CN115270860A - Transformer abnormity diagnosis method, system and diagnosis equipment - Google Patents
Transformer abnormity diagnosis method, system and diagnosis equipment Download PDFInfo
- Publication number
- CN115270860A CN115270860A CN202210842002.XA CN202210842002A CN115270860A CN 115270860 A CN115270860 A CN 115270860A CN 202210842002 A CN202210842002 A CN 202210842002A CN 115270860 A CN115270860 A CN 115270860A
- Authority
- CN
- China
- Prior art keywords
- voiceprint
- transformer
- vibration
- signal
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L17/00—Speaker identification or verification
- G10L17/04—Training, enrolment or model building
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L17/00—Speaker identification or verification
- G10L17/26—Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
Abstract
The invention provides a transformer abnormity diagnosis method, a system and diagnosis equipment, which relate to the technical field of transformer abnormity diagnosis, and are characterized in that contrast learning is used for obtaining multi-mode feature representation from transformer operation voiceprint and vibration data, and then extracted short-time features are passed through a self-encoder to obtain feature vectors of two types of data; namely a voiceprint signal and a vibration signal; performing feature fusion on the learned voiceprint signal and the learned vibration signal by using a self-encoder, and obtaining a voiceprint vibration fusion feature vector through the extracted short-time features by using the self-encoder; and inputting the voiceprint vibration fusion characteristic vector into an LSTM classification network, and outputting to obtain an abnormal classification result of the transformer. The method is based on a voiceprint vibration signal multi-mode feature fusion technology, combines the running sound signal of the transformer with the vibration signal feature, and remarkably reduces the influence of environmental noise interference, thereby effectively improving the accuracy of transformer abnormity identification.
Description
Technical Field
The invention relates to the technical field of transformer abnormity diagnosis, in particular to a transformer abnormity diagnosis method, a transformer abnormity diagnosis system and diagnosis equipment based on sound and vibration combined signal analysis.
Background
With the gradual expansion of the scale of the ultra-high voltage transmission project, the number of transformer substation equipment is rapidly increased, and the outstanding contradiction between the rapid increase of the workload and the relative shortage of personnel is faced in the operation and maintenance work of the equipment. The number and voltage grade of direct current transmission lines are continuously increased, the direct current magnetic biasing problem generated by alternating current and direct current hybrid transmission is increasingly severe, and the direct current magnetic biasing problem of a three-phase group type single-phase transformer commonly adopted in extra-high voltage transmission engineering is particularly prominent. Because the magnetic hysteresis loop of the iron core has nonlinear characteristics, direct current magnetic bias can cause serious distortion of exciting current, so that the iron core and metal accessories of the power transformer can be locally overheated, further the thermal aging of an insulating material of the transformer is accelerated, the magnetostriction is aggravated, the vibration of the iron core of the transformer is strengthened, and the problems of mechanical structure stability such as the looseness of an iron core clamping piece and the like can be caused if the magnetic hysteresis loop of the iron core is not timely processed.
At present, a voiceprint recognition mode can be adopted for fault diagnosis of the transformer, but the existing transformer fault recognition method based on voiceprint recognition has the following defects that a running sound signal of the transformer is easily interfered by environmental background noises (such as human speaking sound, footstep sound and wind sound) and the like in the process of collecting and transmitting, only a single collected equipment sound signal is subjected to recognition analysis, and the accuracy rate of abnormal transformer recognition is low. In addition, because the number of abnormal equipment labeling samples is small, the conventional self-supervision method directly learns feature representation from label data, the generalization capability of the model is not high, and the label-free learning data cannot be fully utilized.
Disclosure of Invention
The invention provides a transformer abnormity diagnosis method, which is based on a voiceprint vibration signal multi-mode feature fusion technology and combines transformer operation sound signals with vibration signal features to remarkably reduce the influence of environmental noise interference, thereby effectively improving the accuracy of transformer abnormity identification.
