A method and system for diagnosing open-circuit faults in three-phase inverter IGBTs

Abstract

This invention proposes a method and system for diagnosing open-circuit faults in three-phase inverter IGBTs, relating to the technical field of power electronic device fault diagnosis. The method involves lightweight processing of acquired historical data to obtain feature vectors; training a pre-defined three-phase inverter IGBT open-circuit fault diagnosis model based on these feature vectors; real-time acquisition of three-phase current data to be processed, followed by bounded normalization to obtain normalized current amplitudes; and generating IGBT open-circuit fault judgment results based on these normalized current amplitudes. Real-time data related to the three-phase inverter IGBT open-circuit fault is collected based on the IGBT open-circuit fault judgment results, and this real-time data is input into the trained three-phase inverter IGBT open-circuit fault diagnosis model to output the three-phase inverter IGBT open-circuit fault diagnosis results. This invention effectively achieves lightweight, efficient, and accurate diagnosis of three-phase inverter IGBT open-circuit faults.

Description

Technical Field

[0001] This invention relates to the technical field of fault diagnosis of power electronic devices, and in particular to a method and system for diagnosing open-circuit faults in three-phase inverter IGBTs. Background Technology

[0002] Asynchronous motor traction drive systems are widely used in key fields such as rail transit and electric vehicles. These systems operate under wide speed conditions for extended periods. Their core power conversion device is a three-phase inverter, whose main function is to convert DC power into three-phase AC power with controllable frequency and amplitude, providing stable drive for the asynchronous motor.

[0003] As a key power switching device in a three-phase inverter, the Insulated Gate Bipolar Transistor (IGBT) plays a crucial role in power conversion and on / off control. However, during long-term operation, it is subject to continuous electrothermal stress, making it highly susceptible to single or multiple open-circuit faults, which become a major hidden danger affecting the stable operation of the traction drive system.

[0004] Currently, some progress has been made in the diagnosis of IGBT open-circuit faults in three-phase inverters. However, the following problems still exist in practical engineering applications: On the one hand, IGBT open-circuit faults are low-probability events. Existing diagnostic methods generally adopt a "continuous sampling-continuous inference" working mode. During most of the normal operation of the system, feature extraction and fault diagnosis algorithm execution continue, resulting in long-term occupation of computing resources, increasing the load on the controller, and causing frequent fluctuations in memory and power consumption. On the other hand, existing research focuses on fault detection and location identification, neglecting the time window for the traction motor to transition from a stable operating state to a regenerative braking state after an IGBT open-circuit fault occurs. When the motor enters an unstable state, the fault characteristics will change significantly or even disappear, causing diagnostic methods that rely on steady-state characteristics to fail due to the lack of effective data. In addition, the accuracy and computational efficiency of existing fault diagnosis methods are closely related to data length and feature dimension. Most of the existing fault diagnosis methods are based on complex mathematical models or deep learning models, overemphasizing diagnostic accuracy indicators while neglecting the computational efficiency of the models, resulting in large computational loads and high time complexity. Meanwhile, some existing fault diagnosis methods only consider IGBT open-circuit faults under constant-speed operating conditions, failing to cover multiple operating scenarios such as acceleration and braking that may occur in actual operation, resulting in poor applicability of the diagnostic methods. Therefore, how to achieve lightweight, efficient, and accurate diagnosis of IGBT open-circuit faults in three-phase inverters is an important problem that urgently needs to be solved. Summary of the Invention

[0005] To address the limitations of existing technologies in achieving lightweight, efficient, and accurate diagnosis of IGBT open-circuit faults in three-phase inverters, this invention proposes a method and system for diagnosing IGBT open-circuit faults in three-phase inverters, which can effectively achieve lightweight, efficient, and accurate diagnosis of IGBT open-circuit faults in three-phase inverters.

[0006] To achieve the above-mentioned technical effects, the technical solution of the present invention is as follows:

[0007] A method for diagnosing open-circuit faults in three-phase inverter IGBTs includes the following steps:

[0008] S1. Obtain historical data related to IGBT open-circuit faults in three-phase inverters;

[0009] S2. Perform lightweight processing on the historical data to obtain feature vectors;

[0010] S3. Based on the feature vector, train the preset three-phase inverter IGBT open-circuit fault diagnosis model to obtain the trained three-phase inverter IGBT open-circuit fault diagnosis model.

[0011] S4. Real-time acquisition of three-phase current data to be processed, bounded normalization processing of the three-phase current data to obtain normalized current amplitude, and generation of IGBT open circuit fault judgment result based on the normalized current amplitude.

[0012] S5. Based on the IGBT open-circuit fault judgment result, collect real-time data related to the three-phase inverter IGBT open-circuit fault, input the real-time data into the three-phase inverter IGBT open-circuit fault diagnosis model, and output the three-phase inverter IGBT open-circuit fault diagnosis result.

