Transformer fault diagnosis method, diagnosis device and electronic equipment

Abstract

The invention provides a transformer fault diagnosis method and device and electronic equipment. The transformer fault diagnosis method comprises the steps of obtaining a real-time sound signal of a transformer; dividing the real-time sound signal into a plurality of sound signal segments according to a preset duration; acquiring an initial feature vector of the sound signal segment; and determining a target fault diagnosis result by using the initial feature vector and a preset transformer fault diagnosis model. According to the transformer fault diagnosis method, the feature vectors which better represent the mechanical faults of the transformer can be extracted, the accuracy of identifying the faults of the transformer is improved, the probability of major faults of equipment is reduced, and meanwhile, the personnel operation and maintenance cost and the safety can be reduced.

Description

Technical Field

[0001] The present disclosure relates to the technical field of transformer fault diagnosis, and more specifically, to a transformer fault diagnosis method, a diagnosis device, and an electronic device. Background Art

[0002] A transformer is one of the important devices in the power system. The normal operation of the transformer can ensure the stability and reliability of the power system. If the transformer fails, it will not only affect normal production and people's stable life, but also cause huge economic losses. In severe cases, it may even pose a threat and damage to personal safety.

[0003] However, due to the low probability of transformer failures, the existing fault detection methods are usually regular maintenance or after-failure repair, and cannot timely detect the potential fault risks of the transformer. In addition, the existing fault detection methods usually require manual participation in analysis, which is not only inefficient but also easily affected by human factors. Summary of the Invention

[0004] To this end, the present disclosure provides a transformer fault diagnosis method, a diagnosis device, and an electronic device to solve at least one of the above problems.

[0005] In a first aspect of the present disclosure, a transformer fault diagnosis method is provided. The transformer fault diagnosis method includes: obtaining a real-time sound signal of the transformer; dividing the real-time sound signal into a plurality of sound signal segments according to a preset duration; obtaining an initial feature vector of the sound signal segment; and determining a target fault diagnosis result by using the initial feature vector and a preset transformer fault diagnosis model.

[0006] Optionally, the step of determining a target fault diagnosis result by using the initial feature vector and a preset transformer fault diagnosis model includes: inputting the initial feature vectors of the plurality of sound signal segments into the fault diagnosis model to obtain a plurality of initial fault diagnosis results; counting the plurality of initial fault diagnosis results, and using the initial fault diagnosis result with the largest count value as the target fault diagnosis result.

[0007] Optionally, the preset transformer fault diagnosis model is trained through the following steps: Obtain the historical sound signals of the transformer, where the historical sound signals include abnormal sound signals and normal sound signals; Process the normal sound signals according to the historical fault sound signals of the transformer and preset noise signals to obtain fault sound signals and noise-added sound signals; Extract features from the fault sound signals, the noise-added sound signals, and the abnormal sound signals to obtain fault training features, noise-added training features, and abnormal training features; Use the fault training features, the noise-added training features, and the abnormal training features to train a preset single-classification model to obtain the preset transformer fault diagnosis model.

[0008] Optionally, the step of processing the normal sound signals according to the historical fault sound signals of the transformer and preset noise signals to obtain fault sound signals and noise-added sound signals includes: Adjust the frequency spectrum of the normal sound signals according to the historical fault sound signals to obtain the fault sound signals; Perform data superposition on the preset noise data and the normal sound signals to obtain the noise-added sound signals.

[0009] Optionally, the step of adjusting the frequency spectrum of the normal sound signals according to the historical fault sound signals to obtain the fault sound signals includes: Determine the frequency range and amplitude range of the corresponding fault frequency spectrum according to the historical fault sound signals; Perform time-frequency conversion on the normal sound signals to obtain a first frequency spectrum; Adjust the amplitude of the first frequency spectrum within the frequency range according to the frequency range and amplitude range of the fault frequency spectrum to obtain a second frequency spectrum; Perform inverse time-frequency conversion on the second frequency spectrum to obtain the fault sound signals.

[0010] Optionally, the step of performing data superposition on the preset noise data and the normal sound signals to obtain the noise-added sound signals includes: Perform time-frequency conversion on the normal sound signals and the preset noise signals respectively, and determine the maximum amplitude of the frequency spectrum of the normal sound signals and the maximum amplitude of the frequency spectrum of the preset noise signals; Calculate the ratio of the maximum amplitude of the frequency spectrum of the normal sound signals to the maximum amplitude of the frequency spectrum of the preset noise signals; Scale the preset noise signals according to the ratio to obtain a first sound signal; Superpose the normal sound signals and the first sound signal to obtain the noise-added sound signals.

