A transformer residual magnetism detection method, device, system and storage medium
Through the detection method based on voiceprints, using deep learning models and signal processing technology, real-time detection of the residual magnetron of the transformer is achieved without shutdown, solving the problems of high detection environment requirements, poor versatility, and shutdown in the existing technology, and improving detection Efficiency and safety.
Patent Information
- Application Number
- CN202111320746.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-09
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-11-09
AI Technical Summary
The prior art is difficult to effectively detect and eliminate residual magnetism in the transformer, resulting in overheating of the iron core, increasing reactive consumption, and damage to the relay protection. The detection environment requirements are high, poor versatility is poor, and shutdown is required for inspection, which is costly.
The soundprint-based detection method is adopted to obtain the sound signal during the operation of the transformer, and use the audio acquisition equipment and deep learning models or signal processing models to determine whether there is residual magnetism in the transformer, real-time detection without shutdown is achieved.
Real-time, no shutdown detection of the transformer's residual magnetism is achieved, which reduces detection costs, improves the universality and efficiency of detection, prompt early warning and demagnetization treatment, and extends the service life of the transformer.
Smart Images

Figure CN114019434B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of power equipment detection, and in particular to a transformer residual magnetism detection method, device, system and storage medium. Background Art
[0002] When the transformer is running, the residual magnetism in the iron core of the transformer causes the iron core to generate a large amount of harmonics in the excitation current to form an excitation surge current. The excitation surge current will increase the reactive power consumption of the transformer and may also cause false triggering or even damage to the relay protector. In addition, the high saturation of the iron core increases the leakage flux, causing overheating of metal structural parts and the oil tank. Local overheating will cause aging of the insulating paper and decomposition of the transformer oil, affecting the life of the transformer. Therefore, in order to ensure the safe operation of the winding and the stable operation of the power system, it is necessary to pay attention to the impact of the residual magnetism of the transformer. Detecting whether there is residual magnetism in the transformer is an urgent problem to be solved. Summary of the invention
[0003] The present application provides a transformer residual magnetism detection method, device, system and storage medium, which can realize residual magnetism detection of the transformer without stopping the machine for detection.
[0004] In order to solve the above technical problems, the technical solution adopted in the present application is: to provide a transformer residual magnetism detection method based on voiceprint, the method comprising: obtaining a first sound signal when the transformer is working; detecting the first sound signal to obtain a detection result of whether the transformer has residual magnetism.
[0005] In order to solve the above technical problems, another technical solution adopted in the present application is: to provide a residual magnetism detection device, which comprises a memory and a processor connected to each other, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, it is used to implement the transformer residual magnetism detection method based on voiceprint in the above technical solution.
[0006] In order to solve the above technical problems, another technical solution adopted in the present application is: to provide a residual magnetism detection system, which includes a residual magnetism detection device and an audio acquisition device, the audio acquisition device is used to collect the sound signal when the transformer is working, and send the sound signal to the residual magnetism detection device, so that the residual magnetism detection device can detect whether there is residual magnetism in the transformer based on the sound signal, wherein the residual magnetism detection device is the residual magnetism detection device in the above technical solution.
[0007] In order to solve the above technical problems, another technical solution adopted in the present application is: providing a computer-readable storage medium, which is used to store a computer program. When the computer program is executed by a processor, it is used to implement the transformer residual magnetism detection method based on voiceprint in the above technical solution.