[0013] Preferably, the historical data includes the DC bus voltage U dc First phase current I a Second phase current I b Third phase current I c Motor speed n s The time T1 when the motor first enters regenerative braking and the time T2 when the motor speed converges.

[0014] Preferably, the step of performing lightweight processing on the historical data to obtain a feature vector includes:

[0015] S21. The historical data is downsampled to obtain downsampled historical data;

[0016] S22. Use a preset fixed time window to extract downsampled historical data to obtain the extracted data;

[0017] S23. Filter the intercepted data to obtain filtered data;

[0018] S24. Regarding the DC bus voltage U in the filtered data... dc First phase current I a Second phase current I b Third phase current I c and motor speed n s Feature extraction is performed to obtain the feature vector X.

[0019] Preferably, the expression for the feature vector X is as follows:

[0020]

[0021] in, Indicates the first phase current I a The half-wave asymmetry characteristic, Indicates the second phase current I b The half-wave asymmetry characteristic, Indicates the third phase current I c The half-wave asymmetry characteristic, Indicate I a I b I c The minimum value among the effective values ​​and I a I b and I c The ratio of the sum of effective values, Indicate I a I b and I c Negative sequence component of current Indicate I a I b and I c Positive sequence component of current Indicates DC bus voltage U dc The relative fluctuation amplitude, Indicates the motor speed n s The standard deviation.

[0022] Preferably, the three-phase inverter IGBT open-circuit fault diagnosis model includes sequentially connected

[0023] The input layer, hidden layer, and output layer of the three-phase inverter IGBT open-circuit fault diagnosis model are trained based on the feature vectors, including:

[0024] S31. The input layer inputs the feature vector X into the hidden layer;

[0025] S32. The hidden layer performs a nonlinear transformation on the input feature vector X and outputs a transformation matrix. as follows:

[0026]

[0027] in, This indicates that the input weights are initialized randomly. This represents the hidden layer bias, and g(.) represents the sigmoid activation function.

[0028] S33. The output layer outputs a label Y corresponding to the fault category of the feature vector X, and the time when the motor first enters regenerative braking. Convergence time of motor speed As a unified output vector Based on the unified output vector and the transformation matrix Relationship:

[0029]

[0030] The output weight β is solved to complete the training of the three-phase inverter IGBT open-circuit fault diagnosis model.

[0031] Preferably, the bounded normalization process of the three-phase current data to obtain the normalized current amplitude includes:

[0032] S411. Perform low-pass filtering on the three-phase current data within a sliding window to obtain low-pass filtered three-phase current data. Perform fast Fourier transform on the low-pass filtered three-phase current data to obtain the spectral information of the three-phase current data. Extract the frequency corresponding to the main spectral peak from the spectral information as the fundamental frequency of the window. ;

[0033] S412. Based on the fundamental frequency of the window Update the length of the sliding window as follows:

[0034]

[0035] in, This represents the rounding function. This indicates the sampling frequency after downsampling. Indicates the preset number of electrical cycles;

[0036] S413. Perform arctangent normalization on the three-phase current data within the length-updated sliding window, and calculate the normalized current amplitude for each phase. as follows:

[0037]

[0038] in, , Represents the arctangent function. This represents the sampled value of the three-phase current data within the sliding window after the length update.

[0039] Preferably, generating the IGBT open-circuit fault judgment result based on the normalized current amplitude includes:

[0040] S421. Determine the normalized current amplitude. If the current exceeds a preset threshold, then the normalized current amplitude within the sliding window will be adjusted. If the percentage of samples exceeding the preset threshold is used as the percentage indicator, execute S422; otherwise, determine that the three-phase inverter IGBT is working normally.

[0041] S422. Determine whether the number of phases with the percentage index lower than the preset percentage threshold is not greater than 1. If yes, it is determined to be a single-phase IGBT open circuit fault; if no, it is determined to be a multi-phase IGBT open circuit fault.

[0042] Preferably, when the IGBT open circuit fault judgment result is a single-phase IGBT open circuit fault, the single-phase IGBT open circuit fault is located according to the positive and negative half-cycle amplitude distribution characteristics of the current phase current.

[0043] Preferably, when the IGBT open-circuit fault diagnosis result is a multi-phase IGBT open-circuit fault, real-time data related to the three-phase inverter IGBT open-circuit fault is collected, the real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. The three-phase inverter IGBT open-circuit fault diagnosis result includes a fault category prediction label. Predicted time for the motor to first enter regenerative braking Convergence prediction time of motor speed .

[0044] This invention also proposes a three-phase inverter IGBT open-circuit fault diagnosis system, comprising:

[0045] The module acquires historical data related to IGBT open-circuit faults in three-phase inverters.

[0046] The lightweight processing module is used to perform lightweight processing on the historical data to obtain feature vectors;

[0047] The training module is used to train the preset three-phase inverter IGBT open-circuit fault diagnosis model based on the feature vector, so as to obtain the trained three-phase inverter IGBT open-circuit fault diagnosis model.

