuit fault. This allows for the acquisition of the spatial distribution of the virtual voltage vector received by the five-phase PMSM under any phase open-circuit fault condition, such as... Figure 3 As shown in the figure. Analysis indicates that the virtual voltage vector actually received by the motor under fault conditions has a significant spatial phase shift compared to normal operating conditions. If the voltage source inverter continues to output the virtual voltage vector according to the sector division rules and vector selection strategy under normal operating conditions, the electromagnetic torque and flux linkage actually acting on the motor will deviate significantly from the expected values, significantly worsening the system's operating performance.

[0039] To address this, this implementation establishes a remapping mechanism for the virtual voltage vector under fault conditions. This mechanism comprehensively considers multiple factors, including torque ripple caused by open-circuit faults, unmodeled dynamics, and external disturbances, to reselect and optimize the virtual voltage vector. This mechanism ensures that the base voltage vector output by the voltage source inverter, after being coupled through the fault topology, ultimately synthesizes a virtual voltage vector at the motor end that perfectly matches the desired value, thereby effectively maintaining the system's dynamic response performance and operational stability under fault conditions.

[0040] In a five-phase PMSM system, unmodeled dynamics, external disturbances, and open-circuit faults can introduce disturbances that affect system stability. To achieve high-performance control, these disturbances need to be observed and compensated for in real time. Therefore, this invention designs a Disturbance Observer (DOB) to estimate the lumped disturbance caused by multiple factors. To simplify the system structure and facilitate DOB integration, the motor model and DOB are unified and established on the stator flux orientation. M-T In the coordinate system, M-T coordinate system as Figure 4 As shown.

[0041] Therefore, from the two-phase stationary coordinate system to M-T The Park matrix of the coordinate system is: (1); In the formula: Represented as stator flux linkage and Angle between axes.

[0042] In the modeling process, the back EMF pulsations caused by open-circuit faults, changes in motor parameters, and unmodeled dynamics (including the third harmonic component of the permanent magnet flux linkage) are all treated as system disturbances. Based on this, a five-phase PMSM is established under normal and open-circuit fault conditions. M-T The unified mathematical model in the coordinate system is expressed as follows: (2); In the formula: u M , u T , i M , i T , , They are respectively M-T Voltage, current, and stator flux linkage components in the coordinate system; The stator flux linkage angular velocity; The rotor flux linkage angular velocity; p This represents the number of pole pairs of the motor. R s The stator resistance of the motor; JThis refers to the moment of inertia of the motor. f M , f T , The disturbance quantity can also be expressed as: (3); in, T l Represents the load torque, Δ R s For parameters R s The disturbance, Δ T l External disturbance; , and It is the unmodeled part of the motor.

[0043] In the stator flux linkage coordinate system, the stator flux linkage vector is... M If the axes coincide, then: (4); in, δ The load angle.

[0044] Therefore, equation (2) can be expressed as: (5); Further simplification and rewriting of the motor model shown in equation (5) yields: (6); In the formula: , n The rotational speed is given.

[0045] Define the system variables as follows: (7); The second-order mathematical model of the perturbated five-phase PMSM can then be expressed as: (8); In the formula: k =-2.5 pi T ; ; .

[0046] According to equation (8), d Since the 1 and control inputs are on the same channel, the DOB design can be represented as follows: (9); In the formula: and They are d 1 and d The estimated value of 2; p 11 and p 12 As an auxiliary variable; k 11 and k 12 This is an optional constant.

[0047] According to equation (9), the structural block diagram of DOB is as follows: Figure 5 As shown.

[0048] Disturbance components d 1 includes the external disturbances experienced by the system. Based on this, the constructed disturbance observer can be used to estimate the torque ripples caused by open-circuit faults in real time, and the observation results can be introduced into the torque reference channel as a feedforward compensation.

[0049] It should be noted that, although this disturbance observer is based on M-T Coordinate system design, its output d Theoretically, a value of 1 can also be directly fed forward to a rotor-based rotating coordinate system ( In the traditional DBDTC (coordinate system), torque ripple caused by open-circuit faults is suppressed. However, the traditional DBDTC... The calculation process in the coordinate system is quite complex and difficult to implement. Considering the controlled five-phase PMSM... d shaft and q The axial inductances are approximately equal (both are...). If the traditional DBDTC is changed from Coordinate system transformation to Coordinate systems can effectively simplify the structure of control systems and significantly reduce the computational burden of algorithms.

[0050] Therefore, by substituting equation (4) into equation (5) and discretizing it, we can finally obtain: (10); In the formula: u M ( k ), u T ( k ), i M ( k ), i T ( k ) 、 The first k time M-TVoltage, current, and stator flux linkage components in the coordinate system; , The first k Stator flux linkage and torque angle at time +1.

