A method for diagnosing open-circuit faults in three-level inverters based on predicted current residuals

Abstract

This invention discloses a method for diagnosing open-circuit faults in three-level inverters based on predicted current residuals. Specifically, it involves: establishing a current prediction model for the three-level inverter feed and determining the cost function; determining the inverter output voltage vector corresponding to the minimum cost function, inputting the corresponding switching state into the inverter, and predicting the stator current of the permanent magnet motor at the next moment; comparing the stator current of the permanent magnet motor with the predicted stator current to obtain the predicted current residual; and judging whether an open-circuit fault has occurred in the inverter by comparing the predicted current residual with a fault detection threshold; when an open-circuit fault is detected, identifying the switch corresponding to the minimum predicted current residual, and determining the faulty switch by comparing the number of elements in the faulty switch counting matrix with the counting threshold. This invention offers fast diagnostic speed, high accuracy, and requires no additional auxiliary hardware circuitry; it also exhibits excellent robustness and dynamism, with flexible fault detection and fault location threshold settings, adapting to complex operating conditions.

Description

Technical Field

[0001] This invention belongs to the field of inverter fault diagnosis, specifically relating to a method for diagnosing open-circuit faults in three-level inverters based on predicted current residuals. Background Technology

[0002] The traction inverter is the core power source of a train's traction drive system, and its control performance and reliability are crucial for the safe and efficient operation of high-speed trains. Currently, the Japanese Shinkansen and my country's CRH2 and CRH380AL high-speed trains use a three-level diode-neutral-point-clamped (NPC) topology. However, the three-level NPC topology suffers from uneven losses and the resulting junction temperature imbalance, limiting the inverter's switching frequency and output current. In 2001, T. Bruckner et al. proposed using active power switching devices to replace the clamping diodes in the NPC topology, thus giving rise to the three-level active neutral-point-clamped (ANPC) inverter topology.

[0003] The main advantages of the ANPC inverter topology are its numerous redundant switching states, its output voltage exhibiting the same harmonic characteristics as the NPC, and the addition of a zero-level output mode and commutation path, providing new degrees of freedom for balancing the loss distribution of power devices. As the number of power devices in a three-level traction inverter increases, the probability of failure also increases, with open-circuit faults in power devices being one of the most common types of inverter failures. Currently, extensive research has been conducted both domestically and internationally on open-circuit fault diagnosis in three-level inverters. Existing research methods are mainly divided into analytical model-based methods, signal processing-based methods, and data-driven methods. Analytical model-based methods have strong universality but require accurate model establishment and are highly sensitive to parameters; signal processing-based methods have weak model dependence but poor universality and complex data calculations; data-driven methods involve large computational loads, require prior database establishment, and have high hardware requirements. All three traditional methods currently applied to inverter open-circuit fault diagnosis have certain limitations, and the numerous redundant switching states in ANPC inverters further increase the difficulty of diagnosing open-circuit faults in power devices, often requiring auxiliary hardware circuits, which increases system cost and complexity. Summary of the Invention

[0004] To address the difficulty in locating open-circuit faults in existing three-level ANPC inverters, this invention provides a method for diagnosing open-circuit faults in three-level inverters based on predicted current residuals, leveraging the advantages of model predictive control's multi-objective rolling optimization, good dynamic performance, and strong robustness.

[0005] The present invention provides a method for diagnosing open-circuit faults in a three-level inverter based on predicted current residuals, comprising the following steps:

[0006] S1. Establish a current prediction model for a permanent magnet motor drive system powered by a three-level inverter, and determine the cost function J for finite set model prediction of current control based on the current prediction model and the control objective.

[0007] S2. Number the voltage vector output by the three-level inverter, based on the permanent magnet motor stator current i at time k. d (k), i q (k) and the three-level inverter capacitor voltage u c1 (k), u c2 (k) Obtain the minimum cost function J min The corresponding inverter output voltage vector, and the corresponding switching state (S) of the voltage vector. a S b S c The input is fed into the inverter, and the stator current i of the permanent magnet motor at time k+1 is predicted simultaneously. d (k+1),i q (k+1).