[0048] The acquisition module is used to acquire the three-phase current data to be processed in real time, perform bounded normalization processing on the three-phase current data to obtain the normalized current amplitude, and generate the IGBT open circuit fault judgment result based on the normalized current amplitude.

[0049] The diagnostic module is used to collect real-time data related to the open-circuit fault of the three-phase inverter IGBT based on the IGBT open-circuit fault judgment result, input the real-time data into the three-phase inverter IGBT open-circuit fault diagnosis model, and output the three-phase inverter IGBT open-circuit fault diagnosis result.

[0050] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0051] This invention proposes a method and system for diagnosing open-circuit faults in three-phase inverters' IGBTs. First, historical data related to open-circuit faults in the three-phase inverters is acquired. This historical data undergoes lightweight processing to obtain feature vectors, effectively simplifying feature dimensions and reducing the computational load of data processing, thus achieving a lightweight diagnostic process. Second, the real-time acquired three-phase current data is subjected to bounded normalization, and an IGBT open-circuit fault judgment result is generated based on the normalized current amplitude. The fault diagnosis process is only initiated when an anomaly occurs, avoiding redundant calculations caused by continuous diagnosis in a fault-free state, reducing computational load and resource consumption. Then, based on the IGBT open-circuit fault judgment result, real-time data related to the open-circuit faults in the three-phase inverters' IGBTs are acquired. This real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. This achieves efficient and accurate diagnosis of open-circuit faults in three-phase inverters, providing a strong guarantee for the stable and reliable operation of asynchronous motor traction drive systems. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating a method for diagnosing open-circuit faults in a three-phase inverter IGBT, as proposed in an embodiment of the present invention.

[0053] Figure 2 This is a flowchart illustrating the principle of a three-phase inverter IGBT open-circuit fault diagnosis method proposed in this embodiment of the invention.

[0054] Figure 3 This diagram illustrates the structure of the three-phase inverter system in the asynchronous motor traction drive system proposed in this embodiment of the invention.

[0055] Figure 4 This is a flowchart illustrating the training process of the three-phase inverter IGBT open-circuit fault diagnosis model proposed in this embodiment of the invention.

[0056] Figure 5This diagram illustrates the structure of a three-phase inverter IGBT open-circuit fault diagnosis system proposed in this embodiment of the invention. Detailed Implementation

[0057] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.

[0058] It is understandable to those skilled in the art that some well-known details may be omitted from the accompanying drawings;

[0059] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0060] Example 1

[0061] This invention proposes a method for diagnosing open-circuit faults in three-phase inverter IGBTs, using an asynchronous motor traction drive system as an example for illustration. (See [link]). Figure 1 and Figure 2 The method includes the following steps:

[0062] S1. Obtain historical data related to IGBT open-circuit faults in three-phase inverters;

[0063] In S1, see Figure 3 A three-phase inverter system for an asynchronous motor traction drive system was established using Matlab / Simulink simulation software. The three-phase inverter is a three-phase two-level voltage source inverter. The simulation covers different speeds and IGBT open-circuit fault combinations in the asynchronous motor traction drive system. Simulations were performed on IGBTs under steady-state constant speed, acceleration, and deceleration conditions with and without multi-phase open-circuit faults. Historical data related to IGBT open-circuit faults in the three-phase inverter were continuously collected over multiple cycles. The historical data includes the DC bus voltage U. dc First phase current I a Second phase current I b Third phase current I c Motor speed n s The time T1 when the motor first enters regenerative braking and the convergence time T2 of the motor speed, and the third-phase current I. c To utilize the first phase current I a Second phase current I b The structure is based on the constraint of the third-phase current I in a three-phase three-wire system. c , And by changing the time of the fault occurrence, data on open circuit faults occurring at different times are collected.

[0064] S2. Perform lightweight processing on the historical data to obtain feature vectors;

[0065] The process of lightweighting the historical data to obtain feature vectors includes:

[0066] S21. The historical data is downsampled to obtain downsampled historical data, thus reducing the data length. Before executing S21, the sampling time interval is set in the simulation system. According to the sampling time interval Original sampling frequency as follows:

[0067]

[0068] The collected historical data is downsampled to obtain the downsampled sampling frequency. ;

[0069] S22. Determine the time of failure in the annotation. Then, a preset fixed time window is used. The historical data collected by downsampling is truncated to obtain the truncated data; the historical data collected by S1 is desampled and downsampled to obtain the downsampled sampling frequency. The fault trigger time is determined in the annotation. Then, within a preset fixed time window: The signal is intercepted internally.

[0070] S23. Filter the intercepted data to obtain filtered data;

[0071] In S23, the inverter PWM cycle is combined. With sampling period The three-phase currents in the extracted data are subjected to carrier period integrated mean (FPIM) filtering, according to the formula:

[0072]

[0073] Calculate the number of sampling points within a carrier period The three-phase current signals within the window are respectively processed using lengths of... The filtered current is obtained by the moving average and used as the filtered data. Since FPIM will shorten the data length, the DC bus voltage is used. With rotational speed Time alignment is performed to ensure that current, voltage, and speed data correspond at the same physical moment.