[0008] Through the above scheme, the beneficial effects of the present application are: the present application adopts a method for detecting whether a transformer has residual magnetism based on voiceprint, first obtaining a first sound signal when the transformer is running; then detecting the first sound signal to obtain a detection result of whether the transformer has residual magnetism, which is convenient for subsequent timely early warning when residual magnetism is detected in the transformer, and convenient for relevant personnel to demagnetize the transformer in time to alleviate the adverse effects caused by residual magnetism; the residual magnetism detection scheme provided by the present application can determine whether the transformer has residual magnetism through the sound signal when the transformer is running, has low requirements for the detection environment, and has good versatility; and there is no need to shut down for detection, and real-time residual magnetism detection of the transformer can be realized, which is easy to use and low in cost. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0010] Figure 1 It is a flow chart of an embodiment of a transformer residual magnetism detection method based on voiceprint provided by the present application;
[0011] Figure 2 It is a flow chart of another embodiment of a transformer residual magnetism detection method based on voiceprint provided by the present application;
[0012] Figure 3 yes Figure 2 A schematic diagram of the flow chart of step 23 in the embodiment shown;
[0013] Figure 4 yes Figure 3 A schematic diagram of the flow chart of step 31 in the embodiment shown;
[0014] Figure 5 is a schematic diagram of a first reconstruction error distribution and a second reconstruction error distribution provided by the present application;
[0015] Figure 6 yes Figure 3 Another schematic diagram of the process of step 31 in the embodiment shown;
[0016] Figure 7 It is a structural schematic diagram of an embodiment of a residual magnetism detection device provided by the present application;
[0017] Figure 8 It is a structural schematic diagram of an embodiment of a residual magnetism detection system provided by the present application;
[0018] Fig. 9It is a structural schematic diagram of an embodiment of a computer-readable storage medium provided by the present application. DETAILED DESCRIPTION
[0019] The present application is further described in detail below in conjunction with the accompanying drawings and examples. It is particularly noted that the following examples are only used to illustrate the present application, but are not intended to limit the scope of the present application. Similarly, the following examples are only some embodiments of the present application rather than all embodiments, and all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.
[0020] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0021] It should be noted that the terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined as "first", "second" and "third" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.
[0022] The winding DC resistance test, which is essential in the routine test of transformer condition inspection, will cause residual magnetism to remain in the transformer core. Therefore, it is necessary to detect whether there is residual magnetism before or during the operation of the transformer. There are many technical solutions for residual magnetism detection and elimination in related technologies. Commonly used residual magnetism detection solutions include the following solutions:
[0023] ① Use a residual magnetism detection device that does not need to invade the inside of the transformer. The residual magnetism detection device can be a device based on harmonic measurement. It uses a fast Fourier transform (FFT) algorithm and an oscilloscope with FFT decomposition function to perform Fourier decomposition and waveform recording on the transformer no-load current to obtain a distribution diagram of the harmonic content including even harmonics. Use the distribution diagram to determine whether the transformer has residual magnetism.
[0024] ② By inputting a sweep frequency signal into the high-voltage winding of the transformer to check the frequency response of the low-voltage winding, an amplitude-frequency response curve is obtained. By comparing the amplitude-frequency response curve with the reference amplitude-frequency response curve in the state without residual magnetism, it can be intuitively determined whether there is residual magnetism in the iron core of the transformer.
[0025] For the above-mentioned technical solution ①, the residual magnetism detection equipment has high requirements for the detection environment and has the problem of poor versatility. It is difficult to be compatible with transformers of different brands and models, and the entire detection process is time-consuming and inefficient. For the above-mentioned technical solution ②, since it is necessary to invade the interior of the transformer, it is necessary to shut down for detection, which has limitations and high detection costs. In order to solve these problems, the present application provides a method for detecting whether there is residual magnetism in a transformer based on voiceprints. The technical solution adopted in the present application is described in detail below.
[0026] See also Figure 1 , Figure 1 : is a flow chart of an embodiment of a transformer residual magnetism detection method based on voiceprint provided by the present application, the method comprising:
[0027] Step 11: Obtain a first sound signal when the transformer is working.
[0028] An audio acquisition device is used to perform audio acquisition processing on the transformer working in the current environment to obtain the sound signal of the transformer (i.e., the first sound signal) to achieve voiceprint acquisition; specifically, the audio acquisition device can be a sensor, which is deployed at the winding of the transformer. The sensor can obtain the sound signal of the transformer and transmit it to the detection platform (i.e., the residual magnetism detection device) through the message queue telemetry transmission protocol (Message Queuing Telemetry Transport, MQTT). The residual magnetism detection device can be a computer or a cloud server or other device with processing capabilities.