[0008] S3. Compare the measured stator current i of the permanent magnet motor at time k+1. dm (k+1),i qm (k+1) and the predicted stator current i d (k+1),i q (k+1) yields the predicted current residual Δi dq By predicting the current residual Δi dq The inverter is judged to have an open-circuit fault by comparing it with the fault detection threshold α.

[0009] S4. When an open-circuit fault is detected in the inverter, the faulty switching transistor is located: First, an output voltage model under a single-switching transistor fault is established. Based on the output voltage model under the fault condition, a current prediction model for the permanent magnet motor drive system under the single-switching transistor fault condition is obtained. The stator current at the next moment is predicted, and the faulty switching transistor counting matrix T is defined. fault The element k of the fault switch counting matrix xj and the inverter's switching transistor T xj One-to-one correspondence.

[0010] S5. Calculate the predicted current residual under single-switch fault conditions in the inverter, find the switch corresponding to the minimum predicted current residual, and count the elements in the fault switch counting matrix corresponding to the switch. Perform fault location condition judgment. Within one current fundamental cycle, when the maximum element k of the fault switch counting matrix... xj The counting threshold k exceeds the fault location threshold. det This indicates that the switch corresponding to the largest element of the fault switch counting matrix is ​​the fault switch.

[0011] Furthermore, step S1 specifically includes the following steps:

[0012] S1-1. The mathematical model of the stator voltage of the permanent magnet motor is established based on the synchronous rotating coordinate system of the dq axis, and is expressed as follows:

[0013]

[0014] Among them, ud, u q R is the stator voltage along the dq axis. S i is the stator winding resistance. d i q L is the stator current along the dq axis. d L q For the stator inductance along the dq axis, ω e ψ is the electrical angle of the permanent magnet motor. f It is a permanent magnet flux linkage.

[0015] S1-2. The stator voltage equation above is discretized using the first-order forward Euler method, resulting in the following prediction model for the permanent magnet motor current fed by the three-level inverter:

[0016]

[0017]

[0018] Among them, i d (k+1),i q (k+1) represents the predicted value of the dq-axis stator current at sampling time k+1, i d (k), i q (k) represents the sampled value of the dq-axis stator current at sampling time k, u d (k), u q (k) represents the dq-axis stator voltage at sampling time k, T s Indicates the control cycle.

[0019] The DC-side capacitor voltage prediction model is as follows:

[0020]

[0021]

[0022] Among them, u c1 (k+1), u c2 (k+1) represents the predicted voltage values ​​of DC bus capacitors C1 and C2 at sampling time k+1. c1 (k), u c2(k) represents the voltage sampling values ​​of DC bus capacitors C1 and C2 at sampling time k, i c1 (k), i c2 (k) represents the current sampling value of DC bus capacitors C1 and C2 at sampling time k, where C is the capacitance value of the DC side capacitor.

[0023] S1-3. Determining the cost function of finite set model predictive current control based on the current prediction model and control objective:

[0024]

[0025] in, This is the d-axis current reference value. This is the q-axis current reference value, λ dc The weighting factor represents the neutral point potential balance.

[0026] Furthermore, step S2 specifically includes the following steps:

[0027] S2-1, the three-level inverter has a total of 27 switching states, each group of switching states (S a S b S c ) corresponds to a set of output voltage vectors (u d u q ), for the voltage vector (u) output of the three-level inverter d u q The inverter output voltage vector is numbered as V, where V∈{1,2,…,27}.

[0028] S2-2. Calculate the cost function value {J1, ..., J} corresponding to each group of voltage vectors. 27}

[0029] S2-3, Find the corresponding minimum cost function J miｎ Inverter switching state (S) a S b S c Then, the switch state is directly used as the inverter control signal input to control the switching transistor of the inverter to turn on and off.

[0030] S2-4, Based on the inverter switching state (S a S b S c ) and the stator current i of the permanent magnet motor at time k. d (k), i q (k), predicting the stator current at time k+1 as i d (k+1),i q (k+1).