[0011] Optionally, the step of extracting features from the fault sound signal, the noise-added sound signal, and the abnormal sound signal to obtain fault training features, noise-added training features, and abnormal training features includes: dividing the fault sound signal, the noise-added sound signal, and the abnormal sound signal into multiple fault sound signal segments, multiple noise-added sound signal segments, and multiple abnormal sound signal segments respectively according to a preset duration; performing the following steps for each sound signal segment: performing frame division on the sound signal segment to obtain multiple frames of data; performing windowing processing on each frame of data according to a preset window and a preset window sliding step to obtain windowed data; performing time-frequency transformation on the windowed data to obtain a frequency spectrum sequence; and extracting features from the frequency spectrum sequence by using a Mel filter bank to obtain the training features of the sound signal segment.

[0012] A second aspect of the present disclosure provides a transformer fault diagnosis device, which includes: a signal acquisition unit configured to acquire a real-time sound signal of a transformer; a signal division unit configured to divide the real-time sound signal into multiple sound signal segments according to a preset duration; a feature vector acquisition unit configured to acquire an initial feature vector of each sound signal segment; and a diagnosis result determination unit configured to determine a target fault diagnosis result by using the initial feature vector of each sound signal segment and a preset transformer fault diagnosis model.

[0013] A third aspect of the present disclosure provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the above-mentioned transformer fault diagnosis method is implemented.

[0014] A fourth aspect of the present disclosure provides an electronic device, which includes: at least one processor; and at least one memory storing a computer program, wherein when the computer program is executed by the at least one processor, the above-mentioned transformer fault diagnosis method is implemented.

[0015] According to the transformer fault diagnosis method, diagnosis device, and electronic device of the embodiments of the present disclosure, the real-time sound signal of a transformer can be processed, the real-time sound signal can be diagnosed by using a preset transformer fault diagnosis model, and the fault condition of the transformer can be determined according to the target fault diagnosis result. In the above-mentioned transformer fault diagnosis method, diagnosis device, and electronic device, a feature vector that better characterizes the mechanical fault of the transformer can be extracted, the accuracy of identifying the fault of the transformer can be improved, the probability of a major fault of the device can be reduced, and at the same time, the personnel operation and maintenance cost and safety can also be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart showing a transformer fault diagnosis method according to an exemplary embodiment of the present disclosure.

[0017] Figure 2 It is a flowchart showing a method for training a preset transformer fault diagnosis model according to an exemplary embodiment of the present disclosure.

[0018] Figure 3 It is a block diagram showing a transformer fault diagnosis device according to an exemplary embodiment of the present disclosure.

[0019] Figure 4 It is a block diagram showing an electronic device according to an exemplary embodiment of the present disclosure. Detailed implementation manners

[0020] The present disclosure provides the following detailed implementation manners to help readers obtain a comprehensive understanding of the methods, devices, and / or systems described herein. However, after understanding the disclosure of the present disclosure, various changes, modifications, and equivalents of the methods, devices, and / or systems described herein will be apparent. For example, the order of operations described herein is merely exemplary and is not limited to those set forth herein. Rather, it can be changed as will be apparent after understanding the disclosure of the present disclosure, except for operations that must occur in a specific order. In addition, descriptions of features known in the art may be omitted for greater clarity and conciseness.

[0021] The features described herein can be implemented in different forms and should not be construed as limited to the examples described herein. On the contrary, the examples described herein are provided only to illustrate some of the many feasible ways of implementing the methods, devices, and / or systems described herein, and many feasible ways will be apparent after understanding the disclosure of the present disclosure.

[0022] As used herein, the term "and / or" includes any one of the associated listed items and any combination of any two or more of them.

[0023] Although terms such as "first", "second", and "third" may be used herein to describe various components, components, regions, layers, or parts, these components, components, regions, layers, or parts should not be limited by these terms. On the contrary, these terms are only used to distinguish one component, component, region, layer, or part from another component, component, region, layer, or part. Therefore, without departing from the teachings of the examples, the first component, the first component, the first region, the first layer, or the first part referred to in the examples described herein may also be referred to as the second component, the second component, the second region, the second layer, or the second part.