The transformer abnormity diagnosis method comprises the following steps:
step one, obtaining a learning multi-mode feature representation from transformer operation voiceprint and vibration data by using contrast learning, and obtaining feature vectors of two types of data by using extracted short-time features through a self-encoder; namely a voiceprint signal and a vibration signal;
secondly, performing feature fusion on the learned voiceprint signals and vibration signals by using a self-encoder, and obtaining voiceprint vibration fusion feature vectors through the extracted short-time features by using the self-encoder;
and thirdly, inputting the voiceprint vibration fusion feature vector into an LSTM classification network, and outputting to obtain an abnormal classification result of the transformer.
It is further noted that, in the step one, a certain short-time characteristic is recorded as
；
Extracting a section of voiceprint and vibration data of the transformer operation, dividing the voiceprint and vibration data into a plurality of overlapped vibrations, and expressing the extracted voiceprint and vibration data by short-time characteristics as
Where (N) is the sequence number of the subframe;
and visualizing the short-time features into a spectrogram, wherein the spectrogram reflects the frequency change condition of the voiceprint vibration signal in continuous time, and the spectrogram features are extracted by utilizing a CNN convolutional neural network.
It is further noted that, in the step one, the comparative learning is used for learning reasonable feature representation from the running voiceprint and vibration data of the transformer;
comparing the running voiceprint and vibration data of the transformer with positive samples and negative samples in a feature space respectively, and calculating feature representation of the samples;
for any sample x, the goal of contrast learning is to learn an encoder f such that:
Wherein
Is a positive sample similar to x and,
is a negative sample dissimilar to x, score is a metric function to measure the similarity between samples;
if the vector inner product is used to calculate the similarity of two samples, the loss function of contrast learning can be expressed as:
Where the corresponding sample x has 1 sample and N-1 negative samples.
It should be further noted that, in the second step,
inputting the extracted voiceprint signal spectrogram into an automatic-supervision learning model, and outputting the voiceprint signal spectrogram as an initial voiceprint feature map;
inputting the extracted vibration signal spectrogram into an automatic supervision learning model, and outputting the vibration signal spectrogram into an initial vibration characteristic diagram;
training an auto-encoder by adopting a reconstruction loss function based on pixel errors;
and respectively inputting the initial voiceprint characteristic diagram and the initial vibration characteristic diagram into a self-encoder F, and outputting the initial voiceprint characteristic diagram and the initial vibration characteristic diagram as fusion characteristic vectors to realize the characteristic fusion of the voiceprint signal characteristic and the vibration signal characteristic.
The initial feature map is the output of the CNN model before the last layer.
It should be further noted that step three further includes: dividing the voiceprint signal characteristics and the vibration signal characteristics into N frames, and obtaining N characteristic vectors through characteristic fusion;
setting the size of the feature vector as P, and inputting data into a matrix of N x P;
and inputting the feature fusion into an LSTM network classifier, and outputting the result as a transformer abnormal category identification result.
It is further noted that, the first step further includes:
constructing a normal state database by collecting the state data of the transformer in a preset period;
acquiring first voice print data by acquiring a first voice signal of a transformer in normal operation in a preset period, and performing denoising and AD conversion; carrying out audio frequency spectrum analysis on the first voiceprint data, extracting first audio features and storing the first audio features in a first audio database; acquiring a second sound signal of the operation of the transformer in real time, and carrying out denoising processing and AD conversion to obtain second acoustic data; carrying out audio frequency spectrum analysis on the second voiceprint data, and extracting a second audio characteristic; and identifying and judging whether a second audio characteristic exists in the first audio database through a neural network: if not, an abnormal alarm signal is sent out.
The invention also provides a transformer abnormity diagnosis system, which comprises: the device comprises an extraction module, a comparison learning module, a self-encoder, an LSTM classification network module and an output display module;
the extraction module is used for extracting a section of voiceprint and vibration data of the operation of the transformer;
the comparison learning module is used for obtaining a learning multi-mode feature representation from the transformer operation voiceprint and vibration data and outputting the extracted short-time feature to the self-encoder;
processing the running voiceprint and vibration data of the transformer by the self-encoder to obtain feature vectors of the two types of data; performing feature fusion on the voiceprint signal and the vibration signal to obtain a voiceprint vibration fusion feature vector;
and the LSTM classification network module processes the voiceprint vibration fusion characteristic vector and outputs an abnormal classification result of the transformer through the output display module.