[0074] S24. Regarding the DC bus voltage U in the filtered data... dc First phase current I a Second phase current I bThird phase current I c and motor speed n s Feature extraction is performed to obtain the feature vector X.

[0075] At the known fault trigger time Under the premise that the feature vector X is a seven-dimensional feature vector, the expression of the feature vector X is as follows:

[0076]

[0077] in, Indicates the first phase current I a The half-wave asymmetry characteristic, Indicates the second phase current I b The half-wave asymmetry characteristic, Indicates the third phase current I c The half-wave asymmetry characteristic, Indicate I a I b I c The minimum value among the effective values ​​and I a I b and I c The ratio of the sum of effective values, Indicate I a I b and I c Negative sequence component of current Indicate I a I b and I c Positive sequence component of current Indicates DC bus voltage U dc The relative fluctuation amplitude, Indicates the motor speed n s The standard deviation. A normalized statistic representing the energy difference between the positive and negative half-waves of the three-phase current, characterizing the impact of IGBT open-circuit faults on current half-wave loss. The calculation expression is as follows:

[0078]

[0079] in, The energy of the phase current in a positive wave. The energy of the phase current in the negative wave. For the phase current in the first k The sampled values ​​of each sampling point;

[0080] Used to characterize the impact of IGBT open-circuit faults on the phase current amplitude distribution. The calculation expression is as follows:

[0081]

[0082] Used to characterize the impact of IGBT open-circuit faults on negative sequence components. The calculation expression is as follows:

[0083]

[0084] Used to indicate the DC side voltage fluctuation after an IGBT open-circuit fault occurs. The calculation expression is as follows:

[0085]

[0086] in, The total number of sampling points. In the first k DC bus voltage U at each sampling point dc , DC bus voltage U dc The average value;

[0087] Used to represent the impact of an IGBT open-circuit fault on the speed of an asynchronous motor. The calculation expression is as follows:

[0088]

[0089] in, In the first k The motor speed n at each sampling point s , Motor speed n s The average value;

[0090] S3. Based on the feature vector, train the preset three-phase inverter IGBT open-circuit fault diagnosis model to obtain the trained three-phase inverter IGBT open-circuit fault diagnosis model.

[0091] The three-phase inverter IGBT open-circuit fault diagnosis model is an Extreme Learning Machine (ELM), comprising an input layer, a hidden layer, and an output layer connected in sequence. The model is trained based on the feature vectors. (See [link to relevant documentation]). Figure 4 ,include:

[0092] S31. The input layer inputs the feature vector X into the hidden layer;

[0093] S32. The hidden layer performs a nonlinear transformation on the input feature vector X and outputs a transformation matrix. as follows:

[0094]

[0095] in, This indicates that the input weights are initialized randomly. The hidden layer bias is represented by g(.), and the sigmoid activation function is represented by g(.). The number of hidden layer nodes is limited to 30 through five-fold cross-validation to restrict the model parameters and computational cost, thus achieving lightweight modeling.

[0096] S33. The output layer outputs a label Y corresponding to the fault category of the feature vector X, and the time when the motor first enters regenerative braking. Convergence time of motor speed As a unified output vector Based on the unified output vector and the transformation matrix Relationship:

[0097]

[0098] The output weight β is calculated to complete the training of the three-phase inverter IGBT open-circuit fault diagnosis model. The regenerative braking time is included. Characterized by the time it takes for the motor to first enter regenerative braking state after an open-circuit fault, and the motor speed convergence time. The system characterizes the time during which the motor operates stably in regenerative braking after an open-circuit fault occurs, providing a quantitative reference for the protection decisions of the control system.

[0099] Here, S3 is the output label Y of the fault category corresponding to feature vector X and the time when the motor first enters regenerative braking. Convergence time of motor speed A training dataset is constructed to train a pre-defined three-phase inverter IGBT open-circuit fault diagnosis model, forming a lightweight three-phase inverter IGBT open-circuit fault diagnosis model for multi-phase IGBT open-circuit fault location and identification, regenerative braking, and speed convergence time prediction.

[0100] S4. Real-time acquisition of three-phase current data to be processed, bounded normalization processing of the three-phase current data to obtain normalized current amplitude, and generation of IGBT open circuit fault judgment result based on the normalized current amplitude.

[0101] S5. Based on the IGBT open-circuit fault judgment result, collect real-time data related to the three-phase inverter IGBT open-circuit fault, input the real-time data into the three-phase inverter IGBT open-circuit fault diagnosis model, and output the three-phase inverter IGBT open-circuit fault diagnosis result.

