Power electronic converter online fault diagnosis method, medium and electronic device

Abstract

The application discloses an online fault diagnosis method, medium and electronic equipment for a power electronic converter. The method comprises the following steps: collecting capacitor voltage and output current data of the power electronic converter offline, and setting a fault label; processing the offline collected data by using a multi-feature slice improved extreme learning machine method, and training an offline power electronic converter fault diagnosis model according to the processed data and the fault label; collecting capacitor voltage and output current data sequences of the power electronic converter online; after processing the online data by using the multi-feature slice improved extreme learning machine method, inputting the processed data into the trained power electronic converter fault diagnosis model, and realizing online fault diagnosis of the power electronic converter. The application can realize open-circuit fault diagnosis of the power electronic converter, can perform noise processing on the collected signals, can reduce the training time of the fault diagnosis model, and can improve the fault diagnosis accuracy.

Description

Technical Field

[0001] This invention relates to the field of power electronics technology, and in particular to an online fault diagnosis method, medium, and electronic equipment for power electronic converters. Background Technology

[0002] Compared to two-level converter topologies, multilevel converter topologies offer advantages such as lower switching losses and lower output voltage harmonic content. In practical applications, the reliability of power electronic converters is crucial for the normal and safe operation of the system. Power semiconductor devices in power electronic converters are most prone to failure, especially short-circuit and open-circuit faults. Short-circuit faults cause overcurrent, which can be quickly detected by standard protection circuits. However, open-circuit faults lead to an increase in the harmonic content of the terminal voltage or current, which can have a destructive impact on inverter loads such as transformers or motors over the long term. Therefore, the system should have rapid open-circuit fault diagnosis capabilities to prevent serious consequences caused by the failure to detect open-circuit faults in a timely manner.

[0003] Furthermore, in engineering applications, the real signals acquired from power electronic converters are often present in the form of abundant noise. Processing noisy signals is a significant challenge for machine learning methods. Simultaneously, when acquiring multiple signals, substantial noise will reduce the accuracy of fault diagnosis and increase the training time of the diagnostic model. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the first objective of this invention is to provide an online fault diagnosis method for power electronic converters. This method can diagnose open-circuit faults in power electronic converters, and can perform noise processing on the acquired signals, reducing the training time of the fault diagnosis model and improving the accuracy of fault diagnosis.

[0005] A second objective of this invention is to provide a computer-readable storage medium.

[0006] A third objective of this invention is to provide an electronic device.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0008] An online fault diagnosis method for power electronic converters includes:

[0009] Offline acquisition of capacitor voltage and output current data of power electronic converters, and setting fault tags;

[0010] The offline collected data is processed using a multi-feature slicing improved extreme learning machine method, and a power electronic converter fault diagnosis model is trained offline based on the processed data and the fault labels.

[0011] Online acquisition of capacitor voltage and output current data sequences of power electronic converters;

[0012] After processing the online data using the multi-feature slice improved extreme learning machine method, the data is input into the trained power electronic converter fault diagnosis model to achieve online fault diagnosis of the power electronic converter.

[0013] Preferably, the offline data acquisition includes normal data and fault data of capacitor voltage and output current, wherein the fault data is the data when all switches of the power electronic converter independently experience open circuit faults.

[0014] Preferably, the offline sampling frequency is the same as the switching frequency of the switching transistor in the power electronic converter.

[0015] Preferably, the step of processing the offline collected data using the multi-feature slicing improved extreme learning machine method, and training the power electronic converter fault diagnosis model offline based on the processed data and the fault labels, includes:

[0016] The offline acquired data is subjected to a first average noise removal process, followed by a second average lightweight measurement signal processing. The second-processed data is then subjected to vertical slicing to obtain multiple training samples. These training samples are then input into the power electronic converter fault diagnosis model. Based on the corresponding fault labels, the corresponding training targets are obtained to complete the training of the power electronic converter fault diagnosis model.