The invention also provides a diagnosis device for realizing the transformer abnormity diagnosis method, which comprises the following steps:
a memory for storing a computer program and a transformer abnormality diagnosis method;
a processor for executing the computer program and the transformer abnormality diagnosis method to realize the steps of the transformer abnormality diagnosis method;
the display screen is used for displaying the diagnosis process and the diagnosed transformer abnormity classification result;
and the communication module is used for transmitting the information of the diagnosis process and the information of the diagnosed transformer abnormity classification result to the server.
According to the technical scheme, the invention has the following advantages:
according to the transformer abnormity identification method and system based on the voiceprint vibration signals, the voiceprint vibration signal multi-mode feature fusion technology is adopted, the running sound signals and the vibration signal features of the transformer are effectively combined, the influence of environmental noise interference in sound signal analysis is reduced, and therefore the accuracy of transformer abnormity identification is effectively improved.
The feature representation method provided by the invention can learn feature representation from unlabeled abnormal data and normal sample data on the basis of labeled sample data, the features can better represent voiceprint vibration data, the feature representation learned by the method is used for classification model training, and the generalization capability of the model can be effectively improved.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings used in the description will be briefly introduced, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a flow chart of a method for diagnosing transformer anomalies;
FIG. 2 is a schematic diagram of vocal print vibration signal feature fusion;
FIG. 3 is a schematic diagram of a voiceprint vibration fusion feature defect detection network structure;
fig. 4 is a schematic diagram of a transformer abnormality diagnosis system.
Detailed Description
The transformer abnormity diagnosis method provided by the invention is used for diagnosing the sound veins and vibration data of the transformer operation and realizing the diagnosis of the transformer state.
It can be understood that the transformer is a key device in the power system, and the reliability of the transformer is significant to ensure the safe and stable operation of the power grid. For a running transformer, continuous mechanical vibration generated by a winding, an iron core and the like is radiated to the outside through air to form a sound wave signal, and abundant transformer state information is contained. Meanwhile, the sound sensor or the patch type acceleration sensor for collecting sound and vibration signals has the advantages of flexible installation, no interference in running state and the like, so that the detection of the abnormal state of the transformer based on the sound and vibration signals gradually becomes a research hotspot. The invention aims to solve the problems that the traditional signal analysis means is difficult to obtain key characteristic quantities widely applied to various transformers, and an effective fault voiceprint recognition method is not formed.
The invention provides a transformer abnormity diagnosis method, system and diagnosis equipment based on sound and vibration combined signal analysis. The method acquires the running sound data and the relevant information of the equipment through the preset basic data acquisition bottom layer framework, and judges the abnormality according to the running sound data by adopting a preset sound identification model to obtain an abnormality identification result.
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The units and algorithm steps of each example described in the embodiment disclosed in the method for diagnosing transformer abnormality based on combined sound and vibration signal analysis provided by the present invention can be implemented by electronic hardware, computer software, or a combination of the two, and in order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functions in the above description. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The block diagram shown in the attached drawing of the transformer abnormity diagnosis method and system based on sound and vibration combined signal analysis provided by the invention is only a functional entity, and does not necessarily correspond to a physically independent entity. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.
In the transformer abnormity diagnosis method and system based on sound and vibration combined signal analysis provided by the invention, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may also be an electric, mechanical or other form of connection.
The transformer abnormity diagnosis method and system based on sound and vibration combined signal analysis are realized based on diagnosis equipment, and the diagnosis equipment comprises the following steps: a memory for storing a computer program and a transformer abnormality diagnosis method; a processor for executing the computer program and the transformer abnormality diagnosis method to realize the steps of the transformer abnormality diagnosis method; the display screen is used for displaying the diagnosis process and the diagnosed transformer abnormity classification result; and the communication module is used for transmitting the information of the diagnosis process and the information of the diagnosed transformer abnormity classification result to the server.