[0102] In S5, when the IGBT open-circuit fault judgment result is a multi-phase IGBT open-circuit fault, real-time data related to the three-phase inverter IGBT open-circuit fault is collected. This real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. The three-phase inverter IGBT open-circuit fault diagnosis result includes a fault category prediction label. Predicted time for the motor to first enter regenerative braking Convergence prediction time of motor speed To locate open-circuit fault IGBTs, a mapping rule is predefined between the fault category prediction label Y' and the open-circuit fault of IGBTs in the three-phase inverter. All open-circuit fault conditions of two-transistor combinations are enumerated, and a unique label is assigned to each fault combination, so that each value of the fault category prediction label Y' corresponds one-to-one with all two-transistor open-circuit faults. After the model outputs the fault category prediction label Y', the specific bridge arm where the IGBT open-circuit fault occurred can be located through the predefined mapping rule.

[0103] This embodiment proposes a method for diagnosing open-circuit faults in three-phase inverters' IGBTs. First, historical data related to IGBT open-circuit faults in the three-phase inverter is acquired. This historical data is then processed to obtain feature vectors, effectively simplifying the feature dimensions and reducing the computational load of data processing, thus achieving a lightweight diagnostic process. Second, the real-time acquired three-phase current data is subjected to bounded normalization, and an IGBT open-circuit fault judgment result is generated based on the normalized current amplitude. The fault diagnosis process is only initiated when an anomaly occurs, avoiding redundant calculations caused by continuous diagnosis in a fault-free state, reducing computational load and resource consumption. Then, based on the IGBT open-circuit fault judgment result, real-time data related to the three-phase inverter's IGBT open-circuit fault is acquired. This real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. This achieves efficient and accurate diagnosis of three-phase inverter IGBT open-circuit faults, providing a strong guarantee for the stable and reliable operation of asynchronous motor traction drive systems. In addition, this method is applicable to offline detection and fault device location of open circuit faults in the internal IGBTs of three-phase voltage source inverters in asynchronous motor traction drives under steady-state uniform speed and acceleration / deceleration operation. It also predicts the time from the occurrence of an IGBT open circuit fault to the occurrence of regenerative braking of the motor.

[0104] In this embodiment, the present invention first diagnoses the open-circuit fault of the multiphase IGBT in a two-level voltage source three-phase inverter, and simultaneously predicts the time and speed convergence of the asynchronous motor during regenerative braking.

[0105] Secondly, a fault diagnosis method combining the three-phase inverter IGBT open-circuit fault diagnosis model and waveform characteristics is used to ensure diagnostic accuracy while reducing data length and feature dimensions, achieving lightweight and rapid diagnosis. Furthermore, an abnormal triggering mechanism is added before fault diagnosis to avoid the problem of the system being in fault diagnosis all the time, reducing the consumption of computing resources.

[0106] Example 2

[0107] This embodiment further illustrates the method for diagnosing open-circuit faults in three-phase inverter IGBTs described in the above embodiment.

[0108] S4 describes the bounded normalization process for the three-phase current data to obtain the normalized current amplitude, including:

[0109] Before executing S411, the initial sliding window length is calculated based on the preset number of electrical cycles, and the initial window fundamental frequency is also calculated. The fundamental frequency of the window The sliding sampling window is updated adaptively in S412.

[0110] S411. Perform low-pass filtering on the three-phase current data within a sliding window to obtain low-pass filtered three-phase current data. Perform fast Fourier transform on the low-pass filtered three-phase current data to obtain the spectral information of the three-phase current data. Extract the frequency corresponding to the main spectral peak from the spectral information as the fundamental frequency of the window. ;

[0111] In S411, after obtaining the spectral information of the three-phase current data, the motor speed within that window is used as the basis for... And combined with the number of motor pole pairs The mechanical rotation speed is converted into synchronous electrical frequency. Based on the synchronous electrical frequency, the frequency corresponding to the main spectral peak is extracted from the spectrum information as the fundamental frequency of the window. ;

[0112] S412. Based on the fundamental frequency of the window Update the length of the sliding window as follows:

[0113]

[0114] in, This represents the rounding function. This indicates the sampling frequency after downsampling. This represents the preset number of electric cycles. Exponential smoothing is used to smooth the window length, and the window length change is limited to not exceeding the preset maximum rate of change threshold to avoid sudden changes in the window length. The purpose of updating the length of the sliding window is to ensure that the window always covers the same number of electric cycles under any speed condition.

[0115] S413. Perform arctangent normalization on the three-phase current data within the length-updated sliding window, and calculate the normalized current amplitude for each phase. as follows:

[0116]

[0117] in, , Represents the arctangent function. This represents the sampled value of the three-phase current data within the sliding window after the length update.

[0118] S4, which generates the IGBT open-circuit fault judgment result based on the normalized current amplitude, includes:

[0119] S421. Determine the normalized current amplitude. If the current exceeds a preset threshold, then the normalized current amplitude within the sliding window will be adjusted. Exceeding the preset threshold If the sample proportion is used as the proportion indicator, execute S422; otherwise, determine that the three-phase inverter IGBT is working normally and return to continue executing S412.