[0029] Step 12: Detect the first sound signal to obtain a detection result of whether there is residual magnetism in the transformer.
[0030] After acquiring the first sound signal, the residual magnetism detection device may detect the first sound signal to determine whether there is residual magnetism in the current transformer, and generate a detection result, which may be that there is residual magnetism in the transformer or that there is no residual magnetism in the transformer. Specifically, the currently acquired first sound signal may be compared with a sound signal when there is no residual magnetism in the transformer (recorded as a reference sound signal without residual magnetism) to determine whether there is residual magnetism in the transformer through the first sound signal.
[0031] In a specific embodiment, when residual magnetism exists in the transformer, early warning processing can be performed based on the detection result; specifically, if the residual magnetism detection device determines that residual magnetism exists in the current transformer, the detection result of the residual magnetism of the transformer is directly displayed, or an early warning is issued to remind relevant personnel to promptly perform residual magnetism elimination processing, thereby eliminating safety hazards and reducing the energy consumption of the transformer; specifically, the reminder method can be to generate a warning message, send the warning message to an external electronic device (such as: a mobile terminal or a computer and other devices), or display the warning message, or play the warning message, or call a preset contact (such as: an operator), etc.
[0032] It can be understood that in other embodiments, the sound signal of the transformer when working can be collected at preset intervals to reduce the processing burden of the residual magnetism detection device; for example: after the residual magnetism detection device and the audio acquisition device are deployed, the first sound signal can be obtained at intervals of 60 seconds, for example, the sound signal of the 0th to 10th seconds in the current minute, the sound signal of the 0th to 10th seconds in the next minute, and so on.
[0033] This embodiment proposes a residual magnetism detection method for a power transformer based on voiceprint, which combines the characteristics of the power transformer, and obtains a first sound signal when the transformer is running by pointing the audio acquisition device to the winding of the transformer. By comparing the first sound signal with the reference sound signal without residual magnetism in the state without residual magnetism, it can be intuitively judged whether there is residual magnetism in the iron core of the transformer; since there are differences in the sound signals of the transformer in the residual magnetism state and the sound signal without residual magnetism in the state without residual magnetism, by learning the characteristics of the sound signal without residual magnetism in the state without residual magnetism of the transformer, it can be smoothly judged whether there is residual magnetism in the transformer according to the first sound signal currently obtained, and since the requirements for the detection environment are not high, it has good versatility; in addition, it is sufficient to detect whether there is residual magnetism in real time during the operation of the transformer, without stopping the machine for detection, and it is easy to use.
[0034] See also Figure 2 , Figure 2 : is a flow chart of another embodiment of a transformer residual magnetism detection method based on voiceprint provided by the present application, the method comprising:
[0035] Step 21: Acquire a first sound signal when the transformer is working.
[0036] Step 21 is the same as step 11 in the above embodiment and will not be described again.
[0037] Step 22: Perform feature extraction processing on the first sound signal to obtain the current voiceprint feature.
[0038] A feature extraction method is used to perform feature extraction processing on the new data currently acquired (i.e., the first sound signal) to generate the current voiceprint features; specifically, the current voiceprint features include short-time energy spectrum, spectral entropy or Mel-Frequency Cepstral Coefficients (MFCC). The feature extraction method is a method in the relevant technology that can realize the feature extraction function, which will not be repeated here.
[0039] Step 23: Use the residual magnetism detection model to detect the current voiceprint features to obtain the detection results.
[0040] The residual magnetism detection model includes a deep learning model or a signal processing model. The deep learning model is a network model constructed using a deep learning method. The signal processing model is a model that performs statistics or other processing on the input data. It can be a non-machine learning model, that is, there is no need to train the signal processing model. Specifically, it can be used Figure 3 The scheme shown in the figure is used to obtain the detection results, including the following steps:
[0041] Step 31: Based on the current voiceprint feature, the preset voiceprint library or the output result of the residual magnetism detection model, determine whether the current voiceprint feature meets the preset residual magnetism condition.