[0017] Preferably, the step of performing an average noise removal process on the offline collected data includes:

[0018] Set up K data point windows, and calculate the mean value of each K-sized data window for offline data acquisition, centering on each data point. This yields the processed capacitor voltage and output current data sequence. The window center moves by one data point each time, calculated as follows:

[0019]

[0020]

[0021] Where p is the output current pointer, p = 1, 2, ..., L, L is the total output current, I′ pi I represents the average value of the i-th output current of the p-th output current. pi This represents the i-th sampled current value of the p-th output current, where i = 1, 2, ..., N, and N is the total number of sampled current values. K = 2i + 1 represents the window size of K data points. q is the capacitor voltage pointer, where q = 1, 2, ..., R, and R is the number of capacitors. U′qi U represents the average value of the i-th sampled voltage of the q-th capacitor. qi This represents the i-th sampled voltage value of the q-th capacitor. When there are not enough elements to fill the window, the mean is calculated based on the elements within the window. The calculation method is as follows:

[0022]

[0023]

[0024]

[0025]

[0026] Among them, I' pN U' is the average of the Nth current of the p-th output current. qN This is the average of the Nth sampled voltage of the qth capacitor.

[0027] Preferably, the secondary averaging and lightweight measurement signal processing of the data after the first processing includes:

[0028] Set a window of size K′ data points, and calculate the mean of local K′ data points for the capacitor voltage data sequence and output current data sequence after the first processing, to obtain the capacitor voltage data sequence and output current data sequence after the second processing; wherein, the window moves K′ data points each time, and the calculation method is as follows:

[0029]

[0030]

[0031] …

[0032]

[0033]

[0034]

[0035] …

[0036]

[0037] Among them, I′ pM U' represents the average of the Mth current of the p-th output current after secondary processing, K′ represents the window size, M represents the total number of sampled current values ​​during secondary processing, and U' q ' M This represents the average voltage of the Mth sampled value of the qth capacitor after secondary processing.

[0038] Preferably, the online sampling frequency is the same as the offline sampling frequency.

[0039] To achieve the above objectives, a second aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the above-described online fault diagnosis method for power electronic converters.

[0040] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the above-described online fault diagnosis method for power electronic converters.

[0041] This invention has at least the following technical effects:

[0042] This invention involves offline acquisition of normal and fault data sequences of capacitor voltage and output current of a power electronic converter (PEB), and the setting of fault labels. Then, a multi-feature slicing improved extreme learning machine (ELM) method is applied to train an offline PEB fault diagnosis model. Next, online acquisition of PEB capacitor voltage and output current data sequences is performed, and the online data is processed using the same method. This processed data is then input into the trained PEB fault diagnosis model, enabling online fault diagnosis of the PEB, particularly open-circuit faults. Specifically, the offline training of the PEB fault diagnosis model using the multi-feature slicing improved ELM method involves first performing a first-stage averaging to remove noise from the offline acquired data, followed by a second-stage averaging and lightweight measurement signal processing. The second-stage data is then longitudinally sliced ​​to obtain multiple training samples. These training samples are then input into the PEB fault diagnosis model, and corresponding training targets are obtained based on the fault labels. This completes the training of the PEB fault diagnosis model. This method of noise processing of the acquired signals reduces the training time of the fault diagnosis model and improves the accuracy of fault diagnosis. In addition, the fault diagnosis of power electronic converters based on the above method has the advantages of simple operation, good real-time performance, and strong generalization ability.

[0043] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0044] Figure 1 This is a flowchart of an online fault diagnosis method for power electronic converters according to an embodiment of the present invention.

[0045] Figure 2This is a topology diagram of a four-level active midpoint clamped inverter according to an embodiment of the present invention.

[0046] Figure 3 This is a schematic diagram of the physical structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0047] The following describes this embodiment in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0048] The following description, with reference to the accompanying drawings, illustrates an online fault diagnosis method, medium, and electronic device for a power electronic converter according to this embodiment.