[0024] In the specification, when an element (such as, a layer, a region, or a substrate) is described as being "on" another element, "connected to" or "coupled to" another element, the element can be directly "on" the other element, directly "connected to" or "coupled to" the other element, or there can be one or more other elements therebetween. In contrast, when an element is described as being "directly on" another element, "directly connected to" or "directly coupled to" another element, there can be no other elements therebetween.

[0025] The terms used herein are for the purpose of describing various examples only and are not intended to limit the disclosure. Unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. The terms "comprising", "including" and "having" specify the presence of the stated features, quantities, operations, components, elements and / or combinations thereof, but do not preclude the presence or addition of one or more other features, quantities, operations, components, elements and / or combinations thereof.

[0026] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs after understanding this disclosure. Unless explicitly defined as such herein, terms (such as those defined in a general dictionary) shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure, and shall not be interpreted in an idealized or overly formal manner.

[0027] In addition, in the description of the examples, when a detailed description of a related structure or function that is considered to be well-known would cause an ambiguous interpretation of the disclosure, such detailed description will be omitted.

[0028] As mentioned above, during the operation of the transformer, structures such as the iron core and windings vibrate due to the influence of excitation such as magnetostriction, electromagnetic force, and torque force, and generate mechanical waves, and these mechanical waves are transmitted through media such as solids and liquids in the equipment to the equipment housing or the surrounding ambient air, thereby generating vibrations and audible sound signals, and the equipment state of the equipment can be determined according to the vibration signal and the real-time sound signal of the equipment.

[0029] A transformer fault diagnosis method, a diagnosis device, and an electronic device according to an exemplary embodiment of the present disclosure will be described below with reference to the accompanying drawings to solve or alleviate at least one of the above problems.

[0030] Figure 1 is a flowchart showing a transformer fault diagnosis method according to an exemplary embodiment of the present disclosure.

[0031] Referring to Figure 1 , in step S101, a real-time sound signal of the transformer can be obtained.

[0032] As an example, the sound signal of a transformer is usually the noise generated when the transformer is operating. The sound signal of the transformer can be detected through continuous monitoring to diagnose the health status of the transformer. Specifically, a data collector can be set around the transformer to collect the sound signal of the transformer during operation. At the same time, the grid voltage, current, etc. can also be collected as the operating state data of the transformer. Here, the data collector can be a non-contact pick-up device, which can collect a large amount of sound signals of the transformer, solve the problem of lack of data, and achieve effective real-time monitoring of the transformer.

[0033] In step S102, the real-time sound signal can be divided into multiple sound signal segments according to a preset duration. Here, first, the obtained real-time sound signal can be denoised. For example, the sound signal can be denoised by spectral subtraction, but the present disclosure is not limited thereto. Then, the denoised sound signal can be converted into a sound signal in a preset format (such as but not limited to the.wav format) to obtain a sound signal with lossless quality and easy to edit and process. Finally, the sound signal after format conversion can be divided into multiple sound signal segments according to a preset duration (such as but not limited to 5 seconds).

[0034] As an example, a real-time sound signal with a duration of 1 minute can be obtained, denoised and its sound format can be converted. Then, the 1-minute sound signal can be divided into 12 new sound signal segments according to a preset duration of 5 seconds.

[0035] In step S103, the initial feature vector of the sound signal segment can be obtained. Here, the initial feature vector is an MFCC (Mel Frequency Cepstral Coefficient) feature vector, and the MFCC vector can be extracted for each sound signal segment. Specifically, since humans are relatively insensitive to the high-frequency part of the sound signal, the human perception of sound frequency is non-linear, and the Mel spectrum conforms to this auditory characteristic. Therefore, the MFCC vector of each sound signal segment can be extracted. Further, through mechanism analysis, it can be known that the sounds of transformer mechanical faults mostly concentrate in the low-frequency part above 50 Hz and below 2000 Hz. Therefore, when extracting the MFCC features, the upper and lower limits of the frequency range can be set to 2000 Hz and 50 Hz respectively, and the number of filters can be set to 120. However, the above upper and lower limits of the frequency range and the number of filters are only examples, and the present disclosure is not limited thereto.

[0036] In step S104, the target fault diagnosis result can be determined by using the initial feature vector and the preset transformer fault diagnosis model.

[0037] Specifically, the initial feature vectors of multiple voice signal segments can be input into the fault diagnosis model to obtain multiple initial fault diagnosis results; count the multiple initial fault diagnosis results, and use the initial fault diagnosis result with the largest count value as the target fault diagnosis result.

