CN-122026780-A - Three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction

CN122026780ACN 122026780 ACN122026780 ACN 122026780ACN-122026780-A

Abstract

The invention discloses a three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction, which comprises the steps of reconstructing theoretical stator current by utilizing reference current based on a motor steady-state model, introducing an amplitude per unit algorithm to calculate current residual error, identifying fault occurrence and locking fault phases, collecting system states at fault triggering moments, utilizing a reference voltage vector time sequence prediction model to online deduce a modulation switch sequence in a future detection period to generate dynamic reference values of theoretical action times of each switch state, utilizing an edge triggering mechanism to count actual measurement break times of the fault phases in different switch states in real time, constructing actual measurement feature vectors, and determining fault devices by calculating minimum distances between the actual measurement vectors and theoretical features of each fault mode. The invention can effectively eliminate the influence of motor rotation speed and load change on the diagnosis threshold, solves the problem that the fault characteristics of the inner tube and the clamping diode are overlapped and are difficult to distinguish, and improves the robustness and the accuracy of fault diagnosis.

Inventors

CHEN WEI
ZHANG GUOZHENG
GU XIN
JIN XUEFENG

Assignees

浙江大学先进电气装备创新中心

Dates

Publication Date: 20260512
Application Date: 20260205

Claims (8)