[0031] Furthermore, step S3 specifically includes the following steps:

[0032] S3-1, Using Park transformation to convert the three-phase stator current i collected by the current sensor ａｍ (k+1),i bm (k+1),i ｃｍ (k+1) is transformed into a two-phase rotating dq-axis current i dm (k+1), iq m (k+1).

[0033] S3-2, The transformed stator current i dm (k+1),i qm (k+1) and the predicted stator current i d (k+1),i q (k+1) is used to calculate the residual:

[0034] Δi d =|i d (k+1)-i dm (k+1)|

[0035] Δi q =|i q( k+1)-i qm (k+1)|

[0036]

[0037] S3-3, predict the current residual Δi dq The inverter is compared with a fault detection threshold α to determine whether a fault has occurred. To prevent misdiagnosis due to measurement errors and noise, a fault detection variable K is defined. count .

[0038] S3-4. In each control cycle, when Δi is satisfied... dq >α, Fault detection variable K count Perform the counting.

[0039] S3-5. Perform conditional judgment for inverter open-circuit fault. Within each fundamental current cycle, the fault detection variable K... count With fault judgment threshold K det Comparison is performed, and the fault diagnosis condition is met: K count >K det This indicates that the inverter has an open circuit fault.

[0040] Furthermore, step S4 specifically includes the following steps:

[0041] S4-1. Perform commutation path analysis for open-circuit faults in three-level ANPC inverters and establish an output voltage model for single-switch faults in the inverter.

[0042] S4-2, the three-level ANPC inverter has a total of 18 switching transistors. The switching state of the inverter output is (S a ,S b ,S c Assuming a single switch failure in the inverter, predict the voltage vector under 18 switch failure conditions as V. (1～18) =(u d(1～18) (k+1),u q(1～18) (k+1)), based on the current prediction model, the predicted current vector I is obtained under the assumed single-switch fault condition. (1～18) =(i d(1～18) (k+1),i q(1～18) (k+1)).

[0043] S4-3. Define the fault switch counting matrix as follows:

[0044]

[0045] Where, matrix element k xj x = a, b, c; j = 1, 2, 3, 4, 5, 6 correspond to the switching transistor T of the three-level ANPC inverter. xj , x=a,b,c;j=1,2,3,4,5,6.

[0046] Furthermore, step S5 specifically includes the following steps:

[0047] S5-1, At time k+1, the actual measured current vector (i dm (k+1),i qm (k+1)) and the predicted current vector I under assumed fault conditions (1～18) =(i d(1～18) (k+1),i q(1～18) (k+1) is used to calculate the current vector residual:

[0048] [Δi dq ] (i) =(i d(i) (k+1)-i dm k+1)) 2 +(i q(i) -i qm (k+1)) 2

[0049] S5-2. In each model predictive control cycle, calculate the residual [Δi] between the predicted current vector and the measured current vector. dq ] (1～18) , in [Δi dq ] (1～18)Find the minimum vector residual [Δi] dq ] min Determine the fault switch corresponding to the minimum vector residual, and then use the fault switch counter element k corresponding to this fault switch. xj Counting, represented as:

[0050]

[0051] Where n represents the number of sampling points within one fundamental current cycle;

[0052] S5-3. Conditions for fault location: Within one current fundamental cycle, when the maximum element k of the fault switch counting matrix... xj The counting threshold k exceeds the fault location threshold. det The largest element k in the fault switch counting matrix xj The corresponding three-level ANPC inverter switching transistor T xj If a faulty switch is detected, the alarm will display the sequence of faulty switch transistors, thus completing the diagnosis and location of the faulty switch transistor.

[0053] The beneficial technical effects of this invention are as follows:

[0054] (1) This invention combines the features of model predictive control rolling optimization, which can predict the system current at future moments. Fault judgment can be achieved by comparing the predicted current residual with the fault detection threshold. Fault location is based on the principle of minimizing the predicted current residual, finding the assumed fault state that is closest to the actual state. This fault diagnosis method has fast diagnosis speed and high accuracy, and does not require the addition of auxiliary hardware circuits.