[0038] Here, each initial feature vector can be input into a preset transformer fault diagnosis model, and the initial fault diagnosis result of each initial feature vector can be determined according to the preset transformer fault diagnosis model. When the initial fault diagnosis result is 1, it is determined that the current voice signal segment is a normal voice signal segment. When the initial fault diagnosis result is -1, it is determined that the current voice signal segment is an abnormal voice signal segment. Then, determine the count value of each initial fault diagnosis result, and use the initial fault diagnosis result with the largest count value as the target fault diagnosis result. Taking the above 1-minute real-time voice signal as an example, the 12 divided voice signal segments are respectively input into the preset transformer fault diagnosis model to obtain 12 initial fault diagnosis results. Among them, the number of initial fault diagnosis results with a value of 1 is 10, and the number of initial fault diagnosis results with a value of -1 is 2. Therefore, the value 1 is used as the target fault diagnosis result of the 1-minute real-time voice signal.

[0039] Figure 2 It is a flowchart showing a training method of a preset transformer fault diagnosis model according to an exemplary embodiment of the present disclosure.

[0040] Refer to Figure 2 In step S201, the historical voice signal of the transformer can be obtained, where the historical voice signal includes abnormal voice signals and normal voice signals. Here, the historical voice signal of the transformer can be obtained through a data acquisition device. For example, the voice signals of the transformer in the past two months can be collected, and the collected voice signals can be data-labeled based on multi-source data analysis such as historical experience and working condition data, and the voice signals can be divided into normal voice signals and abnormal voice signals to more accurately and efficiently identify and extract the feature vectors of the voice signals. Here, the duration of collecting the voice signals is only an example, and the present disclosure does not make any restrictions on this.

[0041] Optionally, in order to better train the preset transformer fault diagnosis model, the historical voice signal can be processed. For example, the historical working condition data of the transformer can be obtained, and the historical working condition data includes weather data and operating state data of the substation where the transformer is located. Among them, the acquisition time of the historical working condition data corresponds to the acquisition time of the historical voice signal; then, the abnormal voice signals corresponding to the specific historical working condition data in the historical working condition data in the historical voice signal can be corrected to normal voice signals.

[0042] Here, the weather data may include, for example, the wind speed and rainfall on the current day, etc., and the operating state data of the transformer may include, for example, the voltage and current of the power grid, etc. The weather data and the operating state data are only exemplary examples, and they may also include other data, and the present disclosure does not impose any restrictions thereon.

[0043] On the other hand, the weather data and the operating state data will affect the quality of the sound signal of the transformer. For example, in windy and rainy weather, the sound signal of the transformer will become relatively weak. In addition, the voltage and current fluctuations of the power grid will affect the magnetic flux of the transformer winding and the iron core, thereby affecting the energy and frequency of the sound wave, etc. In practical applications, when encountering windy and rainy weather, the sound signal collected by the data acquisition device is easily judged as an abnormal sound signal. Therefore, the abnormal sound signal can be corrected to a normal sound signal according to the historical weather data and historical working condition data when the historical sound signal is collected.

[0044] In step S202, the normal sound signal can be processed according to the historical fault sound signal of the transformer and the preset noise signal to obtain a fault sound signal and a noise-added sound signal.

[0045] In practical applications, the historical fault sound signal can be a known fault sound signal. By processing the normal sound signal in the historical sound signal with the known fault sound signal, a simulated fault sound signal can be obtained, which can increase the number of fault samples. The preset noise signal can be multiple groups of noise data that meet the conditions selected from the public noise database according to the geographical location and working scenario of the transformer. In this way, the diversity of the training data of the preset transformer fault diagnosis model is increased through the added fault sound signal and noise-added sound signal, and the diagnostic ability of the preset transformer fault diagnosis model is improved.

[0046] According to an embodiment of the present disclosure, in step S202, the spectrum of the normal sound signal can be adjusted according to the historical fault sound signal to obtain a fault sound signal; and the preset noise data and the normal sound signal can be superimposed to obtain a noise-added sound signal.