The diagnostic device interacts with a server over a network to receive or send messages or the like. The terminal device may be a variety of electronic devices having a display screen including, but not limited to, smart phones, tablets, portable and desktop computers, digital computers, and the like.
In particular, the processes described below with reference to the flowcharts may be implemented as computer software programs, according to embodiments of the present disclosure. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable storage medium, the computer program containing program code for performing the method illustrated by the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication section, and/or installed from a removable medium. The computer program, when executed by a Central Processing Unit (CPU), performs various functions defined in the methods and/or apparatus of the present application.
The transformer abnormality diagnosis method executed by the diagnosis device includes: as shown in figure 1 of the drawings, in which,
s101, obtaining a learning multi-mode feature representation from the transformer operation voiceprint and vibration data by using contrast learning, and obtaining feature vectors of two types of data by using the extracted short-time features through a self-encoder; namely a voiceprint signal and a vibration signal;
specifically, the multi-modal characterization of the voiceprint vibration signal is performed in the following manner: with a certain short-term characteristic of
. For a segment of a voiceprint (vibration) signal, which is divided into overlapping vibrations, the short-time characteristics of the segment can be expressed as
Where (N) is the sequence number of the subframe.
And visualizing the short-time frequency domain characteristics into a spectrogram, reflecting the frequency change condition of the voiceprint vibration signal in continuous time, and extracting the characteristics by using a CNN convolutional neural network.
Comparative learning is used to learn reasonable feature representations from voiceprint vibration data. The basic idea of contrast learning is to compute the feature representation of a sample by comparing the data with positive and negative examples samples, respectively, in a feature space. For any data x, the goal of contrast learning is to learn an encoder f such that:
Wherein
Is a positive sample similar to x and,
is and x isSimilar negative samples, score is a metric function to measure the similarity between samples. If the vector inner product is used to calculate the similarity of two samples, the loss function of contrast learning can be expressed as:
In the method of the present invention, there are 1 sample and N-1 negative samples corresponding to sample x. Specifically, in terms of voiceprint vibration data, a subframe is extracted from a small amount of labeled fault data to serve as a negative sample; and extracting any subframe from the unmarked data as a positive sample, thereby forming a pair. And (4) for two frames of input, using the spectrogram as an input feature, and respectively using the CNN network for training. The convolutional layer output before the last layer of the CNN network is expressed as a feature. The method can learn and express from the label-free data, so that the closer the homologous data is in the expression space, the farther the different source data is in the expression space, the more the feature expression of the voiceprint vibration sample is improved, and the accuracy of transformer fault identification is obviously improved.
S102, performing feature fusion on the learned voiceprint signals and the learned vibration signals by using a self-encoder, and obtaining voiceprint vibration fusion feature vectors through the extracted short-time features by using the self-encoder;
as shown in fig. 2, feature fusion is performed on the learned voiceprint and vibration signals using an autoencoder. Obtaining a voiceprint vibration fusion feature vector by the extracted short-time features through a self-encoder, wherein the voiceprint vibration fusion feature vector comprises the following steps:
inputting the extracted voiceprint signal spectrogram into an automatic supervision learning model, and outputting the voiceprint signal spectrogram into an initial voiceprint characteristic diagram; the initial feature map is the output of the CNN model before the last layer.
Inputting the extracted vibration signal spectrogram into an automatic supervision learning model, and outputting the vibration signal spectrogram into an initial vibration characteristic diagram;
training an auto-encoder by adopting a reconstruction loss function based on pixel errors;
and respectively inputting the initial voiceprint characteristic diagram and the initial vibration characteristic diagram into a self-encoder F, and outputting the initial voiceprint characteristic diagram and the initial vibration characteristic diagram as fusion characteristic vectors to realize the characteristic fusion of the voiceprint signal characteristic and the vibration signal characteristic.