[0055] (2) This invention combines the advantages of analytical model-based and signal processing-based methods, and has good robustness and dynamism. The fault detection and fault location threshold settings are flexible and can adapt to complex operating conditions. It has important theoretical significance and engineering application value for promoting the application and reliable operation of ANPC inverters in the next generation of high-speed train traction drive systems and reducing operation and maintenance costs. Attached Figure Description

[0056] Figure 1 Schematic diagram of a permanent magnet motor drive system powered by a three-level ANPC inverter;

[0057] Figure 2 This is a block diagram illustrating the principle of the three-level inverter open-circuit fault diagnosis method based on predicted current residual of the present invention.

[0058] Figure 3 This is a flowchart of the open-circuit fault diagnosis method for three-level inverters based on predicted current residuals according to the present invention.

[0059] Figure 4This is a flowchart illustrating the steps of the online diagnosis method for open-circuit faults in a three-level ANPC inverter according to an embodiment of the present invention.

[0060] Figure 5 The three-level ANPC inverter T is an embodiment of the present invention. a1 Diagnostic results under open-circuit fault of switching transistor;

[0061] Figure 6 The three-level ANPC inverter T is an embodiment of the present invention. a3 Diagnostic results under open-circuit fault of switching transistor. Detailed Implementation

[0062] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0063] This invention uses a permanent magnet synchronous motor drive system powered by a three-level active neutral point clamped (3L-ANPC) inverter as an example. The main circuit schematic of this three-level ANPC inverter-powered permanent magnet synchronous motor drive system is shown below. Figure 1 As shown, the system includes a three-phase three-level ANPC inverter and a permanent magnet motor. The DC bus is connected from the positive terminal of DC bus capacitor C1 to the negative terminal of DC bus capacitor C2. The DC side capacitors C1 and C2 are 6mF, the DC bus voltage is 300V, the permanent magnet motor load is 5N.m, the reference operating speed is 600r / min, and the fault diagnosis threshold α = 0.1, K... det =3,k det The value of is related to the motor speed and the sampling frequency; in this embodiment, k is taken as . det =250, and the parameters of the permanent magnet motor used are shown in Table 1:

[0064] Table 1 Permanent Magnet Motor Parameters

[0065]

[0066]

[0067] like Figure 2 As shown, the diagnostic method provided by this invention is applied to a permanent magnet motor drive system powered by a three-level ANPC inverter under model predictive current control: it includes a model predictive current control strategy module, a fault detection module, and a fault location module. The specific control flow is as follows... Figure 3 As shown.

[0068] In this embodiment, a model predictive current control strategy module is used to drive and control the permanent magnet motor. The control method includes the following steps:

[0069] The mathematical model of the stator voltage of the permanent magnet motor is established based on the synchronous rotating coordinate system of the dq axis, and is expressed as follows:

[0070]

[0071] Among them, u d u q R is the stator voltage along the dq axis. S i is the stator winding resistance. d i q L is the stator current along the dq axis. d L q For the stator inductance along the dq axis, ω e ψ is the electrical angle of the permanent magnet motor. f For permanent magnet flux linkage;

[0072] The stator voltage equations above are discretized using the first-order forward Euler method, resulting in the following prediction model for the permanent magnet motor current fed by a three-level inverter:

[0073]

[0074]

[0075] Among them, i d (k+1),i q (k+1) represents the predicted value of the dq-axis stator current at sampling time k+1, i d (k), i q (k) represents the sampled value of the dq-axis stator current at sampling time k, u d (k), u q (k) represents the dq-axis stator voltage at sampling time k, T s Indicates the control cycle;

[0076] The DC-side capacitor voltage prediction model is as follows:

[0077]

[0078]

[0079] Among them, u c1 (k+1), u c2 (k+1) represents the predicted voltage values ​​of DC bus capacitors C1 and C2 at sampling time k+1. c1 (k), u c2 (k) represents the voltage sampling values ​​of DC bus capacitors C1 and C2 at sampling time k, i c1 (k), i c2(k) represents the current sampling value of DC bus capacitors C1 and C2 at sampling time k, where C is the capacitance value of the DC side capacitor;

[0080] Determine the cost function for finite set model predictive current control based on the current prediction model and control objective:

[0081]

[0082] in, This is the d-axis current reference value. This is the q-axis current reference value, λ dc The weighting factor represents the neutral point potential balance.