[0051] Because of the load angle The load angle is relatively difficult to obtain, thus significantly increasing the difficulty of calculating the stator voltage. With constant stator and rotor flux linkages, torque is the only factor. Therefore, based on the stator flux linkage formula and the magnetic co-energy product method, the expression for torque can be written as: (11); In the formula, Let be the rotor flux linkage. By differentiating and discretizing equation (11) over time, we obtain: (12); in, T e For torque reference, T e ( k ) is the first k instantaneous torque; These are the given torque angles; Let be the amplitude of the permanent magnet flux linkage. Transforming the above equation, we obtain the relationship between the virtual load angle and the torque: (13); Furthermore, let the stator flux given and torque angle given in equation (10) be equal to the first... k The stator flux linkage and torque angle at time +1 are used to obtain the stator voltage reference. Therefore, the reference stator voltage can be expressed as: (14); In the formula, u M , u T for M-T Reference stator voltage in the coordinate system; For stator flux linkage.

[0052] Using the inverse matrix of equation (1), M-T Reference stator voltage conversion in coordinate system to In coordinate system, (15); in, , for Reference stator voltage in the coordinate system; TMT -1 for M-T The Park inverse matrix of the coordinate system.

[0053] The voltage reference values ​​on the M-axis and T-axis are directly related only to the changes in stator flux linkage and torque, respectively. This characteristic enables voltage command based on the stator flux linkage coordinate system to achieve complete decoupled control of stator flux linkage and torque.

[0054] Compared with the traditional DBDTC, the simplified DBDTC used in this scheme only needs to calculate two state variables: torque and stator flux linkage. This avoids the complex calculation of load angle in the traditional method, thus significantly reducing computational complexity and processor burden.

[0055] Furthermore, since both the constructed perturbation observer and the simplified DBDTC are based on M-T Coordinate system design, the system does not need to perform d-q coordinate system to M-T Additional coordinate system transformations allow all calculations to be performed directly in a unified coordinate system, further simplifying the algorithm structure. Meanwhile, the perturbation components obtained from DOB observations... d 1. It can be directly introduced as a feedforward quantity into the simplified torque control loop of DBDTC, effectively suppressing torque pulsation caused by open circuit faults and the third harmonic component of back EMF, and improving the control performance of the system under fault conditions.

[0056] Based on 10 virtual voltage vectors V 3 exist The spatial distribution of the subspace is defined by the angle bisectors of adjacent vectors, dividing the plane into 10 uniformly distributed sectors, each with an angle of 36 degrees. ° Determined by equations (14) and (15) The subspace stator voltage reference vector can be selected as the optimal virtual voltage vector based on its sector location, with V3 corresponding to that sector chosen as the optimal virtual voltage vector. To achieve precise modulation, the duty cycle of the optimal V3 needs to be calculated, specifically using a geometric projection method: projecting the reference voltage vector onto the selected optimal V3. 3 The direction, through this projection relationship, determines its time ratio, and the process is as follows: Figure 6 As shown, the specific duty cycle formula can be expressed as: (16); In the formula: Reference vector and The angle between axes; For optimal V 3 and The angle between axes.

[0057] when Optimal V of subspace 3While the duty cycle and operating time are determined, suppression of the third harmonic current is not considered. Therefore, it is necessary to further improve the control strategy mentioned above to enable it to have closed-loop regulation and suppression capabilities for harmonic currents.

[0058] Five-phase PMSM x-y The voltage model in the subspace can be expressed as: (17); In the formula, u xy , i xy They are respectively x-y Voltage and current components in the subspace; L s3 for x-y The inductive component of the subspace, They are respectively x-y The permanent magnet flux component of the subspace.

[0059] As can be seen from equation (17), the harmonic current is mainly affected by the nonlinear relationship and the third harmonic of the permanent magnet flux linkage. By discretizing equation (17), the predicted harmonic current vector can be obtained: (18); According to the analysis of equation (18), in order to effectively suppress the third harmonic current, it is necessary to... x-y Apply a vector to the subspace that is the same as the current harmonic current vector. I xy (k) Voltage vectors in opposite directions u xy (k) To visually demonstrate this control principle, Figure 4 .5 describes the spatial distribution relationship of each vector from a geometric perspective. After decoupling is achieved through Clarke transform, Subspace and x-y The fundamental and third harmonic components in the subspace are independent of each other, and the third harmonic components of the current, permanent magnet flux linkage, and back electromotive force are... x-y All subspaces are 3 ω r The electric angular velocity of the rotation is synchronized. Therefore, u xy , i xy and exist x-y Subspaces with the same electric angular frequency can have their vector composition relationship uniformly characterized as follows: Figure 7 The vector diagram shown.