[0049] Figure 1 This is a flowchart of an online fault diagnosis method for power electronic converters according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0050] Step S1: Collect the capacitor voltage and output current data of the power electronic converter offline and set fault tags.

[0051] In this embodiment, the offline data acquisition includes normal data (health data) of capacitor voltage and output current, and fault data. The fault data refers to the fault data when all switches in the power electronic converter experience an independent open-circuit fault. The offline sampling frequency is the same as the switching frequency of the switches in the power electronic converter.

[0052] Step S2: The offline collected data is processed using the multi-feature slicing improved extreme learning machine method, and the power electronic converter fault diagnosis model is trained offline based on the processed data and fault labels.

[0053] Among these methods, a multi-feature slicing improved extreme learning machine approach is used to process offline collected data, and a power electronic converter fault diagnosis model is trained offline based on the processed data and fault labels, including:

[0054] The offline acquired data is subjected to a first average to remove data noise, and then a second average to quantize the measurement signal. The second-processed data is then longitudinally sliced ​​to obtain multiple training samples. These training samples are then input into the power electronic converter fault diagnosis model. Based on the corresponding fault labels, the corresponding training targets are obtained to complete the training of the power electronic converter fault diagnosis model.

[0055] In this embodiment, the offline collected data undergoes an average noise removal process, including:

[0056] Set up K data point windows, and calculate the mean value of each K-sized data window for offline data acquisition, centering on each data point. This yields the processed capacitor voltage and output current data sequence. The window center moves by one data point each time, calculated as follows:

[0057]

[0058]

[0059] Where p is the output current pointer, p = 1, 2, ..., L, L is the total output current, I′ pi I represents the average value of the i-th output current of the p-th output current. pi This represents the i-th sampled current value of the p-th output current, where i = 1, 2, ..., N, and N is the total number of sampled current values. K = 2i + 1 represents the window size of K data points. q is the capacitor voltage pointer, where q = 1, 2, ..., R, and R is the number of capacitors. U′ qi U represents the average value of the i-th sampled voltage of the q-th capacitor. qi This represents the i-th sampled voltage value of the q-th capacitor. When there are not enough elements to fill the window, the mean is calculated based on the elements within the window. The calculation method is as follows:

[0060]

[0061]

[0062]

[0063]

[0064] Among them, I' pN U' is the average of the Nth current of the p-th output current. qN Let be the average of the Nth sampled voltage of the qth capacitor. From this, we can obtain the DC-side capacitor voltage data sequence U' after one processing step. q The three-phase output current data sequence I'p is respectively {U' q}={U' q1 U' q2 … U' qN} and {I' p}={I' p1 I' p2 … I' pN}

[0065] Furthermore, a second-stage averaging and lightweight measurement signal processing is performed on the data after the first-stage processing, including:

[0066] Set a window of size \(K'\) data points, and successively calculate the means of the local \(K'\) data points of the processed capacitor voltage data sequence and the output current data sequence to obtain the processed capacitor voltage data sequence and the output current data sequence; where the window moves \(K'\) data points each time, and the calculation method is as follows:

[0067]

[0068]

[0069] Where \(I'\) pM represents the \(M\)th current mean of the \(p\)th output current after the second processing, \(K'\) represents the window size, \(M\) represents the total number of sampled current values during the second processing, and \(U''\) qM represents the \(M\)th sampled voltage mean of the \(q\)th capacitor after the second processing, where \(0 < M \ll N\). Thus, the processed DC-side capacitor voltage data sequence \(U'\) q ' and the three-phase output current data sequence \(I'\) p ' are respectively \(\{U''\) q \} = \{U''\) q1 \(U''\) q2 \(\cdots U''\) qM \} and \(\{I''\) p \} = \{I''\) p1 \(I''\) p2 \(\cdots I''\) pM \}.