[0083] The three-level ANPC inverter has a total of 27 switching states (as shown in Table 2), and each switching state (S) a ,S b ,S c ) corresponds to a set of output voltage vectors (u d u q ), for the voltage vector (u) output of the three-level inverter d u q The inverter output voltage vector is numbered as V, where V∈{1,2,…,27};

[0084] Table 2 Switching Status of Three-Level ANPC Inverters

[0085]

[0086] The mathematical model of stator voltage and current of a three-phase permanent magnet synchronous motor based on the dq synchronous rotating coordinate system is expressed as follows:

[0087]

[0088] Among them, u a u b u c U represents the output voltage of phases a, b, and c. d u q θ represents the output voltage along the dq axis. r Expressed as electrical angles of a permanent magnet motor;

[0089] Calculate the objective function value {J1,…,J} corresponding to each group of voltage vectors. 27};

[0090] Find the corresponding minimum cost function J min Inverter switching state (S) a ,S b ,S cThen, the switch state is directly used as the inverter control signal input to control the turn-on and turn-off of the inverter power switch transistor;

[0091] Based on the output switch state (S) a ,S b ,S c ) and the stator current i of the permanent magnet motor at time k. d (k), i q (k), predicting the stator current at time k+1 as i d (k+1),i q (k+1).

[0092] like Figure 4 As shown, the diagnostic method provided by the present invention includes a fault detection module and a fault location module. In this embodiment, the fault detection and location of the inverter are realized based on the method of predicting current residual.

[0093] Fault detection and location methods include the following steps:

[0094] The three-phase stator current i acquired by the current sensor is transformed using Park transformation. am (k+1),i bm (k+1),i cm (k+1) is transformed into a two-phase rotating dq-axis current i dm (k+1),i qm (k+1);

[0095] The transformed stator current i dm (k+1), i qm (k+1) and the stator current i predicted at the previous moment d (k+1), i q (k+1) is used to calculate the residual:

[0096] Δi d =|i d (k+1)-i dm (k+1)|

[0097] Δi q =|i q (k+1)-i qm (k+1)|

[0098]

[0099] The predicted current residual Δi dq The inverter is compared with a fault detection threshold α to determine whether a fault has occurred. To prevent misdiagnosis due to measurement errors and noise, a fault detection variable K is defined. count ;

[0100] In each control cycle, when Δi is satisfied... dq >α, Fault detection variable K count Perform the counting;

[0101] To determine the conditions for an inverter open-circuit fault, the fault detection variable K is set within each fundamental current cycle. count With fault judgment threshold K det Comparison is performed, and the fault diagnosis condition is met: K count >K det This indicates that the inverter has an open circuit fault.

[0102] When an open-circuit fault is detected in the inverter, the faulty switching transistor is located. First, an output voltage model is established under a single-switching transistor open-circuit fault in the inverter. Based on the output voltage model under the fault, a current prediction model for the permanent magnet motor drive system under a single-switching transistor open-circuit fault in the inverter is obtained to predict the stator current at the next moment. The faulty switching transistor counting matrix T is defined. fault The element k of the fault switch counting matrix xj and the switching transistor T of the three-level inverter xj One-to-one correspondence.

[0103] An open-circuit fault in the inverter will cause the theoretical output phase voltage and the actual output phase voltage of the system to differ. Therefore, the predicted current obtained based on the current prediction model will deviate from the actual measured current. Based on this principle, in order to locate the faulty switch of the three-level ANPC inverter, it is necessary to perform a commutation path analysis under an open-circuit fault in the three-level ANPC inverter and establish an output voltage model under a single switch open-circuit fault in the three-level ANPC inverter. Considering the single switch fault, the phase voltages output by the inverter before and after the fault are shown in Table 3.

[0104] Table 3. Phase voltages of the three-level ANPC inverter before and after a fault.