[0042] After obtaining the voiceprint feature corresponding to the first sound signal (ie, the current voiceprint feature), the current voiceprint feature, the pre-established preset voiceprint library or the output result of the residual magnetism detection model can be combined to determine whether there is residual magnetism in the current transformer.
[0043] In a specific embodiment, the deep learning model is an autoencoding model, which includes an encoder and a decoder. Figure 4 The scheme shown in the figure determines whether the current voiceprint feature meets the preset residual magnetism condition, including the following steps:
[0044] Step 41: Use an encoder to encode the current voiceprint feature to obtain a coded feature.
[0045] The current voiceprint feature is input into the encoder so that the encoder encodes the current voiceprint feature to generate a coding feature, and then the coding feature is input into the decoder.
[0046] Step 42: Use a decoder to decode the encoded features to obtain reconstructed voiceprint features.
[0047] After receiving the encoded feature, the decoder decodes the encoded feature to generate a feature obtained by reconstructing the input current voiceprint feature (ie, the reconstructed voiceprint feature).
[0048] Step 43: Based on the reconstructed voiceprint features and the preset voiceprint library, determine whether the current voiceprint features meet the preset residual magnetism conditions.
[0049] The preset voiceprint library includes the first test sample data, and the first test sample data includes a plurality of voiceprint features without residual magnetism, and the voiceprint features without residual magnetism are voiceprint features corresponding to the sound generated when the transformer is running without residual magnetism; all voiceprint features without residual magnetism can be used to determine the reference voiceprint features without residual magnetism; then the error between the reference voiceprint features without residual magnetism and the reconstructed voiceprint features is calculated to obtain the current reconstruction error (i.e., singularity); then it is determined whether the current reconstruction error is greater than the preset error threshold, and the preset error threshold is used by the residual magnetism detection device to judge the newly input first sound signal; if the current reconstruction error is greater than the preset error threshold, it indicates that the current reconstruction error is large, and at this time it can be determined that the current voiceprint feature meets the preset residual magnetism condition; if the current reconstruction error is less than / equal to the preset error threshold, it indicates that the current reconstruction error is small, and at this time it is determined that the current voiceprint feature does not meet the preset residual magnetism condition. Specifically, the preset error threshold can be a threshold set according to experience or actual application requirements, or determined by the training process of the autoencoder model.
[0050] Furthermore, a non-residual magnetism voiceprint feature can be directly randomly selected from all non-residual magnetism voiceprint features as a reference non-residual magnetism voiceprint feature; or a self-encoding model can be used to process all non-residual magnetism voiceprint features to obtain a reference non-residual magnetism voiceprint feature, for example: a latent space encoding method with memory slot storage is used to learn the non-residual magnetism voiceprint features of historical input, and the main components of the non-residual magnetism voiceprint features are stored as reference non-residual magnetism voiceprint features. Furthermore, in the process of inputting the non-residual magnetism voiceprint feature into the self-encoding model each time, when the decoder is used for decoding processing, the encoding feature corresponding to the non-residual magnetism voiceprint feature is decoded in combination with the reference non-residual magnetism voiceprint feature to generate the corresponding decoded feature.
[0051] In a specific implementation, the training process of the autoencoder model is as follows:
[0052] 1) Construct the first test sample data.
[0053] Acquire the sound signal of the transformer in the no-residual magnetism state, and form all the sound signals into a data set; then filter the data set to obtain the filtered no-residual magnetism sound signal; extract and process the filtered no-residual magnetism sound signal to obtain the voiceprint feature of the no-residual magnetism sound signal (including short-time energy spectrum, spectral entropy or MFCC, etc.) to construct the first test sample data, which includes the no-residual magnetism sound signal and the no-residual magnetism voiceprint feature corresponding to the no-residual magnetism sound signal.
[0054] Furthermore, since equipment such as transformers are deployed outdoors, there may be a variety of sounds in the environment where the equipment is located. In order to improve the effect of the autoencoder model, the sound signals with large noise in the data set can be removed and relatively pure sound signals can be screened out, thereby ensuring the effectiveness of the data required to train the autoencoder model.
[0055] 2) Inputting the first test sample data into the autoencoder model to train the autoencoder model.