[0120] S422. Determine whether the number of phases with the percentage index lower than the preset percentage threshold is not greater than 1. If yes, it is determined to be a single-phase IGBT open circuit fault; if no, it is determined to be a multi-phase IGBT open circuit fault.

[0121] An anomaly is detected and the fault trigger time is determined when any phase ratio indicator falls below a preset phase ratio threshold. When a single-phase IGBT open-circuit fault occurs, the fault is located based on the amplitude distribution characteristics of the positive and negative half-cycles of the current phase current. When a multi-phase IGBT open-circuit fault occurs, real-time data related to the three-phase inverter IGBT open-circuit fault, such as the real-time DC bus voltage, is collected. First real-time phase current Second real-time phase current Third real-time phase current and real-time motor speed The real-time feature vector is obtained by performing lightweight processing on the real-time data as in step S2. Input the three-phase inverter IGBT open-circuit fault diagnosis model and output the three-phase inverter IGBT open-circuit fault diagnosis results; the three-phase inverter IGBT open-circuit fault diagnosis results include fault category prediction labels. Predicted time for the motor to first enter regenerative braking Convergence prediction time of motor speed .

[0122] This embodiment proposes a method for diagnosing open-circuit faults in three-phase inverters' IGBTs. The AC side current signal is periodically averaged to suppress high-frequency ripple interference caused by PWM modulation. The fundamental frequency of the current is estimated by combining zero-crossing information and motor speed. A sliding sampling length matching the electrical cycle is dynamically determined, and the current within the window is bounded and normalized. The fault diagnosis process is triggered only when the current amplitude is abnormal. A multi-mode diagnosis strategy is adopted based on the number of abnormal phases: when a single-phase abnormality is detected, the amplitude distribution characteristics of the positive and negative half-cycles of the current in that phase are analyzed to achieve rapid and regular location of the single-phase IGBT open-circuit fault; when multiple simultaneous abnormalities are detected, time-domain features are extracted from the window current, DC bus voltage, and motor speed within a preset fixed time window to construct a low-dimensional feature vector. This vector is then combined with an extreme learning machine as the three-phase inverter IGBT open-circuit fault diagnosis model to complete the identification and location of multi-phase open-circuit faults in the IGBTs, simultaneously achieving time prediction for regenerative braking and speed convergence.

[0123] It should also be noted that the method in this embodiment incorporates a fault detection step before fault diagnosis, only entering the fault diagnosis process when an anomaly occurs. This avoids redundant calculations caused by continuous diagnosis in a fault-free state, reducing computational load and resource consumption. To ensure that an effective current cycle can be collected within each sampling window, a method combining Fast Fourier Transform (FFT) and motor speed for frequency estimation of current is proposed. This dynamically determines the data sampling window that matches the electrical cycle, ensuring that the data is concise yet effective. A sub-mode diagnosis strategy is adopted based on the number of abnormal phases. In single-phase anomalies, the amplitude distribution characteristics of the positive and negative half-cycles of the current are used for rapid location. In multi-phase anomalies, time-domain features are extracted from the data within a fixed time window to construct a low-dimensional feature vector. Combined with the ELM algorithm, fault diagnosis and regenerative braking and speed convergence time prediction are performed simultaneously, providing a reference for the protection decision of the traction drive system.

[0124] Compared with existing fault diagnosis methods, this method also has the following advantages:

[0125] (i) It reduces the data length and feature dimensions required for fault diagnosis, ensuring the reliability of diagnosis while ensuring the lightweight model parameters, reducing the computational load of the system, and predicting the convergence time of regenerative braking and speed, thus providing a reliable indicator for the control system to implement protection measures.

[0126] (ii) Use FFT and motor speed to estimate the fundamental frequency, determine the length of the dynamic sampling window that matches the electrical phase period, ensure data validity, and reduce data length.

[0127] (iii) By introducing an abnormal detection triggering mechanism before fault diagnosis, subsequent diagnosis is only started when an abnormal current is detected, thus avoiding continuous execution of diagnosis during the system operation phase and reducing computational load and resource consumption.

[0128] (iv) While completing the fault diagnosis, predict the time for the system to enter regenerative braking and speed convergence after the IGBT open circuit fault, so as to provide a quantitative reference for the protection decision of the control system.

[0129] In addition, in terms of algorithms, various optimization algorithms such as recurrent neural networks with feature reconstruction (FR-RNN), multimodal deep residual filter networks (DRFN), and rough set theory (RST) are applied to data regression prediction. The improvement of a single algorithm or the use of hybrid algorithms can solve the problem of fault category classification accuracy to a certain extent.

[0130] Example 3

[0131] See Figure 5 This embodiment proposes a three-phase inverter IGBT open-circuit fault diagnosis system, including:

[0132] The module acquires historical data related to IGBT open-circuit faults in three-phase inverters.