[0060] like Figure 7As shown in (a), when the permanent magnet flux has an ideal sinusoidal distribution (i.e. )hour, i xy ( k The change of +1 depends only on the currently applied voltage vector. u xy ( k ) and harmonic current i xy ( k According to the principle of vector composition, for (k+1)T s To constantly reduce the amplitude of harmonic currents, it is necessary to make u xy ( k )and i xy ( k The angle between them is greater than 90°, meaning that it remains constant within each control cycle. u xy ( k )and I xy (k) The angle is obtuse, thus ensuring that the applied voltage can effectively suppress harmonic currents.

[0061] like Figure 7 As shown in (b), when the permanent magnet flux linkage contains a third harmonic component... i xy ( k +1) Simultaneously affected u xy ( k ), i xy ( k )and The combined effects. In this case, for ( k +1) T s Harmonic suppression must be achieved at all times, but control is still necessary. u xy ( k )and I xy ( k The angle between the two sides is obtuse, and this method can also effectively suppress harmonic currents. Therefore, by selecting the angle with... I xy Voltage vectors with opposite polarities u xy A stable harmonic current suppression mechanism can be established to ensure that the system can achieve high-performance harmonic suppression under different magnetic flux conditions.

[0062] according to Figure 3 shown x-yThe characteristics of the virtual voltage vector distribution in the subspace show that, regardless of whether the five-phase PMSM is in normal operation or in an open-circuit fault state in any one phase, x-y Virtual voltage vector V2 in subspace 3' V5 3' V7 3' V 10 3' They are always located in the first to fourth quadrants, respectively. Based on the principle of harmonic current suppression and the analysis of equation (18), in order to effectively suppress the harmonic current at time (k+1), it is necessary to make x-y Voltage vector applied to subspace u xy ( k ) and the current harmonic current vector I xy ( k Maintain an obtuse angle relationship.

[0063] Based on the above principle, through detection I xy ( k The polarity of the harmonic current vector can be used to quickly determine the quadrant in which it is located, and then the virtual voltage vector V corresponding to its diagonal quadrant can be selected. 3' As a control variable, this control strategy is applicable to any single-phase open-circuit fault condition without changing the control structure. It achieves directional suppression of harmonic components through diagonal quadrant vector selection, ultimately establishing... x-y Qualitative closed-loop control of subspace harmonic currents. The corresponding virtual voltage vector optimization relationship is shown in Table 5.

[0064] Table 5. Selection Table of Virtual Voltage Vectors in Harmonic Subspace: ; Determine the proposed strategy and x-y The optimal virtual voltage vector in the subspace and its corresponding duty cycle are input into the PWM module to generate a PWM waveform, enabling high-performance operation of the five-phase PMSM under normal conditions and under open-circuit faults in any one phase.

[0065] The torque and current waveforms of the motor in steady state are as follows: Figure 8 As shown, it can be seen that this applies to both normal operation and the sequential occurrence of open-circuit faults from phase A to phase E. x-y The use of subspace virtual voltage vectors can improve the sinusoidal nature of phase currents, effectively suppress harmonic currents, and eliminate the need for fault phase diagnosis and fault-tolerant algorithm injection. This allows for direct, fault-tolerant control of faulty motors without the need for fault diagnosis, ensuring high reliability of the motor control strategy. Under fault-tolerant conditions, the load step waveform is as follows: Figure 9As shown, torque tracking is timely and accurate when the load increases by a step from 3 Nm to 5 Nm. The experimental waveforms above fully demonstrate that the algorithm of this invention, when applied to a five-phase PMSM with an open-circuit fault in any phase, exhibits excellent load dynamic response performance.

[0066] In summary, this invention integrates a disturbance observer within the DBDTC framework to achieve real-time observation and feedforward compensation of system disturbances and fault components. By constructing a specific harmonic subspace virtual voltage vector and based on the harmonic current suppression principle, a complete harmonic current closed-loop fault-tolerant control architecture is formed. This invention effectively overcomes the inherent defects of traditional DTC methods, such as large torque ripple and insufficient harmonic current suppression capability, which rely on switching meters and only use fundamental space virtual voltage vectors. Experimental results verify that the invented algorithm can effectively suppress torque ripple caused by open-circuit faults in any one phase, and maintain the steady-state and dynamic performance of direct torque control, ensuring stable operation of the motor under normal conditions and under open-circuit fault conditions in any one phase.