[0056] The first test sample data is used to perform self-supervisory training on the autoencoder model to learn the characteristics of the sound signal in the no-residual magnetization state; after the autoencoder model is trained with the first test sample data, the autoencoder model can output the benchmark no-residual magnetization voiceprint characteristics.
[0057] In a specific embodiment, the preset voiceprint library also includes second test sample data, the second test sample data includes multiple residual magnetic voiceprint features, and the residual magnetic voiceprint features are features corresponding to the sound signals generated when the transformer is operating in a residual magnetic state; an autoencoder model is used to process all non-residual magnetic voiceprint features to obtain a first reconstruction error distribution; an autoencoder model is used to process all residual magnetic voiceprint features to obtain a second reconstruction error distribution; and a preset error threshold is determined based on the first reconstruction error distribution and the second reconstruction error distribution.
[0058] Furthermore, each non-residual magnetic voiceprint feature is input into the self-encoding model training to obtain the corresponding first reconstruction feature; the error between each non-residual magnetic voiceprint feature and the corresponding first reconstruction feature is calculated to obtain the first reconstruction error; all first reconstruction errors are counted to obtain the first reconstruction error distribution; each residual magnetic voiceprint feature is input into the self-encoding model training to obtain the corresponding second reconstruction feature; the error between each residual magnetic voiceprint feature and the corresponding second reconstruction feature is calculated to obtain the second reconstruction error; all second reconstruction errors are counted to obtain the second reconstruction error distribution. Specifically, the first reconstruction error and the second reconstruction error respectively conform to certain distributions and can be used to assist in identification and alarm. The first reconstruction error distribution is the relationship between the first reconstruction error and the number / probability corresponding to the first reconstruction error, and the second reconstruction error distribution is the relationship between the second reconstruction error and the number / probability corresponding to the second reconstruction error. The error that can distinguish the first reconstruction error distribution from the second reconstruction error distribution is used as the preset error threshold. For example: Figure 5 As shown, the first reconstruction error distribution is recorded as A, and the second reconstruction error distribution is recorded as B. Since there is no overlapping area between the first reconstruction error distribution A and the second reconstruction error distribution B, the error C falling between the first reconstruction error distribution A and the second reconstruction error distribution B can be used as the preset error threshold. It can be understood that when there is an overlapping area between the first reconstruction error distribution and the second reconstruction error distribution, a certain reconstruction error in the overlapping area can be used as the preset error threshold, and then the self-encoding model is verified using the verification sample data in the preset voiceprint library. When the verified false alarm rate (i.e., the probability that a sound signal without residual magnetism is judged to have residual magnetism) is less than the preset false alarm rate, the verification is stopped; when the verified false alarm rate is greater than the preset false alarm rate, the value of the preset error threshold is adjusted, and the verification is continued until the verified false alarm rate is less than the preset false alarm rate.
[0059] In another specific embodiment, the first sound signal can be processed by using a signal processing model in combination with the voiceprint short-term stabilization technology. Figure 6 The scheme shown specifically includes the following steps:
[0060] Step 61: Obtain the energy mean of the first sound signal.
[0061] The energy mean of the first sound signal can be directly calculated using the short-time energy spectrum in the current voiceprint feature; or the first sound signal can be divided according to frequency to obtain second sound signals in different frequency bands; the energy values of the second sound signals in different frequency bands can be obtained to obtain sub-energy values; all sub-energy values can be averaged to obtain the energy mean of the first sound signal; or the energy value of the first sound signal in a preset frequency band can be calculated to obtain the energy mean, where the preset frequency band is a frequency band where residual magnetism may appear as a voiceprint, such as 4KHz to 8KHz.
[0062] Step 62: Determine whether the difference between the energy mean value and the reference energy value exceeds a preset difference.
[0063] The reference energy value is the energy mean corresponding to the reference no-residual-magnetism voiceprint feature. By comparing the energy mean of the first sound signal with the energy mean corresponding to the reference no-residual-magnetism voiceprint feature, it is determined whether the current transformer has residual magnetism.