1. A three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction is characterized by comprising the following steps: Step S1, collecting real-time operation data of a motor driving system, reconstructing theoretical stator current by utilizing a motor steady-state model, calculating three-phase current residual errors of the system by combining an amplitude per unit algorithm, and judging that an open-circuit fault occurs in any phase and setting a fault mark when the current residual error of the phase exceeds a preset threshold value; Step S2, responding to the triggering signal of the fault sign, latching the reference voltage vector phase, the modulation degree, the carrier direction and the fault phase switching state at the moment of fault occurrence, deducing a modulation switching sequence in a future detection period on line based on the rotation characteristic of the reference voltage vector, counting the theoretical action times of the fault phase in the sequence under different level states, and constructing a theoretical fault characteristic vector; Step S3, monitoring the state jump of a fault phase switch in real time in a detection period after the fault mark is triggered, and accumulating corresponding state fault counters when the switching state of the fault phase switch is detected to be in a specific level state and the corresponding phase current has a current interruption characteristic, so as to construct an actual fault characteristic vector; And S4, constructing theoretical fault feature vectors of each power device in the fault phase bridge arm under the open-circuit fault according to the step S2, calculating the matching error between the actual fault feature vectors and each theoretical fault feature vector, and selecting the power device corresponding to the theoretical fault feature vector with the minimum matching error and meeting the tolerance requirement as a final fault diagnosis result.
2. The method for diagnosing the open-circuit fault of the three-level inverter based on the fault feature vector prediction according to claim 1, wherein the motor driving system adopts an NPC three-level inverter which comprises a three-phase bridge arm, for an x-phase bridge arm, the NPC three-level inverter comprises 4 IGBT tubes S x1 ~S x4 with anti-parallel diodes and 2 clamping diodes D x5 and D x6 , the anti-parallel diodes which are correspondingly connected with the IGBT tubes S x1 ~S x4 are D x1 ~D x4 , the collector of the S x1 is connected with the P end of the direct-current side of the inverter, the emitter of the S x1 is connected with the collector of the S x2 and the cathode of the D x5 , the emitter of the S x2 is connected with the collector of the S x3 and is connected with the x-phase stator winding of the motor as an output end of the x-phase bridge arm, the emitter of the S x3 is connected with the collector of the S x4 and the anode of the D x6 , the emitter of the S x4 is connected with the N end of the direct-current side of the inverter, the anode of the D x5 is connected with the cathode of the D x6 in parallel connection with the neutral point O end of the inverter, and the DC bus is connected between the P end and the O end through direct-current capacitors.
3. The three-level inverter open circuit fault diagnosis method based on fault feature vector prediction according to claim 1, wherein the specific implementation manner of step S1 comprises the following sub-steps: step 11, real-time operation data of a motor driving system are collected in real time, wherein the real-time operation data comprise the rotating speed of a motor, the position angle of a rotor and three-phase actual measurement stator currents; S12, reconstructing theoretical stator current by using a reference current output by an inverter control system and a motor steady-state model through the following expression; Wherein i x_model is x-phase theoretical stator current, x=a, b, c, i d_ref and i q_ref are d-axis reference current and q-axis reference current respectively, θ e is rotor position angle of the motor, K p and K i are proportional gain and integral gain of double closed loop control in the inverter control system respectively, ω ref and ω e are reference angular frequency and actual angular frequency of the motor respectively, and t represents time; Step S13, carrying out per unit treatment on three-phase actual measurement current and three-phase theoretical stator current according to the following expression; , Wherein lambda x and Respectively carrying out per-unit X-phase actual measurement on stator current and X-phase theoretical stator current, wherein I x is the X-phase actual measurement on stator current of the motor, and I base is a current reference value; substep S14 subtracting lambda x And taking the absolute value as an x-phase current residual, and judging that the phase has open-circuit fault and setting a fault mark when any phase current residual exceeds a preset threshold value.
4. The three-level inverter open circuit fault diagnosis method based on fault feature vector prediction according to claim 1, wherein the specific implementation manner of step S2 comprises the following sub-steps: Step S21, latching a reference voltage vector phase theta f , a modulation degree m, a carrier direction D and a fault phase switching state at the fault occurrence time, and predicting a phase angle theta l of a future first switching period according to the following expression; Omega e is the actual angular frequency of the motor, T s is the switching period of the inverter, and L is the number of future switching periods which are expected to be detected and covered; S22, substituting a phase angle theta l and a modulation degree m into a sector for calculation, determining the sector where a reference voltage vector of a first switching period is located in the future, and further obtaining a complete switching sequence S full (l) of the corresponding sector through table lookup according to a seven-segment vector action table of SVPWM; Step S23, determining a residual action sequence S rem of a current switching period according to the carrier direction D and the fault phase switching state, and splicing according to the following expression to obtain a full-time switching action sequence S total ; Sub-step S24 is to count theoretical action times of switching the fault phase switching state to P, O, N three different level states according to the full-time switching action sequence S total , corresponding to K P 、K O 、K N , thereby generating theoretical fault feature vector 。
5. The method of claim 1, wherein in the step S3, in a detection period after the trigger of the fault flag, if the fault phase switch state is switched to the P-level state and the interruption of the DC side P-side of the inverter is detected, the count value C P of the P-level fault counter is increased by 1, if the fault phase switch state is switched to the O-level state and the interruption of the neutral point O-side of the inverter is detected, the count value C O of the O-level fault counter is increased by 1, if the fault phase switch state is switched to the N-level state and the interruption of the DC side N-side of the inverter is detected, the count value C N of the N-level fault counter is increased by 1, if the fault trigger instant fault phase switch state is already in the above-mentioned certain level state, the initial value of the fault counter corresponding to the level state is set to 1, and finally the count values of the fault phase switch state are counted, and the actual fault feature vector is locked 。
6. The method for diagnosing an open circuit fault of a three-level inverter based on fault feature vector prediction as set forth in claim 5, wherein in step S4, a theoretical fault feature vector under an open circuit fault of each power device in a fault phase bridge arm is first constructed, and for any theoretical fault feature vector K fault_i , a matching error E i between the theoretical fault feature vector K fault_i and an actual fault feature vector C real is calculated by the following expression; K P_i 、K O_i 、K N_i represents theoretical action times of switching the switching state of the fault phase to P, O, N three different level states under the open-circuit fault of the ith power device in the fault phase bridge arm respectively, and i is the serial number of the power device in the fault phase bridge arm; then setting a tolerance threshold delta, and positioning a fault device according to the following expression; wherein F is the serial number of the faulty device.
7. A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used for executing the computer program to realize the three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction according to any one of claims 1-6.
8. A computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the three-level inverter open circuit fault diagnosis method based on fault feature vector prediction according to any one of claims 1 to 6.