[0070] Furthermore, as described above, longitudinally slice the processed data, and use the sliced data to input into the extreme learning machine method to train the fault diagnosis model of the power electronic converter. The slicing method and the training method are as follows:

[0071]

[0072] Where \(I''\) LM is the \(M\)th current mean of the \(L\)th output current after the second average lightweighting, and \(U''\) RM is the \(M\)th voltage mean of the \(R\)th capacitor after the second average lightweighting.

[0073] Step S3: Online collect the capacitor voltage and output current data sequences of the power electronic converter.

[0074] Among them, when online collecting the capacitor voltage and output current data sequences of the power electronic converter, the online collection frequency is the same as the offline sampling frequency.

[0075] Step S4: After processing the online data using the multi-feature slicing improved extreme learning machine method, input it into the trained fault diagnosis model of the power electronic converter to achieve online fault diagnosis of the power electronic converter.

[0076] To demonstrate the online fault diagnosis method for power electronic converters proposed in this embodiment, a four-level active neutral-point clamping inverter is used as an example.

[0077] Four-level active midpoint clamp inverter topology as follows Figure 2 As shown, taking phase A switch tube as an example, an open circuit fault is sequentially implanted to conduct an online diagnostic experiment for the open circuit fault. The specific fault diagnosis operation steps are as follows:

[0078] Table 1 Switch Fault Identification Table

[0079] Sa1 1 Sa2 2 Sa3 3 Sa4 4 Sa5 5 Sa6 6 No fault 0

[0080] (1) Offline acquisition of the DC side capacitor voltage and three-phase output current health and fault data sequence of the four-level active neutral point clamped inverter, and setting fault tags. The switch fault identifiers are shown in Table 1.

[0081] (2) Apply the multi-feature slicing improved extreme learning machine method to train a fault diagnosis model for a four-level active midpoint clamp inverter.

[0082] (3) Online acquisition of DC side capacitor voltage and three-phase output current data sequence of four-level active neutral point clamped inverter.

[0083] (4) Apply the feature slice improved extreme learning machine method in step (2) to process online data and input it into the four-level active midpoint clamp inverter fault diagnosis model to diagnose the four-level active midpoint clamp inverter fault.

[0084] It should be noted that, compared with other inventions, the solution proposed in this invention differs in the following ways:

[0085] (1) This invention utilizes the first multi-feature slicing technology to improve the Extreme Learning Machine for diagnosing open-circuit faults of switching transistors in power electronic converters;

[0086] (2) The present invention provides a specific research method, which is as follows: First, offline data sequences of capacitor voltage and output current health and fault data of power electronic converter are collected, and fault labels are set; second, the multi-feature slicing improved extreme learning machine method is applied to train the power electronic converter fault diagnosis model offline; then, online data sequences of capacitor voltage and output current data of power electronic converter are collected; finally, the feature slicing improved extreme learning machine method is applied to process the online data, and the data is input into the power electronic converter fault diagnosis model to diagnose power electronic converter faults online.

[0087] This enables open-circuit fault diagnosis in power electronic converters. Specifically, when training the power electronic converter fault diagnosis model offline using a multi-feature slicing improved extreme learning machine method, the offline acquired data is first averaged to remove noise, then the processed data undergoes a second averaging and lightweight measurement signal processing. Finally, the processed data is vertically sliced ​​to obtain multiple training samples. These training samples are then input into the power electronic converter fault diagnosis model, and corresponding training targets are obtained based on the fault labels. This completes the training of the power electronic converter fault diagnosis model. By processing the noise in the acquired signals using this method, the training time of the fault diagnosis model can be reduced, and the accuracy of fault diagnosis can be improved. Furthermore, the power electronic converter fault diagnosis method of this invention also has the advantages of simple operation, good real-time performance, and strong generalization ability.

[0088] Furthermore, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the online fault diagnosis method for power electronic converters provided by the methods described above.