[0064] Step 63: If the difference between the energy mean and the reference energy value exceeds a preset difference, it is determined that the current voiceprint feature meets the preset residual magnetism condition.
[0065] If the difference between the energy mean of the first sound signal detected and the energy mean corresponding to the baseline no-residual magnetism voiceprint feature exceeds the preset difference, it indicates that the difference between the two is large. At this time, it can be determined that the current voiceprint feature meets the preset residual magnetism condition, that is, there is residual magnetism in the transformer; if the difference between the energy mean of the first sound signal detected and the energy mean corresponding to the baseline no-residual magnetism voiceprint feature does not exceed the preset difference, it indicates that the difference between the two is not large. At this time, it can be determined that the current voiceprint feature does not meet the preset residual magnetism condition, that is, there is no residual magnetism in the transformer; the preset difference is a value set according to experience or application needs, for example: 0, that is, when the energy mean of the first sound signal is different from the energy mean corresponding to the baseline no-residual magnetism voiceprint feature, it is determined that there is residual magnetism in the transformer and an alarm is issued.
[0066] By calculating the energy mean of the first sound signal through the short-time energy spectrum of different frequency bands and comparing the energy mean with the benchmark energy value, the ability to capture instantaneous and abnormal sound signals can be enhanced, thereby achieving more accurate capture of instantaneous voiceprints.
[0067] Step 32: If the current voiceprint feature meets the preset residual magnetism condition, it is determined that the transformer has residual magnetism.
[0068] Step 33: If the current voiceprint feature does not meet the preset residual magnetism condition, it is determined that the transformer does not have residual magnetism.
[0069] Step 24: When residual magnetism exists in the transformer, a warning message is generated, information related to residual magnetism detection is displayed and / or played, and / or information related to residual magnetism detection is sent to an external electronic device.
[0070] The detection-related information includes warning information or detection results, that is, when residual magnetism is detected in the transformer, the detection result can be displayed, played, or sent to a preset external electronic device; or the warning information can be displayed, played, or sent to a preset external electronic device to remind relevant personnel to perform onomatopoeia elimination processing on the transformer to improve the safety of transformer operation and service life.
[0071] It can be understood that all newly acquired data (i.e., sound signals) are input into the deep learning model or the signal processing model. If the detection result of any one of the deep learning model and the signal processing model is that there is residual magnetism in the transformer, then it is determined that there is residual magnetism in the transformer to avoid underreporting and improve the accuracy of residual magnetism detection.
[0072] Step 25: When there is no residual magnetism in the transformer, the transformer information is displayed and stored.
[0073] If it is determined that the transformer does not have residual magnetism, detailed information of the transformer can be displayed and stored, and the detailed information includes the transformer model, sound signal, voiceprint characteristics and the detection result of the absence of residual magnetism.
[0074] In other specific embodiments, in order to facilitate processing, the first sound signal can also be divided to obtain multiple third sound signals; then the third sound signal is input into the residual magnetism detection model, and the specific implementation process is similar to the process of processing the first sound signal in the above embodiment, which will not be repeated here. For example, the length of the first sound signal is 10 seconds. After receiving the first sound signal, the residual magnetism detection device performs a segment extraction process on the first sound signal to divide the 10-second first sound signal into 9 2-second third sound signals, that is, the 0th to 2nd seconds are the first third sound signal, the 1st to 3rd seconds are the second third sound signal, and so on, and finally 9 third sound signals are obtained, and these 9 third sound signals are input into the residual magnetism detection model for residual magnetism detection processing; it can be understood that if the detection result of any of the 9 third sound signals is that the transformer has residual magnetism, it is determined that the transformer has residual magnetism.
[0075] The present embodiment proposes a residual magnetism detection method for a power transformer based on voiceprint, and inputs the currently acquired first sound signal into a deep learning model and / or a signal processing model, so that the deep learning model and / or the signal processing model compares the first sound signal with a reference sound signal with no residual magnetism, and determines whether the first sound signal and the reference sound signal with no residual magnetism are significantly different. If the difference between the two is significantly large, it is determined that there is residual magnetism in the transformer; the scheme adopted in the present embodiment can perform residual magnetism detection when the transformer is running, without stopping the transformer for detection, and can perform real-time online detection; and combining the detection results of the deep learning model with the signal processing model helps to improve the detection accuracy and prevent missed detection.