[0047] According to an embodiment of the present disclosure, the step of adjusting the spectrum of the normal sound signal according to the historical fault sound signal to obtain a fault sound signal may include: determining the frequency range and amplitude range of the corresponding fault spectrum according to the historical fault sound signal; performing time-frequency conversion on the normal sound signal to obtain a first spectrum; adjusting the amplitude of the first spectrum within the frequency range according to the frequency range and amplitude range of the fault spectrum to obtain a second spectrum; performing inverse time-frequency conversion on the second spectrum to obtain a fault sound signal. In this way, by adjusting the amplitude of the normal sound signal within the fault spectrum range to be the same as the amplitude of the historical fault signal through the amplitude of the historical fault signal within the fault spectrum range, a fault sound signal is obtained.

[0048] According to an embodiment of the present disclosure, the step of superimposing preset noise data and the normal sound signal to obtain a noise-added sound signal may include: performing time-frequency conversion on the normal sound signal and the preset noise signal respectively, and determining the maximum amplitude of the spectrum of the normal sound signal and the maximum amplitude of the spectrum of the preset noise signal; calculating the ratio of the maximum amplitude of the spectrum of the normal sound signal to the maximum amplitude of the spectrum of the preset noise signal; scaling the preset noise signal according to the ratio to obtain a first sound signal; and superimposing the normal sound signal and the first sound signal to obtain a noise-added sound signal. In this way, time-frequency domain conversion is performed on the normal sound signal and the preset noise signal, the ratio of the maximum amplitude of the spectrum of the normal sound signal to the maximum amplitude of the spectrum of the noise sound signal is calculated, and the preset noise data is scaled according to this ratio to obtain a first sound signal with the same frequency as the normal sound signal. Then, the first sound signal and the normal sound signal are superimposed to obtain a noise-added sound signal.

[0049] Return to reference Figure 2 , in step S203, feature extraction may be performed on the fault sound signal, the noise-added sound signal, and the abnormal sound signal to obtain fault training features, noise-added training features, and abnormal training features.

[0050] According to an embodiment of the present disclosure, the fault sound signal, the noise-added sound signal, and the abnormal sound signal may be divided into multiple fault sound signal segments, multiple noise-added sound signal segments, and multiple abnormal sound signal segments respectively according to a preset duration; for each sound signal segment, the following steps are performed: performing frame division processing on the sound signal segment to obtain multiple frames of data; performing windowing processing on each frame of data according to a preset window and a preset window sliding step length to obtain windowed data; performing time-frequency transformation on the windowed data to obtain a spectrum sequence; and performing feature extraction on the spectrum sequence by using a Mel filter bank to obtain the training features of the sound signal segment.

[0051] As an example, before feature extraction, the fault sound signal, the noise-added sound signal, and the abnormal sound signal can be divided into multiple sound signal segments with a preset duration (such as but not limited to 5 seconds). Each sound signal segment is converted into a preset format (such as but not limited to the.wav format), and then the sound signal segments converted into the preset format are framed, with each frame containing data from 10 milliseconds to 30 milliseconds. For each frame of data, windowing is performed according to a preset window size (such as 200 ms) and a preset window step size (such as 100 ms). The fast Fourier transform (FFT) is performed on the windowed data of each frame, and the transformed power spectrum is passed through a series of Mel filter banks (which can be 120) to simulate the different sensitivities of the human ear to different frequencies. Then, the logarithm of the energy output by each Mel filter bank is taken to obtain the Mel log power spectrum coefficients. To reduce the correlation of the signal, the discrete cosine transform is performed on the Mel log power spectrum coefficients of each frame to obtain the Mel frequency cepstrum coefficients, that is, the MFCC features.

[0052] In step S204, the preset single-classification model can be trained using the fault training features, the noise-added training features, and the abnormal training features to obtain a preset transformer fault diagnosis model. Here, the preset single-classification model can be a one-class support vector machine (One-Class SVM (Support Vector Machine)). Specifically, the principle of the single-classification model is as follows. First, the normal data can be mapped into a high-dimensional feature space so that the normal data points can be surrounded by a hyperplane. This hyperplane is called the decision boundary. By maximizing the distance between the hyperplane and the normal data, an optimal separating hyperplane is found so that the abnormal points are as far away from the hyperplane as possible. This means that the decision boundary should be as far away from the normal data points as possible. For new data points, by calculating the distance between them and the hyperplane, it is determined whether the data points are abnormal. The data points with a larger distance are more likely to be abnormal points.