[0067] It should be understood that the above embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention by those skilled in the art in various equivalent forms fall within the scope defined by the appended claims.

Claims

1. A method for deadbeat-tolerant direct torque control of a five-phase permanent magnet synchronous motor based on harmonic current closed-loop suppression, characterized in that, Includes the following steps: Step 1: Using the Clarke transformation matrix, the output voltage of the voltage source inverter used to drive the five-phase permanent magnet synchronous motor (PMSM) when the power devices of the upper bridge arm are in different switching states is transformed into a two-phase stationary coordinate system, thereby obtaining the fundamental voltage vector and its spatial distribution characteristics located in the fundamental space and the third harmonic space under standard operating conditions. Step 2, to suppress third harmonic current i x =0 and i y =0 as the principle, xy The virtual voltage vector in space is synthesized to zero, and thus in A virtual voltage vector V is obtained in space, synthesized from large and medium base voltage vectors. i 3 subscript i =1~10; therefore, in The space uses large and medium-sized basic voltage vectors to synthesize a virtual voltage vector; Step 3, using closed-loop control xy Based on the principle of subspace harmonic components, in xy Using large and medium-sized base voltage vectors in space, a virtual voltage vector V is synthesized at a certain ratio. i 3’ subscript i =1~10, making The virtual voltage vector synthesis in space is zero, avoiding the introduction of... Voltage components in space; Step 4: When any one phase of the five-phase PMSM experiences an open-circuit fault, the base voltage vector received by the faulty PMSM in the fundamental and third harmonic space can be obtained by mapping the voltage actually borne by the non-faulty phase to the two-phase stationary coordinate system through a modified transformation matrix constructed by removing fault-related terms from the traditional Clarke transformation matrix. Therefore, the base voltage vector actually received by the motor is no longer consistent with the base voltage vector output by the inverter. In the case of phase A being open-circuited, these vectors lack phase A state components. Compared with the base voltage vector set of the normal system, the base voltage vectors obtained under fault conditions not only have a spatial position shift, but their total number is also reduced. Step 5: Under any open-circuit fault condition, the virtual voltage vector received by the faulty motor will be offset due to the position offset of the base voltage vector it forms. Therefore, the virtual voltage vector actually received by the PMSM needs to be reselected so that after the voltage source inverter sends out the corresponding base voltage vector, the virtual voltage vector actually obtained by the faulty motor is consistent with its expected optimal virtual voltage vector. Step 6: To accurately observe and compensate for the combined disturbances caused by unmodeled system dynamics, external interference, and open-circuit faults, and to ensure the motor obtains the optimal virtual voltage vector, a disturbance observer (DOB) is designed and introduced for real-time estimation and compensation of the aforementioned lumped disturbances. Simultaneously, to simplify control design and observer construction, the mathematical model of the five-phase PMSM and the DOB are established together in a rotating coordinate system oriented towards stator flux linkage, i.e. MT In a coordinate system, to maintain consistency between the design of the control system and the observer under a unified coordinate system; Step 7, based on dq Traditional deadbeat direct torque control (DBDTC) in coordinate system conversion to MT The coordinate system is discretized, and the stator flux and torque setpoints are set to equal the stator flux and torque at the next moment to obtain the stator voltage reference, thereby achieving decoupled control of the stator flux and torque. Furthermore, the disturbance components observed by DOB are... d 1. Direct feedforward to torque reference effectively suppresses torque pulsation caused by open circuit faults and the third harmonic component of back EMF; Step 8, based on in The 10 V-shaped subspaces are constructed 3 The distribution of adjacent V 3 The angle bisector serves as the sector boundary, dividing the area into 10 sectors. The V corresponding to the sector containing the stator voltage reference vector is selected. 3 That is, the optimal V 3 The optimal V is calculated using the geometric method. 3 Duty cycle; Step 9, based on xy The spatial distribution of the subspace virtual voltage vector under normal motor conditions and open-circuit fault conditions of any phase is determined. Based on the principle of harmonic current suppression, a harmonic subspace virtual voltage vector selection table is preset, and then the closed-loop suppression of harmonic current is achieved through this table.

