A model predictive fault-tolerant control method for dual three-phase permanent magnet synchronous motor considering switching delay adaptation

Abstract

The application discloses a kind of double three-phase permanent magnet synchronous motor model prediction fault-tolerant control methods considering delay adaptation, the subspace after single-phase open circuit fault of motor is decoupled by order reduction decoupling matrix, and the stator voltage equation of motor is established;Select the virtual voltage vector of adjacent three voltage vectors on fundamental subspace synthesis zero component on harmonic subspace, based on stator voltage equation, select the adjacent two virtual voltage vectors of reference voltage vector, realize full area tracking to reference voltage vector by duty cycle modulation;For the delay period from fault occurrence to fault-tolerant algorithm switching, the disturbance model is established on dq axis to open-circuit voltage residual error, and the generalized proportional integral observer containing disturbance influence is directly established, and the disturbance is observed in real time;In addition, the delay time is estimated based on high-frequency current distortion energy, and the observer gain is dynamically adjusted.The application significantly reduces the harmonic content under fault-tolerant steady state, and reduces the torque ripple amplitude of fault transient.

Description

Technical Field

[0001] This invention relates to the field of fault-tolerant control of dual three-phase permanent magnet synchronous motors, specifically a model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors (DTP-PMSM) that considers adaptive switching delay. This method helps to reduce the harmonic content in fault-tolerant steady state, reduce the torque ripple amplitude during switching delay, and improve the smoothness of motor operation. Background Technology

[0002] Dual three-phase permanent magnet synchronous motors (PMSMs), with their high torque density, large power rating, and superior fault tolerance, have become a core choice for high-performance drive systems in demanding reliability applications such as ship electric propulsion, aerospace, and new energy vehicles. However, the increased number of phases also increases the probability of failure during actual operation. Theoretically, switching transistor faults, short-circuit faults, and open-circuit faults in dual three-phase PMSMs can be converted into phase loss faults through hardware isolation. Therefore, scholars both domestically and internationally have conducted in-depth research on phase loss fault-tolerant control strategies for dual three-phase PMSMs.

[0003] Research on phase-loss fault-tolerant control for DTP-PMSM mainly focuses on harmonic suppression and torque output stability. Model predictive control utilizes the voltage vectors acting in the fundamental and harmonic subspaces of a six-phase two-level inverter, employing vector synthesis techniques to significantly reduce harmonic content. When a motor fault occurs, the fundamental and harmonics couple, causing changes in the subspace voltage vector distribution. Therefore, a model predictive fault-tolerant control method suitable for motor faults needs to be designed. Furthermore, during the transient fault occurrence, there is an inherent delay between diagnostic location and fault-tolerant algorithm activation. During this stage, the missing voltage of the faulty phase generates a continuous voltage residual, leading to torque ripple and current distortion. Existing technologies do not design a global compensation scheme for this delay period, and the operational stability of the motor during fault transients remains unresolved. Summary of the Invention

[0004] To address the issues of high harmonic content and large torque ripple during the fault-tolerant switching delay in existing dual three-phase permanent magnet synchronous motors after an open-circuit fault, this invention provides a model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors that considers adaptive switching delay.

[0005] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0006] A model predictive fault-tolerant control method for a dual three-phase permanent magnet synchronous motor considering adaptive switching delay includes:

[0007] The fundamental and harmonic subspaces of a dual three-phase permanent magnet synchronous motor after a single-phase open-circuit fault are decoupled using a reduced-order decoupling matrix, and the stator voltage equation of the motor after decoupling is established.

[0008] Three adjacent voltage vectors in the fundamental subspace are synthesized into a virtual voltage vector with zero components in the harmonic subspace. Based on the stator voltage equation, two virtual voltage vectors adjacent to the reference voltage vector are selected, and voltage tracking of the reference voltage vector is achieved in the whole region through duty cycle modulation.

[0009] During the delay between the occurrence of a single-phase open-circuit fault and the switching of the motor fault-tolerant algorithm, a disturbance model is established for the open-circuit voltage residual on the dq axis. The fault location information is skipped, and a generalized proportional-integral observer containing the disturbance effect is directly established to observe the disturbance effect of the open-circuit voltage residual on the dq axis in real time.

[0010] Further technical solutions include: designing an online estimation method based on high-frequency current distortion energy to estimate different degrees of delay time, and dynamically adjusting the observer gain using the delay time to achieve delay-adaptive observation.