[0105]

[0106] The three-level ANPC inverter has a total of 18 switching transistors. Based on the switching state of the inverter output, (S... a ,S b ,S c Assuming a single-switch open-circuit fault condition in the inverter, predict the voltage vector V under 18 assumed single-switch open-circuit fault conditions. (1～18) =(u d(1～18) (k+1),u q(1～18) (k+1)), based on the current prediction model, 18 predicted current vectors I under the assumption of a single-switch open-circuit fault are obtained. (1～18) =(i d(1～18) (k+1),i q(1～18)(k+1));

[0107] Define the fault counting matrix:

[0108]

[0109] Where the fault counting matrix element k xj (x = a, b, c; j = 1, 2, 3, 4, 5, 6) corresponds to the three-level ANPC inverter switch T. xj (x=a,b,c;j=1,2,3,4,5,6).

[0110] At time k+1, the actual measured current vector (i) will be... dm (k+1), i qm (k+1)) and the predicted current vector I under assumed fault conditions (1～18) =(i d(1～19) (k+1),i q(1～19) (k+1) is used to calculate the current vector residual:

[0111] [Δi dq ] (i) =(i d(i) (k+1)-i dm k+1)) 2 +(i q(i) -i qm (k+1)) 2

[0112] Within each model predictive control cycle, calculate the residual [Δi] between the predicted current vector and the measured current vector. dq ] (1～18) , in [Δi dq ] (1～18) Find the minimum vector residual [Δi] dq ] min Determine the fault switch corresponding to the minimum vector residual, and then use the fault switch counter element k corresponding to this fault switch. xj Counting, represented as:

[0113]

[0114] Where n represents the number of sampling points within one fundamental current cycle;

[0115] The condition for fault location is determined when the maximum element k of the fault switch counting matrix is ​​within one current fundamental cycle. xj The counting threshold k exceeds the fault location threshold. det The largest element k in the fault switch counting matrix xj The corresponding three-level ANPC inverter switching transistor T xjIf a faulty switch is detected, the alarm will display the sequence of faulty switch transistors, thus completing the diagnosis and location of the faulty switch transistor.

[0116] like Figure 5 and Figure 6 As shown, the open-circuit fault diagnosis method for a three-level ANPC inverter is used to diagnose the switching transistor T. a1 T a3 Online diagnosis of open circuit faults, including i a i b i c These represent the stator currents of phases a, b, and c of the permanent magnet motor, Δi dq k represents the predicted current residual. a1 and k a3 These represent the corresponding switching transistors T. a1 T a3 The fault location count value, where D and L represent the fault detection and fault location flag signals, respectively. In the sampling results of this embodiment, let the output current display of the current sensors of phases a, b, and c be 1A / div, and let the current residual Δi dq The display value is 0.05A / div. Let the fault location count variable k be... a1 and k a3 The display value is 200 / div. The inverter fault detection flag signal D is displayed as 1 / div. The inverter fault location flag signal L is displayed as 1 / div.

[0117] Three-level ANPC inverter T a1 Diagnostic results under open-circuit fault of switching transistor as follows Figure 5 As shown, when the switching transistor T a1 A fault occurred at 0.08s, and the three-phase stator current began to distort, with current i... a During the positive half-cycle, the predicted current residual Δi dq The count variable K begins to exceed the threshold α. count Start increasing, when it exceeds the fault detection threshold K det When the fault detection flag D changes from 0 to 2, the fault location subroutine is triggered. The largest element of the fault switch counting matrix is ​​k. a1 k a1 The current increases continuously within a fundamental current cycle, until the maximum element k of the fault switch counting matrix... a1 The counting threshold k exceeds the fault location threshold. det The fault location flag L changes from 0 to 1, indicating that the inverter switching transistor T... a1 It is a faulty switching transistor.

[0118] Three-level ANPC inverter T a3 Diagnostic results under open-circuit fault of switching transistor as follows Figure 6 As shown, when the switching transistor Ta3 A fault occurred at 0.1s, and the three-phase stator current began to distort, with current i... a During the negative half-cycle, the predicted current residual Δi dq The fault count variable K begins to exceed the threshold α. count Start increasing, when it exceeds the fault detection threshold K det When the fault detection flag D changes from 0 to 2, the fault location subroutine is triggered. The largest element of the fault switch counting matrix is ​​k. a3 k a3 The current increases continuously within a fundamental current cycle, until the maximum element k of the fault switch counting matrix... a3 The counting threshold k exceeds the fault location threshold. det When the fault location flag L changes from 0 to 3, it indicates that the inverter switching transistor T... a3 It is a faulty switching transistor.