[0053] In this way, the one-class classification model is trained using the known normal sound signals and abnormal sound signals. Since the currently collected data is mainly normal samples, the one-class classification model is used for model training and diagnosis, learning the center of the normal sound signals, and identifying the abnormal sound samples far from the center as abnormal, which can timely and effectively detect the potential fault risks of the transformer and avoid serious equipment failures and accidents.

[0054] According to the embodiments of the present disclosure, during the transformer fault diagnosis process, feature vectors that can better characterize the mechanical faults of the transformer can be extracted, the accuracy of identifying transformer faults can be improved, the probability of major equipment failures can be reduced, and at the same time, the personnel operation and maintenance costs and safety can also be reduced.

[0055] The method for diagnosing transformer faults according to the exemplary embodiments of the present disclosure has been described above. Next, a transformer fault diagnosis device according to the exemplary embodiments of the present disclosure will be described in conjunction with Figure 3 a description of a transformer fault diagnosis device according to the exemplary embodiments of the present disclosure.

[0056] Figure 3 FIG. is a block diagram showing a transformer fault diagnosis device according to the exemplary embodiments of the present disclosure.

[0057] As Figure 3 shown, the transformer fault diagnosis device 300 includes: a signal acquisition unit 301, a signal division unit 302, a feature vector acquisition unit 303, and a diagnosis result determination unit 304.

[0058] The signal acquisition unit 301 can acquire the real-time sound signal of the transformer; the signal division unit 302 can divide the real-time sound signal into multiple sound signal segments according to a preset duration; the feature vector acquisition unit 303 can acquire the initial feature vector of each sound signal segment; the diagnosis result determination unit 304 can use the initial feature vector of each sound signal segment and a preset transformer fault diagnosis model to determine the target fault diagnosis result.

[0059] As an example, the diagnosis result determination unit 304 can input the initial feature vectors of multiple sound signal segments into the fault diagnosis model to obtain multiple initial fault diagnosis results; count the multiple initial fault diagnosis results, and use the initial fault diagnosis result with the largest count value as the target fault diagnosis result.

[0060] The transformer fault diagnosis device 300 may further include a model training unit 305. The model training unit 305 can acquire the historical sound signal of the transformer, where the historical sound signal includes abnormal sound signals and normal sound signals; process the normal sound signal according to the historical fault sound signal of the transformer and a preset noise signal to obtain a fault sound signal and a noisy sound signal; perform feature extraction on the fault sound signal, the noisy sound signal, and the abnormal sound signal to obtain fault training features, noisy training features, and abnormal training features; use the fault training features, the noisy training features, and the abnormal training features to train a preset single-classification model to obtain a preset transformer fault diagnosis model.

[0061] As an example, the model training unit 305 can acquire the historical operating condition data of the transformer, where the historical operating condition data includes the weather data and the operating state data of the substation where the transformer is located, and the acquisition time of the historical operating condition data corresponds to the acquisition time of the historical sound signal; correct the abnormal sound signals in the historical sound signal corresponding to the specific historical operating condition data in the historical operating condition data to normal sound signals.

[0062] As an example, the model training unit 305 may adjust the spectrum of the normal sound signal according to the historical fault sound signal to obtain a fault sound signal; superimpose the preset noise data and the normal sound signal to obtain a noise-added sound signal.

[0063] As an example, the model training unit 305 may determine the frequency range and amplitude range of the corresponding fault spectrum according to the historical fault sound signal; perform time-frequency conversion on the normal sound signal to obtain a first spectrum; adjust the amplitude of the first spectrum within the frequency range according to the frequency range and amplitude range of the fault spectrum to obtain a second spectrum; perform inverse time-frequency conversion on the second spectrum to obtain a fault sound signal.

[0064] As an example, the model training unit 305 may perform time-frequency conversion on the normal sound signal and the preset noise signal respectively, and determine the maximum amplitude of the spectrum of the normal sound signal and the maximum amplitude of the spectrum of the preset noise signal; calculate the ratio of the maximum amplitude of the spectrum of the normal sound signal to the maximum amplitude of the spectrum of the preset noise signal; scale the preset noise signal according to the ratio to obtain a first sound signal; superimpose the normal sound signal and the first sound signal to obtain a noise-added sound signal.