[0011] A further technical solution, based on the stator voltage equation, selects two virtual voltage vectors adjacent to the reference voltage vector, and achieves continuous voltage tracking of the reference voltage vector across the entire region through duty cycle modulation. Specifically:

[0012] According to the dq-axis stator voltage equation, let s d =d i d / dt、 s q =d i q / dt, we obtain the rate of change of current under the action of each virtual voltage vector:

[0013]

[0014] in, s d0 The slope of the d-axis current corresponding to the zero voltage vector. s q0 The slope of the q-axis current corresponding to the zero voltage vector. u di Represents virtual voltage vector u i Components along the d-axis, u dj Represents virtual voltage vector u j Components along the d-axis, u qi Represents virtual voltage vector u i Components on the q-axis, u qj Represents virtual voltage vector u jComponents on the q-axis, s di for u di The corresponding current slope, s dj for u dj The corresponding current slope, s qi for u qi The corresponding current slope, s qj for u qj The corresponding current slope, L d and L q These are the d-axis and q-axis inductances, respectively. i d and i q These are the d-axis and q-axis currents, respectively. R s For stator resistance, θ e The motor position angle, ψ f This represents the amplitude of the permanent magnet flux linkage. Electric angular velocity, virtual voltage vector u i , u j For adjacent virtual voltage vectors;

[0015] Combining the discretized dq-axis stator voltage equations, the current value at the next moment can be expressed as:

[0016]

[0017] in, T i Virtual voltage vector u i Duration of action T j Virtual voltage vector u j Duration of action T 0 is the zero vector u The duration of action of 0, This represents the current d-axis current value at the current moment. This represents the current q-axis current value at the current moment. T s The sampling period;

[0018] Let the current value at the next moment be equal to the current reference value: id k+1 = i d ref , i q k+1 = i q ref Based on the principle of maximum torque, let i d ref =0, i q ref The speed is obtained by tracking the reference speed from the outer speed loop, then:

[0019]

[0020] In the formula, intermediate quantities ;

[0021] Depend on T i , T j , T s The synthesized voltage vector is obtained. .

[0022] A further technical solution is that the generalized proportional-integral observer incorporating the effects of disturbances is:

[0023]

[0024] in, x 1d d-axis current i d The first-order differential estimate, x 2d for u d / L d The first-order differential estimate, x 3d This is the first-order differential estimate of the d-axis perturbation. x 1q q-axis current i q The first-order differential estimate, x 2q for u q / L q The first-order differential estimate, x 3qLet be the first-order differential estimate of the q-axis perturbation, and be the estimation error of the d-axis current. e 1d = i d - x 1d The estimation error of the q-axis current is e 1q = i q - x 1q , 1. 2. 3 represents the observer gain. u dn and u qn The normal voltages of the d and q axes calculated by the controller. u d The effect of open-circuit voltage residual on the d-axis, u q This represents the effect of the open-circuit voltage residual on the q-axis.

[0025] A further technical solution yields the following estimated voltage residual value: .

[0026] A further technical solution, based on online estimation of delay time using high-frequency current distortion energy, is as follows:

[0027] The instantaneous sum of squares of the current tracking error is defined as:

[0028]

[0029] In the formula, i d (t) represents the real-time sampled value of the d-axis current. i q (t) represents the real-time sampled value of the q-axis current. i d ref This is the reference value for the d-axis current. i q ref This is the reference value for the q-axis current.

[0030] Calculate the integral energy within the sliding window:

[0031]

[0032] In the formula, T w The length of the sliding window;

[0033] After a single-phase open-circuit fault occurs, the voltage residual causes the current tracking error to increase rapidly. E (t) jumps above the threshold E th ,when E (t) When the threshold is exceeded for the first time, record the current time as the time of the fault occurrence. t fault_oneset Then the delay time .

[0034] A further technical solution involves dynamically adjusting the observer gain using the aforementioned delay time, specifically as follows:

[0035] Observer gain and bandwidth The relationship is:

[0036]

[0037] bandwidth satisfy:

[0038] .

[0039] Further technical solutions, to avoid estimated jitter caused by sudden gain changes, include bandwidth... Perform first-order low-pass filtering:

[0040]

[0041] in, This represents the bandwidth after a first-order low-pass filter.