Description

Three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction Technical Field The invention belongs to the technical field of motor drive system fault diagnosis, and particularly relates to a three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction. Background With the rapid development of the power electronic technology, the neutral point clamped (Neutral Point Clamped, NPC) three-level inverter is widely applied to the fields of high-voltage high-power permanent magnet synchronous motor driving, rail transit traction, new energy power generation and the like by virtue of the advantages of high output voltage waveform quality, low switching stress, small electromagnetic interference and the like. However, because the working environment is complex and bears high voltage and high current, the power switch device and the diode in the inverter are extremely easy to generate open-circuit faults, and once the faults occur, if fault-tolerant measures are not timely cut off and taken, serious distortion of output current, aggravation of torque pulsation and even secondary faults are caused, so that the safe operation of a system is seriously threatened. Therefore, the research on a rapid and accurate open circuit fault diagnosis method has important engineering significance. At present, open circuit fault diagnosis methods for NPC three-level inverters are mainly divided into two types, namely voltage-based detection and current-based detection, wherein the voltage-based method generally needs to be added with an additional voltage sensor or a hardware circuit, the cost and the volume of a system are increased, the current-based method uses the existing current sensor to diagnose by analyzing the characteristics of current waveforms, and the method is a main current research direction at present and mainly comprises a current residual method [Q. -T. An, L. Sun and L. -Z. Sun, "Current Residual Vector-Based Open-Switch Fault Diagnosis of Inverters in PMSM Drive Systems," in IEEE Transactions on Power Electronics, vol. 30, no. 5, pp. 2814-2827, May 2015]、, a current average method [S. K. E. Khil, I. Jlassi, J. O. Estima, N. Mrabet-Bellaaj and A. J. M. Cardoso, "Detection and isolation of open-switch and current sensor faults in PMSM drives, through stator current analysis," 2017 IEEE 11th International Symposium on Diagnostics for Electrical Machines, Power Electronics and Drives, Tinos, Greece, 2017, pp. 373-379], a parker vector trajectory method [N. M. A. Freire, J. O. Estima and A. J. Marques Cardoso, "Open-Circuit Fault Diagnosis in PMSG Drives for Wind Turbine Applications," in IEEE Transactions on Industrial Electronics, vol. 60, no. 9, pp. 3957-3967, Sept. 2013] and the like. Although the existing current diagnosis method has a certain result, the following significant defects still exist in an actual variable frequency speed regulation system: 1. The diagnosis threshold is difficult to set, the existing current residual method generally adopts a fixed current threshold, however, in a motor driving system, the load torque and the rotating speed are dynamically changed, and the fluctuation range of the stator current amplitude is extremely large. Under the light load or low speed working condition, the current deviation caused by the fault may be smaller than a fixed threshold value, so that the false alarm is caused, while during the heavy load or dynamic adjustment, the normal current fluctuation may exceed the threshold value, so that the false alarm is caused. 2. The complexity of the NPC topology structure causes that certain fault modes have similar current characteristics, particularly an inner tube (S x2) open circuit and a clamping diode (D x5) open circuit, both of which are shown as current interruption when the switch is in an O state, and the traditional method often has difficulty in decoupling the two through a single state characteristic, so that positioning errors are easily caused. 3. The prior diagnosis strategy generally assumes that fault characteristics are static in a diagnosis period, however, SVPWM has a sector switching and a complex seven-segment wave generation rule, and when a diagnosis window crosses a sector boundary, a switch sequence is suddenly changed, so that a diagnosis algorithm based on fixed logic is invalid, and the accuracy of diagnosis is affected. Therefore, there is a need for an NPC-type three-level inverter fault diagnosis method that can adapt to the operation condition of a motor, has a prediction capability for SVPWM modulation dynamic switching sequences, and can accurately distinguish between faulty power devices. Disclosure of Invention In view of the above, the invention provides a three-level inverter open-circuit fault diagnosis method based on fault feature vector prediction, which can solve the problems that in the prior art, the diagnosis threshold is greatly influenced by w