[0133] The lightweight processing module is used to perform lightweight processing on the historical data to obtain feature vectors;

[0134] The training module is used to train the preset three-phase inverter IGBT open-circuit fault diagnosis model based on the feature vector, so as to obtain the trained three-phase inverter IGBT open-circuit fault diagnosis model.

[0135] The acquisition module is used to acquire the three-phase current data to be processed in real time, perform bounded normalization processing on the three-phase current data to obtain the normalized current amplitude, and generate the IGBT open circuit fault judgment result based on the normalized current amplitude.

[0136] The diagnostic module is used to collect real-time data related to the open-circuit fault of the three-phase inverter IGBT based on the IGBT open-circuit fault judgment result, input the real-time data into the three-phase inverter IGBT open-circuit fault diagnosis model, and output the three-phase inverter IGBT open-circuit fault diagnosis result.

[0137] In this embodiment, historical data related to open-circuit faults of the three-phase inverter IGBTs are first acquired. This historical data is then processed to obtain feature vectors, effectively simplifying the feature dimensions and reducing the computational load of data processing, thus achieving a lightweight diagnostic process. Secondly, the real-time collected three-phase current data is subjected to bounded normalization, and an IGBT open-circuit fault judgment result is generated based on the normalized current amplitude. The fault diagnosis process is only initiated when an anomaly occurs, avoiding redundant calculations caused by continuous diagnosis in a fault-free state, and reducing computational load and resource consumption. Then, based on the IGBT open-circuit fault judgment result, real-time data related to the three-phase inverter IGBT open-circuit fault is collected. This real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. This achieves efficient and accurate diagnosis of three-phase inverter IGBT open-circuit faults, providing a strong guarantee for the stable and reliable operation of the asynchronous motor traction drive system.

[0138] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for diagnosing open-circuit faults in IGBTs of a three-phase inverter, characterized in that, Includes the following steps: S1. Obtain historical data related to IGBT open-circuit faults in three-phase inverters; S2. Perform lightweight processing on the historical data to obtain feature vectors; S3. Based on the feature vector, train the preset three-phase inverter IGBT open-circuit fault diagnosis model to obtain the trained three-phase inverter IGBT open-circuit fault diagnosis model. S4. Real-time acquisition of three-phase current data to be processed, bounded normalization processing of the three-phase current data to obtain normalized current amplitude, and generation of IGBT open circuit fault judgment result based on the normalized current amplitude. The bounded normalization process for the three-phase current data to obtain the normalized current amplitude includes: S411. Perform low-pass filtering on the three-phase current data within a sliding window to obtain low-pass filtered three-phase current data. Perform fast Fourier transform on the low-pass filtered three-phase current data to obtain the spectral information of the three-phase current data. Extract the frequency corresponding to the main spectral peak from the spectral information as the fundamental frequency of the window. ; S412. Based on the fundamental frequency of the window Update the length of the sliding window. as follows: in, This represents the rounding function. This indicates the sampling frequency after downsampling. Indicates the preset number of electrical cycles; S413. Perform arctangent normalization on the three-phase current data within the length-updated sliding window, and calculate the normalized current amplitude for each phase. as follows: in, , Represents the arctangent function. This represents the sampled value of the three-phase current data within the sliding window after the length update; When the IGBT open circuit fault judgment result is a single-phase IGBT open circuit fault, the single-phase IGBT open circuit fault is located according to the positive and negative half-cycle amplitude distribution characteristics of the current phase current. When the IGBT open-circuit fault diagnosis result is a multi-phase IGBT open-circuit fault, real-time data related to the three-phase inverter IGBT open-circuit fault is collected. This real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. The three-phase inverter IGBT open-circuit fault diagnosis result includes a fault category prediction label. Predicted time for the motor to first enter regenerative braking Convergence prediction time of motor speed ; S5. Based on the IGBT open-circuit fault judgment result, collect real-time data related to the three-phase inverter IGBT open-circuit fault, input the real-time data into the three-phase inverter IGBT open-circuit fault diagnosis model, and output the three-phase inverter IGBT open-circuit fault diagnosis result.

2. The method for diagnosing open-circuit faults in three-phase inverter IGBTs according to claim 1, characterized in that, The historical data includes DC bus voltage U dc First phase current I a Second phase current I b Third phase current I c Motor speed n s The time T1 when the motor first enters regenerative braking and the time T2 when the motor speed converges.

3. The method for diagnosing open-circuit faults in three-phase inverters according to claim 2, characterized in that, The process of lightweighting the historical data to obtain feature vectors includes: S21. The historical data is downsampled to obtain downsampled historical data; S22. Use a preset fixed time window to extract downsampled historical data to obtain the extracted data; S23. Filter the intercepted data to obtain filtered data; S24. Regarding the DC bus voltage U in the filtered data... dc First phase current I a Second phase current I b Third phase current I c and motor speed n s Feature extraction is performed to obtain the feature vector X.