[0042] The beneficial effects of this invention are as follows:

[0043] (1) This invention addresses the issue of a single-phase open-circuit fault in a dual three-phase permanent magnet synchronous motor by establishing a reduced-order decoupled mathematical model of the motor and synthesizing a virtual voltage vector with zero harmonic components in the harmonic subspace using three adjacent voltage vectors in the αβ subspace. Furthermore, the phase of the reference voltage is tracked by combining two adjacent virtual voltage vectors of the reference voltage with the zero vector, thereby achieving continuous vector synthesis across the entire region and reducing the harmonic content during fault-tolerant steady-state operation of the motor.

[0044] (2) The present invention takes into account the inherent delay in the switching of the fault-tolerant algorithm caused by fault diagnosis, sampling lag and other reasons. During this period, the torque pulsation caused by voltage residual is estimated in real time by using the GPI observer. No fault location information is required in the process, which effectively covers the fault-tolerant switching delay window caused by diagnosis and location, and ensures the continuity and stability of motor operation.

[0045] (3) This invention designs an online delay time estimation method based on high-frequency current distortion energy for different degrees of delay time. The gain of the DA-GPI observer is dynamically adjusted according to the estimated delay time to achieve delay adaptive observation and improve the stability of the motor's global operation process. Attached Figure Description

[0046] Figure 1 This is a driving topology diagram for an open-circuit fault in phase F of the present invention;

[0047] Figure 2(a) is a phase diagram of the voltage vector distribution of the inverter αβ subspace under the fault state of phase F of the present invention;

[0048] Figure 2(b) is a phase diagram of the voltage vector distribution of the inverter z-subspace under the fault state of phase F of the present invention;

[0049] Figure 3 A virtual voltage vector diagram with constant amplitude was designed for this invention;

[0050] Figure 4 This is a voltage residual diagram of phase F under different operating conditions of the present invention;

[0051] Figure 5(a) shows the fault-tolerant steady-state phase current waveform obtained based on the traditional fault-tolerant method;

[0052] Figure 5(b) shows the fault-tolerant steady-state phase current waveform obtained based on the improved model prediction fault-tolerant method;

[0053] Figure 6(a) shows the torque and phase current waveforms without considering the adaptive switching delay.

[0054] Figure 6(b) shows the torque and phase current waveforms considering adaptive switching delay.

[0055] Figure 7 This is a flowchart of the control strategy of the present invention. Detailed Implementation

[0056] The present invention will now be described in detail with reference to the accompanying drawings and specific examples, but the scope of protection of the present invention is not limited thereto.

[0057] This invention is a model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors considering adaptive switching delay. First, a virtual voltage vector is constructed using three adjacent voltage vectors to reduce harmonic effects. Further, phase tracking of the reference voltage is performed using two adjacent virtual voltage vectors, and amplitude tracking is performed using a zero vector. The duty cycle is rationally allocated to achieve continuous vector synthesis across the entire region. Second, considering the large torque ripple amplitude during the inherent delay of the fault-tolerant algorithm switching, a GPI observer is used to compensate for this disturbance, skipping the fault diagnosis process and achieving full-process compensation. Finally, considering the poor observation stability of fixed gain for long delays, an online delay time estimation method based on high-frequency current distortion energy is proposed. The gain of the DA-GPI observer is dynamically adjusted to ensure the stability of the observer under different delay times.

[0058] like Figure 7 As shown, the present invention specifically adopts the following technical solution:

[0059] 1. The subspace of a dual three-phase permanent magnet synchronous motor after a single-phase open circuit is decoupled by a reduced-order decoupling matrix, and the mathematical model of the motor after the fault is derived.

[0060] When a single-phase open-circuit fault occurs in a motor, the reduced degrees of control freedom mean that continuing to use the decoupling model under normal operating conditions will cause controller conflicts. For the post-fault structure, a reduced-order decoupling matrix can achieve complete decoupling of the fundamental and harmonic subspaces. Taking an open-circuit F-phase motor as an example... Figure 1 The corresponding reduced-order decoupling matrix is:

[0061] (1)

[0062] The voltage component in the natural coordinate system is transformed to the fundamental subspace in the stationary coordinate system by using a reduced-order decoupling matrix. α β Harmonic subspace z.