[0065] As an example, the model training unit 305 may divide the fault sound signal, the noise-added sound signal, and the abnormal sound signal into multiple fault sound signal segments, multiple noise-added sound signal segments, and multiple abnormal sound signal segments respectively according to a preset duration; for each sound signal segment, perform the following steps: perform frame division on the sound signal segment to obtain multiple frames of data; perform windowing processing on each frame of data according to a preset window and a preset window step size to obtain windowed data; perform time-frequency transformation on the windowed data to obtain a spectrum sequence; use a Mel filter bank to extract features from the spectrum sequence to obtain the training features of the sound signal segment.

[0066] Regarding the device in the above embodiments, the specific manners in which each unit performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0067] Figure 4 is a block diagram showing an electronic device according to an exemplary embodiment of the present disclosure. Referring to Figure 4 , the electronic device 400 includes: a processor 401; a memory 402, where the memory 402 stores computer-executable instructions, and when the computer-executable instructions are run by the processor 401, the processor 401 is caused to execute a transformer fault diagnosis method according to an exemplary embodiment of the present disclosure.

[0068] As an example, the electronic device 400 can be a PC computer, a tablet device, a personal digital assistant, a smart phone, or other devices capable of executing the above instruction set. Here, the electronic device 400 does not have to be a single electronic device, but can also be an aggregate of devices or circuits that can execute the above instructions (or instruction sets) individually or jointly. The electronic device 400 can also be part of an integrated control system or system manager, or can be configured as a portable electronic device that interfaces with a local or remote (e.g., via wireless transmission). In addition, the electronic device 400 can also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, a mouse, a touch input device, etc.). All components of the electronic device 400 can be connected to each other via a bus and / or a network.

[0069] The processor 401 can include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, the processor 401 can also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, etc.

[0070] The processor 401 can run instructions or code stored in the memory 402, where the memory 402 can also store data. The instructions and data can also be sent and received via a network interface device over a network, where the network interface device can employ any known transmission protocol.

[0071] The memory 402 can be integrated with the processor 401, for example, by arranging RAM or flash memory within an integrated circuit microprocessor, etc. In addition, the memory 402 can include a separate device, such as an external disk drive, a storage array, or other storage devices that can be used by any database system. The memory 402 and the processor 401 can be operatively coupled, or can communicate with each other, for example, via an I / O port, a network connection, etc., such that the processor 401 can read files stored in the memory 402.

[0072] According to an embodiment of the present disclosure, there is provided a computer-readable storage medium, which, when the instructions in the computer-readable storage medium are run by a processor, causes the processor to execute a transformer fault diagnosis method according to an exemplary embodiment of the present disclosure.

[0073] The transformer fault diagnosis method according to an embodiment of the present disclosure can be written as a computer program and stored on a computer-readable storage medium. Examples of computer-readable storage media include: read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-RLTH, BD-RE, Blu-ray or optical disc memory, hard disk drive (HDD), solid state drive (SSD), cartridge memory (such as, multimedia card, secure digital (SD) card or extreme digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk and any other device, any other device configured to store a computer program and any associated data, data files and data structures in a non-transitory manner and provide the computer program and any associated data, data files and data structures to a processor or computer such that the processor or computer can execute the computer program. In one example, the computer program and any associated data, data files and data structures are distributed over a networked computer system such that the computer program and any associated data, data files and data structures are stored, accessed and executed in a distributed manner by one or more processors or computers.

[0074] According to an embodiment of the present disclosure, it is possible to extract a feature vector that can better characterize the mechanical faults of a transformer by using the sound signal of an actually operating transformer through a preset single-classification model, and train a preset transformer fault diagnosis model suitable for different regions.

[0075] According to an embodiment of the present disclosure, when an abnormality occurs in the internal components of a transformer, the transformer can be fault diagnosed, and the fault condition of the transformer can be determined and identified, and the diagnosis result can be returned to the monitoring system to achieve real-time diagnosis of transformer abnormalities, reduce the probability of major faults occurring in the equipment, and at the same time reduce the personnel operation and maintenance costs and improve safety.

[0076] The specific embodiments of the present disclosure have been described in detail above. Although some embodiments have been shown and described, those skilled in the art should understand that these embodiments can be modified and varied without departing from the principles and spirit of the present disclosure defined by the claims and their equivalents, and these modifications and variations should also be within the protection scope of the claims of the present disclosure.

Claims

1. A transformer fault diagnosis method, characterized in that, The transformer fault diagnosis method includes: Obtaining the real-time sound signal of the transformer; Dividing the real-time sound signal into multiple sound signal segments according to a preset duration; Obtaining the initial feature vector of the sound signal segment; Using the initial feature vector and a preset transformer fault diagnosis model to determine the target fault diagnosis result.