4. The method for diagnosing open-circuit faults of IGBTs in three-phase inverters according to claim 3, characterized in that, The expression for the feature vector X is as follows: in, Indicates the first phase current I a The half-wave asymmetry characteristic, Indicates the second phase current I b The half-wave asymmetry characteristic, Indicates the third phase current I c The half-wave asymmetry characteristic, Indicate I a I b I c The minimum value among the effective values ​​and I a I b and I c The ratio of the sum of effective values, Indicate I a I b and I c Negative sequence component of current Indicate I a I b and I c Positive sequence component of current Indicates DC bus voltage U dc The relative fluctuation amplitude, Indicates the motor speed n s The standard deviation.

5. The method for diagnosing open-circuit faults of IGBTs in a three-phase inverter according to claim 4, characterized in that, The three-phase inverter IGBT open-circuit fault diagnosis model includes an input layer, a hidden layer, and an output layer connected in sequence. The training of the preset three-phase inverter IGBT open-circuit fault diagnosis model based on the feature vector includes: S31. The input layer inputs the feature vector X into the hidden layer; S32. The hidden layer performs a nonlinear transformation on the input feature vector X and outputs a transformation matrix. as follows: in, This indicates that the input weights are initialized randomly. This represents the hidden layer bias, and g(.) represents the sigmoid activation function. S33. The output layer outputs a label Y corresponding to the fault category of the feature vector X, and the time when the motor first enters regenerative braking. Convergence time of motor speed As a unified output vector Based on the unified output vector and the transformation matrix Relationship: The output weight β is solved to complete the training of the three-phase inverter IGBT open-circuit fault diagnosis model.

6. The method for diagnosing open-circuit faults of IGBTs in a three-phase inverter according to claim 1, characterized in that, The process of generating an IGBT open-circuit fault judgment result based on the normalized current amplitude includes: S421. Determine the normalized current amplitude. If the current exceeds a preset threshold, then the normalized current amplitude within the sliding window will be adjusted. If the percentage of samples exceeding the preset threshold is used as the percentage indicator, execute S422; otherwise, determine that the three-phase inverter IGBT is working normally. S422. Determine whether the number of phases with the percentage index lower than the preset percentage threshold is not greater than 1. If yes, it is determined to be a single-phase IGBT open circuit fault; if no, it is determined to be a multi-phase IGBT open circuit fault.

7. A three-phase inverter IGBT open-circuit fault diagnosis system, characterized in that, include: The acquisition module is used to acquire historical data related to IGBT open-circuit faults in three-phase inverters; The lightweight processing module is used to perform lightweight processing on the historical data to obtain feature vectors; The training module is used to train the preset three-phase inverter IGBT open-circuit fault diagnosis model based on the feature vector, so as to obtain the trained three-phase inverter IGBT open-circuit fault diagnosis model. The acquisition module is used to acquire the three-phase current data to be processed in real time, perform bounded normalization processing on the three-phase current data to obtain the normalized current amplitude, and generate the IGBT open circuit fault judgment result based on the normalized current amplitude. The bounded normalization process for the three-phase current data to obtain the normalized current amplitude includes: S411. Perform low-pass filtering on the three-phase current data within a sliding window to obtain low-pass filtered three-phase current data. Perform fast Fourier transform on the low-pass filtered three-phase current data to obtain the spectral information of the three-phase current data. Extract the frequency corresponding to the main spectral peak from the spectral information as the fundamental frequency of the window. ; S412. Based on the fundamental frequency of the window Update the length of the sliding window. as follows: in, This represents the rounding function. This indicates the sampling frequency after downsampling. Indicates the preset number of electrical cycles; S413. Perform arctangent normalization on the three-phase current data within the length-updated sliding window, and calculate the normalized current amplitude for each phase. as follows: in, , Represents the arctangent function. This represents the sampled value of the three-phase current data within the sliding window after the length update; When the IGBT open circuit fault judgment result is a single-phase IGBT open circuit fault, the single-phase IGBT open circuit fault is located according to the positive and negative half-cycle amplitude distribution characteristics of the current phase current. When the IGBT open-circuit fault diagnosis result is a multi-phase IGBT open-circuit fault, real-time data related to the three-phase inverter IGBT open-circuit fault is collected. This real-time data is input into the three-phase inverter IGBT open-circuit fault diagnosis model, and the three-phase inverter IGBT open-circuit fault diagnosis result is output. The three-phase inverter IGBT open-circuit fault diagnosis result includes a fault category prediction label. Predicted time for the motor to first enter regenerative braking Convergence prediction time of motor speed ; The diagnostic module is used to collect real-time data related to the open-circuit fault of the three-phase inverter IGBT based on the IGBT open-circuit fault judgment result, input the real-time data into the three-phase inverter IGBT open-circuit fault diagnosis model, and output the three-phase inverter IGBT open-circuit fault diagnosis result.

Images