[0063] Furthermore, by transforming the fundamental component related to electromechanical energy into the rotating coordinate system dq through the Park transformation, the stator voltage of the DTP-PMSM when phase F is open can be obtained as follows:

[0064] The stator voltage of the dq axis is:

[0065] (2)

[0066] Among them, matrix , θ e The motor position angle, u d The stator voltage is the d-axis voltage. u qThese are the stator voltages along the q-axis. R s For stator resistance, L d and L q These are the d-axis and q-axis inductances, respectively. i d and i q These are the d-axis and q-axis currents, respectively. ψ f This represents the amplitude of the permanent magnet flux linkage. It represents the electric angular velocity.

[0067] Due to the reduction in degrees of freedom, the harmonic subspace is reduced from two dimensions to one dimension, denoted as the z-axis; the stator voltage along the z-axis is:

[0068] (3)

[0069] In the formula, i z Let the current be along the z-axis. u z The voltage along the z-axis. L z This refers to the leakage inductance of the motor.

[0070] The voltage equation of the dq axis (i.e., equation (2)) is discretized using the forward Euler method to obtain the current prediction model for the next time step:

[0071] (4)

[0072] In the formula, T s The sampling period is This is the predicted value of the d-axis current at the next moment. This is the predicted value of the q-axis current at the next moment. This represents the current d-axis current value at the current moment. This represents the current q-axis current value at the current moment. This represents the current d-axis voltage value. This represents the q-axis voltage value at the current moment.

[0073] 2. Construct virtual voltage vectors and use two virtual voltage vectors adjacent to the reference voltage to combine with the zero vector to achieve continuous vector synthesis across the entire region;

[0074] The relationship between the voltage vector distribution in the fundamental and harmonic subspaces of the stationary coordinate system and the switching states of the inverter can be expressed as:

[0075] (5)

[0076] in, uα The voltage vector component along the α-axis. u β In order to be in β Voltage vector components of the axis, U dc This is the DC bus voltage; S A This refers to the A-phase switch state of the inverter. S B This refers to the B-phase switch state of the inverter. S C This refers to the C-phase switch state of the inverter. S D This refers to the D-phase switch status of the inverter. S E This is the E-phase switch state of the inverter, and S A , S B , S C , S D , S E The value can be 0 or 1.

[0077] The voltage vector distribution in each subspace is shown in Figures 2(a) and (b). To reduce harmonic content, a virtual voltage vector with zero component in the harmonic subspace is synthesized from three adjacent voltage vectors in the fundamental subspace. Figure 3 of vv For example, the virtual voltage vector is derived from the voltage vector in Figure 2(a). v 18 , v 26 and v 27 The duration of each voltage vector is calculated using the following formula:

[0078] (6)

[0079] In the formula, t 1 is v 18 Duration of action t 2 is v 26 Duration of action t 3 is v 27 Duration of action; t 0 is the zero vector v 00 The duration of action is used to process the virtual voltage vector to the same amplitude. V objFor the target virtual voltage vector; superscript " α , β "Indicates the corresponding voltage vector in α , β The components of the axis; , , Voltage vector v 18 exist α , β z-axis components , , Voltage vector v 26 exist α , β z-axis components , , Voltage vector v 27 exist α , β z-axis components , , Voltage vector v 00 exist α , β The z-axis component.

[0080] Solving the system of equations in formula (6), the duration of each voltage vector is obtained as follows:

[0081] (7)

[0082] By utilizing the principle of vector composition, the voltage vector is adjusted using the duty cycle of each voltage vector's action time. v 18 , v 26 and v 27 The amplitude is used to obtain the virtual voltage vector. vv 1.

[0083] like Figure 3 The diagram shows a total of 12 virtual voltage vectors with equal amplitudes and phase differences of 30°. The amplitude of each virtual voltage vector is 0.295. U dc .

[0084] In order to make the controller output voltage ( u d , u qBy tracking the reference voltage and selecting two virtual voltage vectors adjacent to the reference voltage and combining them with the zero vector, a continuous voltage vector synthesis with adjustable amplitude and phase can be achieved across the entire region.

[0085] According to the stator voltage equation (equation (2)), let s d =d i d / dt、 s q =d i q / dt, we obtain the rate of change of current under the action of each virtual voltage vector:

[0086] (8)

[0087] In the formula, s d0 The slope of the d-axis current corresponding to the zero voltage vector. s q0 The slope of the q-axis current corresponding to the zero voltage vector. u di Represents virtual voltage vector u i Components along the d-axis, u dj Represents virtual voltage vector u j Components along the d-axis, u qi Represents virtual voltage vector u i Components on the q-axis, u qj Represents virtual voltage vector u j Components on the q-axis, s di for u di The corresponding current slope, s dj for u dj The corresponding current slope, s qi for u qi The corresponding current slope, s qj for u qj Corresponding current slope; virtual voltage vector u i , u j These are adjacent virtual voltage vectors.