[0076] See also Figure 7 , Figure 7It is a structural schematic diagram of an embodiment of a residual magnetism detection device provided in the present application. The residual magnetism detection device 70 includes a memory 71 and a processor 72 connected to each other. The memory 71 is used to store a computer program. When the computer program is executed by the processor 72, it is used to implement the transformer residual magnetism detection method based on voiceprint in the above embodiment.
[0077] See also Figure 8 , Figure 8 It is a structural schematic diagram of an embodiment of a residual magnetism detection system provided in the present application. The residual magnetism detection system 80 includes an audio acquisition device 81 and a residual magnetism detection device 82. The audio acquisition device 81 is used to collect sound signals when the transformer is working, and send the sound signals to the residual magnetism detection device 82, so that the residual magnetism detection device 82 detects whether there is residual magnetism in the transformer based on the sound signals, wherein the residual magnetism detection device 82 is the residual magnetism detection device in the above-mentioned embodiment.
[0078] like Figure 8 As shown, the residual magnetism detection system 80 also includes an electronic device 83, which is connected to the residual magnetism detection device 82. The electronic device 83 is used to receive the detection results of the residual magnetism detection device 82 and display them. The electronic device 83 can be a device 81 with a display function such as a mobile terminal or a computer.
[0079] In other embodiments, an electric stethoscope device (not shown in the figure) may be further provided in the residual magnetism detection system 80, and the electric stethoscope device is connected to the audio acquisition device 81 and the residual magnetism detection device 82, and is used to store the sound signal collected by the audio acquisition device 81 and transmit it to the residual magnetism detection device 82; it can be understood that the electric stethoscope device can transmit the sound signal collected by the audio acquisition device 81 to the residual magnetism detection device 82 at preset time intervals.
[0080] See also Fig. 9 , Fig. 9 It is a structural diagram of an embodiment of a computer-readable storage medium provided in the present application. The computer-readable storage medium 90 is used to store a computer program 91. When the computer program 91 is executed by the processor, it is used to implement the transformer residual magnetism detection method based on voiceprint in the above embodiment.
[0081] The computer-readable storage medium 90 may be a server, a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other media that can store program codes.
[0082] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only illustrative, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.
[0083] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0084] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0085] The above descriptions are merely embodiments of the present application and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A transformer residual magnetism detection method based on voiceprint, characterized in that: include: Acquire a first sound signal when the transformer is working; Performing feature extraction processing on the first sound signal to obtain current voiceprint features; Based on the current voiceprint feature, the preset voiceprint library and the output result of the residual magnetism detection model, it is judged whether the current voiceprint feature meets the preset residual magnetism condition to obtain a detection result; the residual magnetism detection model includes an autoencoder model; If yes, it is determined that the transformer has residual magnetism; If not, it is determined that there is no residual magnetism in the transformer; The preset voiceprint library includes a plurality of voiceprint features without residual magnetism and a plurality of voiceprint features with residual magnetism, wherein the voiceprint features without residual magnetism are voiceprint features when there is no residual magnetism in the transformer; The step of determining whether the current voiceprint feature meets the preset residual magnetism condition comprises: Determine a reference no-residual-magnetism voiceprint feature by using all the no-residual-magnetism voiceprint features, and calculate the error between the reference no-residual-magnetism voiceprint feature and the reconstructed voiceprint feature to obtain a current reconstruction error; Using the autoencoder model to process all the non-residual magnetism voiceprint features to obtain a first reconstruction error distribution; Using the autoencoder model to process all the residual magnetic voiceprint features to obtain a second reconstruction error distribution; Determining a preset error threshold based on a reconstruction error in an overlapping area of the first reconstruction error distribution and the second reconstruction error distribution; Determine whether the current reconstruction error is greater than a preset error threshold; If so, it is determined that the current voiceprint feature meets the preset residual magnetism condition.