2. The control method according to claim 1, characterized in that, It also includes step 10, which sets the given rotational speed. n and actual speed n The difference is processed by the speed loop PI regulator, and the output is the given torque. T e Simultaneously, the stator flux linkage is obtained through a stator flux linkage observer. And calculate the torque. T e ; give torque T e Observing the disturbance components with the disturbance observer d The sum of 1 and the estimated torque T e The difference is obtained by subtraction Δ T e Given a magnetic flux With estimation of magnetic flux Difference obtained ; Based on the instantaneous torque deviation Δ T e Instantaneous deviation of stator flux And the quadrant in which the harmonic components are located, combined with simplified deadbeat direct torque control, select subspace and xy The optimal virtual voltage vector in the subspace is calculated and the duty cycle is input into the PWM module to generate a PWM waveform. This waveform controls the on and off of the power devices in the voltage source inverter, enabling high-performance deadbeat direct torque control under normal conditions and open-circuit fault conditions of the five-phase PMSM.

3. The control method according to claim 1, characterized in that, In step 2 Synthesis of spatial virtual voltage vectors to suppress xy Based on the principle of harmonic current in space, a virtual voltage vector is synthesized using large and medium fundamental voltage vectors with the same direction in the fundamental frequency space. The synthesis ratio is 0.618:0.382 to ensure that it is within the harmonic current range of the fundamental frequency space. xy The voltage in space is zero, thus yielding the virtual voltage vector synthesized from the aforementioned basic voltage vectors in the fundamental space; these virtual voltage vectors V1 3 -V 10 3 The amplitude is 0.5528 U dc , U dc This is the bus voltage.

4. The control method according to claim 1, characterized in that, In step 3 xy The synthesis of spatial virtual voltage vectors uses xy Large and medium base voltage vectors with the same direction in space are combined into virtual voltage vectors with an action ratio of 0.618:0.

382. These virtual voltage vectors V i 3’ exist Spatial amplitude is zero, subscript i =1~10, in xy The spatial amplitude is 0.5528. U dc In preparation for closed-loop suppression xy Subspace harmonic components are used.

5. The control method according to claim 1, characterized in that, In step 4, for any one phase of the five-phase PMSM experiencing an open-circuit fault, if phase A of the motor experiences an open-circuit fault, the voltage source inverter still feeds the same base voltage vector to the faulty motor as under normal conditions. However, phase A of the motor cannot receive the voltage fed by the voltage source inverter, so the base voltage vector received by the faulty motor is missing the phase A state. For ease of analysis, the voltage vector received by the faulty motor is still named according to the voltage vector received under normal conditions, which is consistent with the naming rules of the voltage vector sent by the voltage source inverter. If an open-circuit fault occurs in one of the other four phases, referring to the case of an open-circuit fault in phase A, the voltage vectors and their distributions in the fundamental and third harmonic spaces received by the faulty PMSM can be obtained.

6. The control method according to claim 1, characterized in that, The specific process in step 5 is as follows: Step 5.1: If an open circuit fault occurs in phase A of the motor, the voltage source inverter does not know whether an open circuit fault has occurred in phase A of the motor, and still sends the same medium and large base voltage vectors as under normal motor conditions to form a virtual voltage vector; since the base voltage vectors of each virtual voltage vector in the α-β and xy spaces remain unchanged before and after the open circuit fault in phase A. Step 5.2: Following the method in Step 5.1, construct the virtual vector for one of the remaining four phases under an open-circuit fault condition; since the base voltage vector received by the fault PMSM will shift under fault conditions, the obtained... and xy The spatial virtual voltage vector distribution also shifts.

7. The control method according to claim 1, characterized in that, Step 7 further includes: establishing on MT The DBDTC in the coordinate system is discretized to achieve decoupled control of stator flux linkage and torque, thereby reducing the computational complexity of the system; and the disturbance components observed by the disturbance observer are... d 1. As a feedforward compensation, it is directly introduced into the simplified DBDTC to suppress torque pulsation caused by open circuit faults and the third harmonic component of back EMF.

8. The control method according to claim 1, characterized in that, In step 8, based on Ten virtual voltage vectors V constructed in the subspace 3 The spatial distribution of adjacent V 3 The plane is divided into 10 sectors using the angle bisector as the boundary; by determining the sector where the stator voltage reference vector is located, the corresponding V for that sector is selected. 3 As the optimal virtual voltage vector; and using the geometric projection method to calculate the optimal V 3 The duty cycle, specifically, is the projection of the stator voltage reference vector onto the selected optimal V. 3 The direction is determined by the duty cycle based on the projection relationship.

9. The control method according to claim 1, characterized in that, In step 9, based on xy The spatial distribution characteristics of the subspace virtual voltage vector under normal and fault conditions of the motor, the preset harmonic subspace virtual voltage vector selection table, and the determination of harmonic current vectors I xy ( k Select the virtual voltage vector V corresponding to the diagonal quadrant of the quadrant where the ) is located. 3' Harmonic suppression is performed to achieve closed-loop control of harmonic currents.

Images