[0088] Based on formulas (4) and (8), the current value at the next moment is expressed as:

[0089] (9)

[0090] In the formula, T i Virtual voltage vector u i Duration of action T j Virtual voltage vector u j Duration of action T 0 is the zero vector u The duration of action of 0, T i , T j , T The numerical calculation of 0 uses the deadbeat principle, setting the current value at the next moment to be equal to the current reference value: i d k+1 = i d ref , i q k+1 = i q ref Based on the principle of maximum torque, set i d ref =0, i q ref The outer speed loop tracks the reference speed. If obtained, then:

[0091] (10)

[0092] In the formula, intermediate quantities .

[0093] Depend on T i , T j , T s The synthesized voltage vector is obtained as follows: .

[0094] The synthesized voltage vector is supplied to the dual three-phase permanent magnet synchronous motor by the inverter.

[0095] Third, when a single-phase open circuit occurs in a dual three-phase permanent magnet synchronous motor, there is an inherent delay in the switching process of the fault-tolerant algorithm. To address the torque ripple problem caused by voltage residuals during this process, this invention proposes a Generalized Proportional Integral (GPI) observer that can estimate the voltage residuals in real time during the fault transient, thereby reducing the impact of voltage residuals on the smoothness of motor operation during the delay period.

[0096] When a single-phase open-circuit fault occurs in a motor, not only will the fundamental current and harmonic current couple, causing current constraint, but the open circuit in the inverter will also prevent the control voltage output by the controller of the faulty phase from being accurately transmitted to the motor terminal, resulting in voltage residual and thus affecting the control effect. Figure 1 The output voltage of the F-phase controller is shown to be... The actual voltage received by the motor terminal is u F The difference between the two is the voltage residual, expressed as .

[0097] Since the fault-tolerant algorithm has not yet switched during the switching delay, the reduced-order decoupling matrix is ​​not yet used, and the normal decoupling matrix is ​​still used for the analysis of voltage residuals. The phase voltages in the natural coordinate system are... u A , u B , u C , u D , u E , u F Mapped to via VSD transformation αβ and z 1 z 2 subspaces, the corresponding expression is:

[0098] (11)

[0099] When only phase F is open-circuited, the voltages of the other five phases are unaffected; therefore, only the phase voltages are affected. u β and u z2 Influenced by voltage residual. To further demonstrate the impact of voltage residual on control performance, the open-circuit voltage residual of phase F can be expressed as:

[0100] (12)

[0101] In the formula, I q This represents the DC component of the q-axis current.i B This is the B-phase current. i C This refers to the C-phase current. The voltage residual of the F-phase affects the entire distributed system ( αβ and z 1 z 2 subspaces can be represented as:

[0102] (13)

[0103] In the formula, u α The effect of the open-circuit voltage residual of phase F on the α axis. u β The effect of the open-circuit voltage residual of phase F on the β axis. u z1 The effect of the open-circuit voltage residual of phase F on the z1 axis. u z2 This is to explain the effect of the open-circuit voltage residual of phase F on the z2 axis.

[0104] Since the electromechanical energy conversion of the motor system only occurs in the fundamental subspace, then for αβ Performing an inverse Park transform on the components yields the effect of the voltage residual on the dq plane:

[0105] (14)

[0106] In the formula, u d This indicates the effect of the open-circuit voltage residual of phase F on the d-axis. u q This indicates the effect of the open-circuit voltage residual of phase F on the q-axis. As can be seen from equation (12), when the effect of the voltage residual is ignored, it will cause voltage distortion in the fundamental subspace, generating second harmonics and second torque pulsation, which in turn affects the control effect of the system.

[0107] u d , u q The general form can be represented as the fault phase identifier and motor position angle. θ e Functions related to current state:

[0108] (15)

[0109] u d , u q Directly superimposed on the normal voltage calculated by the controller u dn , u qn The above results in the actual voltage applied to the motor being... u dn + u d , u qn + u q This is the direct cause of current distortion and torque pulsation after a fault. Traditional methods require identifying the faulty phase to calculate... u d , u q The process inevitably causes fault-tolerant switching delays and introduces torque ripples. The present invention aims to estimate the total value in real time through an observer without relying on fault location information.