Claims

1. A three-level inverter open-circuit fault diagnosis method based on predicted current residual, characterized in that, The method comprises the following steps: S1, establishing a current prediction model of a permanent magnet motor drive system fed by a three-level inverter, and determining a cost function J of the finite set model predictive current control based on the current prediction model and a control target; S2, number the voltage vectors outputted by the three-level inverter, according to the permanent magnet motor stator current i d (k), i q (k) and the three-level inverter capacitor voltage u c1 (k), u c2 (k) to obtain the minimum cost function J min The corresponding inverter output voltage vector, input the switch state (S a , S b , S c ) corresponding to the voltage vector to the inverter, while predicting the permanent magnet motor stator current i d (k+1), i q (k+1) S3, comparing the measured stator current i of the permanent magnet motor at k+1 time dm (k+1), i qm (k+1) and the predicted stator current i d (k+1), i q (k+1), obtaining the predicted current residual Δi dq , through the predicted current residual Δi dq and the comparison of the fault detection threshold α, whether the inverter has an open circuit fault is judged; S4, when the open circuit fault of the inverter is detected, the fault switch tube is positioned: first, the output voltage model under the single switch tube fault of the inverter is established, the permanent magnet motor driving system current prediction model under the single switch tube fault of the inverter is obtained based on the output voltage model under the fault, the stator current at the next moment is predicted, the fault switch tube counting matrix T is defined fault , the elements k xj of the fault switch tube counting matrix and the switch tubes T xj of the inverter are one-to-one corresponding; S5, calculate the predicted current residual under the inverter single switch tube fault condition, find out the switch tube corresponding to the minimum predicted current residual, and count the elements in the fault switch tube count matrix corresponding to the switch tube, and judge the condition of fault location. In a current fundamental cycle, when the maximum element k of the fault switch tube count matrix is greater than the count threshold k xj exceeding the count threshold k of fault location det , it is indicated that the switch tube corresponding to the maximum element of the fault switch tube count matrix is the fault switch tube; S5-1, three-level ANPC inverter has 18 switching tubes, based on the current prediction model, the predicted current vector I (1~18) = (i d(1~18) (k+1),i q (1~18) (k+1)) at the k+1 time instant the actually measured current vector (i dm (k +1), i qm (k +1)) and the predicted current vector I (1~18) =(i d(1~18) (k+1), i q (1~18) (k+1)) under the assumed fault condition are subjected to a current vector residual calculation: ； S5-2, calculating a residual error between the predicted current vector and the measured current vector in each model predictive control cycle [Δi dq ] （1~18） , in [Δi dq ] （1~18） Find the minimum vector residual [Δi] dq ] min Determine the fault switch corresponding to the minimum vector residual, and then use the fault switch counter element k corresponding to this fault switch. xj Counting, represented as: ； Wherein, n represents the number of sampling points in one current fundamental period; S5-3, condition judgment for fault location, when the maximum element k of the fault switch tube counting matrix is greater than the counting threshold k in one current fundamental cycle xj exceeding the counting threshold k for fault location det the maximum element k of the fault switch tube counting matrix xj the corresponding three-level ANPC inverter switch tube T xj is the fault switch tube, the alarm displays the fault switch tube sequence, and the diagnosis and location of the fault switch tube are completed.