2. The fault diagnosis method according to claim 1, wherein, The step of using the initial feature vector and a preset transformer fault diagnosis model to determine the target fault diagnosis result includes: Inputting the initial feature vectors of the multiple sound signal segments into the fault diagnosis model to obtain multiple initial fault diagnosis results; Counting the multiple initial fault diagnosis results, and taking the initial fault diagnosis result with the largest count value as the target fault diagnosis result.

3. The fault diagnosis method according to claim 1, wherein The preset transformer fault diagnosis model is obtained through the following steps: Obtaining the historical sound signal of the transformer, where the historical sound signal includes abnormal sound signals and normal sound signals; Processing the normal sound signal according to the historical fault sound signal of the transformer and a preset noise signal to obtain a fault sound signal and a noise-added sound signal; Performing feature extraction on the fault sound signal, the noise-added sound signal, and the abnormal sound signal to obtain fault training features, noise-added training features, and abnormal training features; Using the fault training features, the noise-added training features, and the abnormal training features to train a preset single-classification model to obtain the preset transformer fault diagnosis model.

4. The fault diagnosis method according to claim 3, wherein The step of processing the normal sound signal according to the historical fault sound signal of the transformer and a preset noise signal to obtain a fault sound signal and a noise-added sound signal includes: Adjusting the frequency spectrum of the normal sound signal according to the historical fault sound signal to obtain the fault sound signal; Performing data superposition on the preset noise data and the normal sound signal to obtain the noise-added sound signal.

5. The fault diagnosis method according to claim 4, characterized in that, The step of adjusting the frequency spectrum of the normal sound signal according to the historical fault sound signal to obtain the fault sound signal includes: Determining the frequency range and amplitude range of the corresponding fault frequency spectrum according to the historical fault sound signal; Performing time-frequency conversion on the normal sound signal to obtain a first frequency spectrum; Adjusting the amplitude of the first frequency spectrum within the frequency range according to the frequency range and amplitude range of the fault frequency spectrum to obtain a second frequency spectrum; Performing inverse time-frequency conversion on the second frequency spectrum to obtain the fault sound signal.

6. The fault diagnosis method according to claim 4, wherein, The step of performing data superposition on the preset noise data and the normal sound signal to obtain the noise-added sound signal includes: Performing time-frequency conversion on the normal sound signal and the preset noise signal respectively, and determining the maximum amplitude of the frequency spectrum of the normal sound signal and the maximum amplitude of the frequency spectrum of the preset noise signal; Calculating the ratio of the maximum amplitude of the frequency spectrum of the normal sound signal to the maximum amplitude of the frequency spectrum of the preset noise signal; Scaling the preset noise signal according to the ratio to obtain a first sound signal; Superposing the normal sound signal and the first sound signal to obtain the noise-added sound signal.

7. The fault diagnosis method according to claim 3, characterized in that, The steps of extracting features from the fault sound signal, the noise-added sound signal, and the abnormal sound signal to obtain fault training features, noise-added training features, and abnormal training features include: Dividing the fault sound signal, the noise-added sound signal, and the abnormal sound signal into multiple fault sound signal segments, multiple noise-added sound signal segments, and multiple abnormal sound signal segments respectively according to a preset duration; Performing the following steps for each sound signal segment: Performing frame division on the sound signal segment to obtain multiple frames of data; Performing windowing on each frame of data according to a preset window and a preset window sliding step length to obtain windowed data; Performing time-frequency transformation on the windowed data to obtain a frequency spectrum sequence; Extracting features from the frequency spectrum sequence by using a Mel filter bank to obtain the training features of the sound signal segment.

8. A transformer fault diagnosis device, characterized in that, The transformer fault diagnosis device includes: A signal acquisition unit configured to acquire the real-time sound signal of the transformer; A signal division unit configured to divide the real-time sound signal into multiple sound signal segments according to a preset duration; A feature vector acquisition unit configured to acquire the initial feature vector of each sound signal segment; A diagnosis result determination unit configured to determine a target fault diagnosis result by using the initial feature vector of each sound signal segment and a preset transformer fault diagnosis model.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the transformer fault diagnosis method according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that, Including: At least one processor; At least one memory storing a computer program, and when the computer program is executed by the at least one processor, the transformer fault diagnosis method according to any one of claims 1 to 7 is implemented.

Images