[0110] Based on the stator voltage equation under normal conditions, the discrete model of the motor considering the influence of voltage residuals can be expressed as:

[0111] (16)

[0112] In motor control, traditional extended state observers or disturbance observers based on first-order models often assume that the disturbance is slowly varying. However, at the instant a single-phase open-circuit fault occurs, the voltage residual... u d , u q It manifests as a step-like abrupt change, with its derivative theoretically being infinite at the moment of the fault, followed by a complex dynamic process over a period of time, rather than a conventional slowly varying signal.

[0113] The Generalized Proportional-Integral (GPI) observer constructs a higher-order disturbance estimation model by introducing an integral element for the disturbance estimation error. For step or ramp disturbances, the GPI observer can achieve zero steady-state error tracking, and its dynamic response speed can be adjusted by pole placement. This invention designs a GPI observer using formula (16) to observe voltage residuals, and its structure is as follows:

[0114] (17)

[0115] In the formula, x 1d d-axis current i d The first-order differential estimate, x 2d for u d / L d The first-order differential estimate of (the ratio of d-axis voltage residual to d-axis inductance). x 3d This is the first-order differential estimate of the d-axis perturbation. x 1q q-axis current i q The first-order differential estimate, x 2q for u q / L q The first-order differential estimate of the ratio of q-axis voltage residual to q-axis inductance. x 3q The first-order differential estimate of the q-axis perturbation is given; the estimation error of the d-axis current is... e 1d = i d - x 1d The estimation error of the q-axis current is e 1q = i q - x 1q ; 1. 2. 3 represents the observer gain. The final estimate of the voltage residual can be expressed as:

[0116] (18)

[0117] This observer relies solely on measurable current ( i d , i q ), the voltage output by the controller ( u dn , u qn and speed information It eliminates the need for fault location information and enables global online estimation of disturbance voltage.

[0118] IV. The above-mentioned traditional GPI observer uses a fixed gain. 1. 2. 3. When the delay time t d For shorter time intervals, a fixed gain can balance speed and stability; however, for longer delays, such as... t d When the voltage residual exceeds 50ms, the persistent voltage residual can cause noise amplification, phase lag, and even integral drift in high-gain observers, while low-gain observers cannot quickly track fault transients. To address this, this invention proposes a Delay Adaptive Generalized Proportional-Integral (DA-GPI) observer. By estimating the delay time online based on the energy of the high-frequency current distortion, the observer gain is dynamically adjusted, ensuring stable operation at different delay times.

[0119] To achieve adaptive adjustment of the observer gain, it is first necessary to estimate online the time length from the occurrence of the fault to the current moment, i.e., the delay time. This invention uses a method based on high-frequency current distortion energy for estimation.

[0120] The instantaneous sum of squares of the current tracking error is defined as:

[0121] (19)

[0122] In the formula, i d This represents the real-time sampled value of the d-axis current. i q This represents the real-time sampled value of the q-axis current. i d ref This is the reference value for the d-axis current. i q ref This is the reference value for the q-axis current.

[0123] To eliminate random noise and highlight distortion caused by faults, calculate the integral energy within a sliding window:

[0124] (20)

[0125] In the formula, the length of the sliding window is... T w Take 2ms. During normal operation, E (t) approaches 0; after the fault occurs, the voltage residual causes the current tracking error to increase rapidly. E (t) jumps above the threshold Eth In this embodiment, a threshold value is used. , A Represents Ampere, s It represents seconds.

[0126] when E (t) When the threshold is exceeded for the first time, record the current time as the time of the fault occurrence. t fault_oneset The delay time is:

[0127] (twenty one)

[0128] In subsequent control cycles, It grows linearly until it is reset after the fault-tolerant algorithm is switched.

[0129] The dynamic response of the DA-GPI observer is determined by its bandwidth. Decision. For a third-order GPI observer, the classic pole placement gives the following relationship between observer gain and bandwidth:

[0130] (twenty two)

[0131] In the initial stage of a fault, high gain is beneficial for quickly tracking step disturbances, but prolonged high gain amplifies current sampling noise, resulting in severe jitter in the estimated value. Low gain offers better operational stability but is less capable of tracking dynamic changes in the voltage residual. Therefore, this invention assigns segmented values ​​to the observer bandwidth, enabling... With the estimated delay time Increase and decrease:

[0132] (twenty three)

[0133] To avoid estimation jitter caused by sudden gain changes, for Perform first-order low-pass filtering:

[0134] (twenty four)

[0135] in, This represents the bandwidth after a first-order low-pass filter.