2. The open-circuit fault diagnosis method for a three-level inverter based on predicted current residual error according to claim 1, characterized in that, The step S1 specifically comprises the following steps: S1-1, establishing a stator voltage mathematical model of the permanent magnet motor based on a dq-axis synchronous rotating coordinate system, and represented as: ； wherein, , is the stator voltage in dq-axes, is the stator winding resistance, , is the stator current in dq-axes, , is the stator inductance in dq-axes, is the permanent magnet motor electrical angle, is the permanent magnet flux linkage; S1-2, performing discretization processing on the above stator voltage equation by using a first-order forward Euler method, and obtaining a current prediction model of the permanent magnet motor fed by the three-level inverter as: ； ； wherein, , represents a dq-axis stator current prediction value at a sampling instant, , represents a dq-axis stator current sample value at a sampling instant, , represents a dq-axis stator voltage at a sampling instant, represents a control period; A direct current side capacitor voltage prediction model is: ； ； in, , This represents the predicted voltage values ​​of DC bus capacitors C1 and C2 at sampling time k+1. , This represents the voltage sampling values ​​of DC bus capacitors C1 and C2 at sampling time k. , This represents the current sampling values ​​of DC bus capacitors C1 and C2 at sampling time k, where C is the capacitance value of the DC side capacitors. S1-3, determining the cost function of the finite set model predictive current control based on the current prediction model and the control target: ； wherein is a d-axis current reference value, is a q-axis current reference value, λ dc denotes a weight factor for neutral point potential balancing.

3. The open-circuit fault diagnosis method for a three-level inverter based on predicted current residual error according to claim 2, characterized in that, The step S2 specifically comprises the following steps: S2-1, 27 groups of switch states of three-level inverter, each group of switch state (S a , S b , S c ) corresponds to a group of output voltage vectors (u d , u q ), the voltage vectors (u d , u q ) output by the three-level inverter are numbered, and the inverter output voltage vector is numbered as V, V∈{1,2,…,27}. S2-2. Calculate the cost function value {J1,…,J} corresponding to each group of voltage vectors. 27 }; S2-3, find the minimum cost function J min of the inverter switch state (S a , S b , S c ), and directly input the switch state as the inverter control signal to control the conduction and turn-off of the inverter switch tube; S2-4, according to the inverter switch state (S a , S b , S c ) and the current k moment permanent magnet motor stator current i d (k), i q (k), predict the k+1 moment stator current i d (k+1), i q (k+1).

4. The open-circuit fault diagnosis method for a three-level inverter based on predicted current residual error according to claim 3, characterized in that, The step S3 specifically comprises the following steps: S3-1, the three-phase stator current i collected by the current sensor is transformed into two-phase rotating dq-axis current i by using park transformation am (k+1), i bm (k+1), i cm (k+1) is transformed into two-phase rotating dq-axis current i dm (k +1), i qm (k +1), i S3-2, the transformed stator current i dm (k +1), i qm (k +1) and the predicted stator current i d (k +1), i q (k +1) ； ； ； S3-3, predict the current residual error Δi dq In comparison with the fault detection threshold α, it is determined whether the inverter is faulty. In order to prevent misdiagnosis due to measurement errors and noise, a fault detection variable K is defined count ; S3-4, in each control cycle, when Δi dq >α, fault detection variable K count counting; S3-5, the condition of inverter open-circuit fault is judged, in each current fundamental cycle, the fault detection variable K count is compared with the fault judgment threshold K det , when the fault judgment condition: K count > K det is met, it indicates that the inverter has an open-circuit fault.

5. The open-circuit fault diagnosis method for a three-level inverter based on predicted current residual error according to claim 4, characterized in that, The step S4 specifically comprises the following steps: S4-1, performing commutation path analysis of the open-circuit fault of the three-level ANPC inverter, and establishing an output voltage model under the single switch tube fault of the inverter; S4-2, the three-level ANPC inverter has 18 switch tubes, and the switching state of the inverter output is (S a , S b , S c ). Assuming that the inverter single switch tube fails, the voltage vector V (1~18) =(u d(1~18) (k+1), u q (1~18) (k+1)) based on the current prediction model, to obtain a predicted current vector I (1~18) =(i d(1~18) (k+1),i q (1~18) (k+1)) under the assumed single switch tube fault condition. S4-3, defining a fault switch tube counting matrix, and represented as: ； where the matrix elements k xj , x = a, b, c; j = 1, 2, 3, 4, 5, 6 correspond to the switching tubes T xj , x = a, b, c; j = 1, 2, 3, 4, 5, 6.

Images