2. The transformer residual magnetism detection method based on voiceprint according to claim 1 is characterized in that: The autoencoding model includes an encoder and a decoder. The step of judging whether the current voiceprint feature meets the preset residual magnetism condition based on the current voiceprint feature, the preset voiceprint library and the output result of the residual magnetism detection model includes: Using the encoder to encode the current voiceprint feature to obtain an encoded feature; Using the decoder to decode the encoded feature to obtain the reconstructed voiceprint feature; Based on the reconstructed voiceprint feature and the preset voiceprint library, it is determined whether the current voiceprint feature meets the preset residual magnetism condition.
3. The transformer residual magnetism detection method based on voiceprint according to claim 2 is characterized in that: The method further comprises: Input each of the voiceprint features without residual magnetism into the autoencoder model training to obtain a corresponding first reconstruction feature; calculate the error between each of the voiceprint features without residual magnetism and the corresponding first reconstruction feature to obtain a first reconstruction error; and perform statistics on all the first reconstruction errors to obtain a distribution of the first reconstruction errors; Input each of the residual magnetic voiceprint features into the autoencoder model training to obtain a corresponding second reconstruction feature; calculate the error between each of the residual magnetic voiceprint features and the corresponding second reconstruction feature to obtain a second reconstruction error; and perform statistics on all the second reconstruction errors to obtain a distribution of the second reconstruction errors.
4. The transformer residual magnetism detection method based on voiceprint according to claim 1 is characterized in that: The method further comprises: Obtaining an energy mean of the first sound signal; Determine whether the difference between the energy mean and the reference energy value exceeds a preset difference, wherein the reference energy value is the energy mean corresponding to the reference no-residual-magnetic voiceprint feature; If so, it is determined that the current voiceprint feature meets the preset residual magnetism condition.
5. The transformer residual magnetism detection method based on voiceprint according to claim 4 is characterized in that: The step of obtaining the energy mean of the first sound signal comprises: Dividing the first sound signal according to frequency to obtain second sound signals in different frequency bands; Acquire energy values of the second sound signals in the different frequency bands to obtain sub-energy values; All the sub-energy values are averaged to obtain an energy mean value of the first sound signal.
6. The transformer residual magnetism detection method based on voiceprint according to claim 1 is characterized in that: The method further comprises: dividing the first sound signal to obtain a plurality of third sound signals; The third sound signal is input into the residual magnetism detection model.
7. The transformer residual magnetism detection method based on voiceprint according to claim 1 is characterized in that: When residual magnetism exists in the transformer, early warning processing is performed based on the detection result.
8. The transformer residual magnetism detection method based on voiceprint according to claim 7 is characterized in that: The step of performing early warning processing based on the detection result includes: When residual magnetism exists in the transformer, generating residual magnetism detection related information, displaying and / or playing the residual magnetism detection related information, and / or sending the residual magnetism detection related information to an external electronic device, the detection related information including warning information or the detection result; When the transformer has no residual magnetism, information of the transformer is displayed and stored.
9. A residual magnetism detection device, characterized in that: It comprises a memory and a processor connected to each other, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, it is used to implement the transformer residual magnetism detection method based on voiceprint according to any one of claims 1 to 8.
10. A residual magnetism detection system, characterized in that: It includes a residual magnetism detection device and an audio acquisition device, wherein the audio acquisition device is used to collect sound signals when the transformer is working, and send the sound signals to the residual magnetism detection device, so that the residual magnetism detection device detects whether there is residual magnetism in the transformer based on the sound signals, wherein the residual magnetism detection device is the residual magnetism detection device according to claim 9.
11. The residual magnetism detection system according to claim 10, characterized in that: The residual magnetism detection system also includes an electronic device, which is connected to the residual magnetism detection device and is used to receive the detection result of the residual magnetism detection device and display it.
12. A computer-readable storage medium for storing a computer program, characterized in that: When the computer program is executed by a processor, it is used to implement the transformer residual magnetism detection method based on voiceprint according to any one of claims 1 to 8.