[0136] In each control cycle, the DA-GPI observer adjusts its parameters according to the current data. Select the bandwidth after first-order low-pass filtering, calculate the observer gain after first-order low-pass filtering, and update the observer in real time.

[0137] Finally, the motor discrete model considering the voltage residual (i.e., equation (16)) is discretized using forward Euler discretization:

[0138] (25)

[0139] In the formula, i d (k+1) is the predicted current value of the d-axis in the (k+1)th period. i q (k+1) is the predicted current value of the q-axis in the (k+1)th period. i d (k) represents the current sample value of the d-axis in the kth period. i q (k) represents the current sample value of the q-axis in the kth period. u d (k) represents the voltage sample value of the d-axis in the kth period. u q (k) represents the voltage sample value of the q-axis in the k-th period. u dn (k) represents the voltage residual sample value on the q-axis in the k-th period. u qn (k) represents the voltage residual sample value of the q-axis in the k-th period.

[0140] According to the no-difference-time theory, let i d (k+1) =i d ref , i q (k+1) = i q ref The optimal voltage reference after compensation is obtained from formula (25):

[0141] (26)

[0142] In the formula, u d com (k) is the optimal d-axis reference voltage command that is finally applied to the inverter after compensation. u q com (k) is the optimal reference voltage command for the q-axis of the inverter after compensation.

[0143] To verify the effectiveness of the proposed model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors that considers adaptive switching delay, a simulation model of the dual three-phase permanent magnet synchronous motor was built in the MATLAB / Simulink environment. The rated speed was set to 1200 r / min and the rated torque to 10 N·m. The process of an open-circuit fault occurring in phase F and switching to the proposed fault-tolerant control strategy was simulated.

[0144] Figure 4 This shows the voltage residual of phase F under different operating conditions. u F The voltage residual is close to zero before the fault occurs; at the moment of the fault, the voltage residual exhibits a step jump and persists during the fault-tolerant algorithm switching delay. After observer compensation, the voltage residual is significantly reduced, and the voltage residual is further reduced after the fault-tolerant algorithm switches.

[0145] Figure 5(a) shows the steady-state phase current waveform obtained using the traditional fault-tolerant control method (without virtual voltage vector synthesis). It is evident that the current exhibits significant distortion, poor sinusoidal characteristics, and high harmonic content. Figure 5(b) shows the steady-state phase current waveform obtained using the method proposed in this invention. The waveform is smooth, exhibits good sinusoidal characteristics, and harmonic distortion is significantly suppressed. This fully demonstrates the superiority of this invention in improving steady-state performance.

[0146] Figure 6(a) shows the torque and phase current waveforms without considering the switching delay (i.e., without the introduction of the DA-GPI observer). At the moment of the fault and within the subsequent delay window, the torque exhibits severe pulsation and a large drop, with the pulsation amplitude reaching 3.4 N·m, and the phase current is severely distorted.

[0147] Figure 6(b) shows the torque and phase current waveforms under the complete scheme of the present invention (including the DA-GPI observer). Comparing with Figure 6(a), it can be seen that after adopting the present invention, the torque ripple amplitude during the delay period is significantly reduced, the ripple amplitude is reduced to 0.95 N·m, and the current distortion is also greatly improved.

[0148] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A model predictive fault-tolerant control method for a dual three-phase permanent magnet synchronous motor considering adaptive switching delay, characterized in that, include: The fundamental and harmonic subspaces of a dual three-phase permanent magnet synchronous motor after a single-phase open-circuit fault are decoupled using a reduced-order decoupling matrix, and the stator voltage equation of the motor after decoupling is established. Three adjacent voltage vectors in the fundamental subspace are synthesized into a virtual voltage vector with zero components in the harmonic subspace. Based on the stator voltage equation, two virtual voltage vectors adjacent to the reference voltage vector are selected, and voltage tracking of the reference voltage vector is achieved in the whole region through duty cycle modulation. During the delay between the occurrence of a single-phase open-circuit fault and the switching of the motor fault-tolerant algorithm, a disturbance model is established for the open-circuit voltage residual on the dq axis. The fault location information is skipped, and a generalized proportional-integral observer containing the disturbance effect is directly established to observe the disturbance effect of the open-circuit voltage residual on the dq axis in real time.

2. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 1, characterized in that, Also includes: For different degrees of delay time, an online estimation method based on high-frequency current distortion energy is designed for estimation. The observer gain is dynamically adjusted using the delay time to achieve delay-adaptive observation.

3. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 2, characterized in that, Based on the stator voltage equation, two virtual voltage vectors adjacent to the reference voltage vector are selected, and continuous voltage tracking of the reference voltage vector over the entire region is achieved through duty cycle modulation, specifically: According to the dq-axis stator voltage equation, let s d =di d / dt、s q =di q / dt, we obtain the rate of change of current under the action of each virtual voltage vector: Among them, s d0 The slope of the d-axis current corresponding to the zero voltage vector, s q0 The slope of the q-axis current corresponding to the zero voltage vector, u di Represents the virtual voltage vector u i The component along the d-axis, u dj Represents the virtual voltage vector u j The component along the d-axis, u qi Represents the virtual voltage vector u i The component on the q-axis, u qj Represents the virtual voltage vector u j The component on the q-axis, s di For u di The corresponding current slope, s dj For u dj The corresponding current slope, s qi For u qi The corresponding current slope, s qj For u qj The corresponding current slope, L d and L q These are the d-axis and q-axis inductances, respectively, and i d and i q These are the d-axis and q-axis currents, respectively, R s For the stator resistance, θ e Let ψ be the motor position angle. f This represents the amplitude of the permanent magnet flux linkage. Electric angular velocity, virtual voltage vector u i u j For adjacent virtual voltage vectors; Combining the discretized dq-axis stator voltage equations, the current value at the next moment can be expressed as: Among them, T i For virtual voltage vector u i Duration of action, T j For virtual voltage vector u j The duration of action, T0 is the duration of action of the zero vector u0. This represents the current d-axis current value at the current moment. T represents the q-axis current value at the current moment. s The sampling period; Let the current value at the next moment be equal to the current reference value: i d k+1 =i d ref i q k+1 =i q ref Based on the principle of maximum torque, let i d ref =0, i q ref The speed is obtained by tracking the reference speed from the outer speed loop, then: In the formula, intermediate quantities ; By T i T j T s The synthesized voltage vector is obtained. .

4. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 3, characterized in that, The generalized proportional-integral observer that incorporates the effects of disturbances is: Where, x 1d For d-axis current i d The first-order differential estimate, x 2d for u d / L d The first-order differential estimate, x 3d x is the first-order differential estimate of the d-axis perturbation. 1q For the q-axis current i q The first-order differential estimate, x 2q for u q / L q The first-order differential estimate, x 3q Let e ​​be the first-order differential estimate of the q-axis perturbation, and e be the estimation error of the d-axis current. 1d =i d -x 1d The estimation error of the q-axis current is e 1q =i q -x 1q , 1.

2. 3 represents the observer gain, u dn and u qn The normal voltages of the d and q axes calculated by the controller. u d The effect of open-circuit voltage residual on the d-axis, u q This represents the effect of the open-circuit voltage residual on the q-axis.

5. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 4, characterized in that, The estimated value of the voltage residual is: .

6. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 3, characterized in that, The online estimation of the delay time based on the high-frequency current distortion energy is as follows: The instantaneous sum of squares of the current tracking error is defined as: In the formula, i d (t) represents the real-time sampled value of the d-axis current, i q (t) represents the real-time sampled value of the q-axis current, i d ref i is the reference value for the d-axis current. q ref This is the reference value for the q-axis current. Calculate the integral energy within the sliding window: In the formula, T w The length of the sliding window; After a single-phase open-circuit fault occurs, the voltage residual causes the current tracking error to increase rapidly, and E(t) jumps above the threshold E. th When E(t) first exceeds the threshold, the current time is recorded as the fault occurrence time t. fault_oneset Then the delay time .

7. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 6, characterized in that, The observer gain is dynamically adjusted using the aforementioned delay time, specifically as follows: Observer gain and bandwidth The relationship is: bandwidth satisfy: 。 8. The model prediction fault-tolerant control method for dual three-phase permanent magnet synchronous motors according to claim 7, characterized in that, To avoid estimated jitter caused by sudden gain changes, bandwidth... Perform first-order low-pass filtering: in, (k) represents the bandwidth after first-order low-pass filtering.

Images