[0089] Figure 3 An example is a schematic diagram of the physical structure of an electronic device. For example... Figure 3 As shown, the electronic device may include a processor 210, a communications interface 220, a memory 230, and a communication bus 240, wherein the processor 210, communications interface 220, and memory 230 communicate with each other via the communication bus 240. The processor 210 can call logic instructions in the memory 230 to execute the aforementioned online fault diagnosis method for power electronic converters.

[0090] Furthermore, the logical instructions in the aforementioned memory 230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0091] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0092] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online fault diagnosis method for a power electronic converter, characterized in that, include: Offline acquisition of capacitor voltage and output current data of power electronic converters, and setting fault tags; The offline collected data is processed using a multi-feature slicing improved extreme learning machine method, and a power electronic converter fault diagnosis model is trained offline based on the processed data and the fault labels. Online acquisition of capacitor voltage and output current data sequences of power electronic converters; After processing the online data using the multi-feature slice improved extreme learning machine method, the data is input into the trained power electronic converter fault diagnosis model to realize online fault diagnosis of power electronic converter. The method of using multi-feature slicing to improve the extreme learning machine to process offline acquired data, and then training an offline fault diagnosis model for power electronic converters based on the processed data and the fault labels, includes: The offline acquired data is subjected to a first average noise removal process, and the processed data is subjected to a second average lightweight measurement signal processing. The processed data is then subjected to vertical slicing to obtain multiple training samples. These training samples are then input into the power electronic converter fault diagnosis model. Based on the corresponding fault labels, the corresponding training targets are obtained to complete the training of the power electronic converter fault diagnosis model. The step of performing an average noise removal process on the offline collected data includes: set up K A window of size equal to the number of data points is used, with each data point as the center of the window. The size of each window for offline data acquisition is calculated as follows: K The average of the data window is used to obtain the capacitor voltage and output current data sequence after one processing step; wherein, the center of the window moves by one data point each time, and the calculation method is as follows: in, p This is the output current pointer. p =1,2,…,L, where L is the total output current. Indicates the first p The first output current i Average current I pi Indicates the first p The first output current i Each sampled current value i =1,2,…, N , N This represents the total number of sampled current values. K =2 i +1 indicates K Window size for each data point q This is a capacitor voltage pointer. q =1,2,…,R, where R is the number of capacitors. Indicates the first q The first capacitor i Average of each sampled voltage, U qi Indicates the sampling number q The first capacitor i Each sampled voltage value; when there are not enough elements to fill the window, the average value is calculated based on the elements within the window. The calculation method is as follows: in, For the first p The first output current N Average current For the first q The first capacitor N Average of each sampled voltage; The secondary averaging and lightweight measurement signal processing of the data after the first processing includes: set up K A window the size of several data points is used to sequentially calculate the local correlation between the capacitor voltage data sequence and the output current data sequence after one processing step. K The average of each data point is used to obtain the capacitor voltage data sequence and output current data sequence after secondary processing; wherein, the window moves each time... K The data points are calculated as follows: in, Indicates the second processing step. p The first output current M Average current K Indicates window size. M This represents the total number of sampled current values ​​during secondary processing. Indicates the second processing step. q The first capacitor M The average of the sampled voltages.

2. The online fault diagnosis method for power electronic converters as described in claim 1, characterized in that, The offline data acquisition includes normal data and fault data of capacitor voltage and output current. The fault data is the data when all switches of the power electronic converter experience an open circuit fault independently.

3. The online fault diagnosis method for power electronic converters as described in claim 2, characterized in that, The offline sampling frequency is the same as the switching frequency of the switching transistor in the power electronic converter.

4. The online fault diagnosis method for power electronic converters as described in claim 3, characterized in that, The online sampling frequency is the same as the offline sampling frequency.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the online fault diagnosis method for power electronic converters as described in any one of claims 1-4.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the online fault diagnosis method for power electronic converters as described in any one of claims 1-4.

Images