The feature vector can simultaneously represent the two original feature graphs due to the characteristics of the self-encoder, and the feature fusion of the voiceprint vibration signal is realized.
And S103, inputting the voiceprint vibration fusion feature vector into an LSTM classification network, and outputting to obtain an abnormal classification result of the transformer.
The LSTM-based abnormity diagnosis classification network is trained on the basis of S102 feature fusion. The LSTM network structure is shown in fig. 3.
Dividing the voiceprint signal characteristics and the vibration signal characteristics into N frames, and obtaining N characteristic vectors through characteristic fusion; setting the size of the feature vector as P, and inputting data into a matrix of N x P; and inputting the feature fusion into an LSTM network classifier, and outputting the result as a transformer abnormal category identification result.
The transformer abnormity diagnosis method based on sound and vibration combined signal analysis provided by the invention fuses the characteristics of the sound veins and the vibration signals of the transformer operation, can effectively reduce noise interference and improve the accuracy of transformer abnormity identification.
According to the method, the voiceprint vibration signal learning characteristic representation is realized by utilizing contrast learning, the voiceprint vibration data can be well represented, and the generalization capability of a classification model is improved.
The invention is based on the LSTM abnormality diagnosis classification network, and can accurately identify the transformer abnormality category.
Based on the above method, the present invention further provides a transformer abnormality diagnosis system, as shown in fig. 4, the system includes: the device comprises an extraction module, a comparison learning module, a self-encoder, an LSTM classification network module and an output display module;
the extraction module is used for extracting a section of voiceprint and vibration data of the operation of the transformer;
the comparison learning module is used for obtaining a learning multi-mode feature representation from the transformer operation voiceprint and vibration data and outputting the extracted short-time feature to the self-encoder;
processing the running voiceprint and vibration data of the transformer by the self-encoder to obtain feature vectors of the two types of data; performing feature fusion on the voiceprint signal and the vibration signal to obtain a voiceprint vibration fusion feature vector;
the LSTM classification network module processes the voiceprint vibration fusion feature vector and outputs an abnormal classification result of the transformer through the output display module.
In the transformer abnormality diagnosis system provided by the invention, the transformer abnormality can be diagnosed by adopting a machine learning method, a deep learning method and the like. The system according to the embodiment of the present invention may be executed by a diagnostic apparatus, but the present invention is not limited thereto.
In the transformer abnormity diagnosis system provided by the invention, aiming at the problem that the current transformer abnormity labeling sample data is less, a method for realizing abnormity early warning by using normal state data to build a database and measuring deviation from a steady state by an error is provided. Acquiring a first sound signal of a transformer in normal operation in a preset period, and performing denoising processing and AD conversion to obtain first voiceprint data; carrying out audio frequency spectrum analysis on the first voiceprint data, extracting first audio frequency characteristics and storing the first audio frequency characteristics in a first audio frequency database; acquiring a second sound signal of the operation of the transformer in real time, and carrying out denoising processing and AD conversion to obtain second acoustic data; carrying out audio frequency spectrum analysis on the second voiceprint data, and extracting a second audio characteristic; and identifying and judging whether a second audio characteristic exists in the first audio database through a neural network: if not, an abnormal alarm signal is sent out.
According to the transformer abnormity identification method and system based on the voiceprint vibration signal, the voiceprint vibration signal multi-mode feature fusion technology is adopted, the transformer operation sound signal and the vibration signal feature are effectively combined, the influence of environmental noise interference in sound signal analysis is reduced, and therefore the transformer abnormity identification accuracy is effectively improved. The feature representation method and the feature representation system can learn feature representation from unlabeled abnormal data and normal sample data on the basis of labeled sample data, the features can well represent voiceprint vibration data, the feature representation learned by the method is used for classification model training, and the generalization capability of the model can be effectively improved.
The transformer abnormity diagnosis method and system based on sound and vibration combined signal analysis provided by the invention are the units and algorithm steps of each example described in combination with the embodiments disclosed herein, and can be realized by electronic hardware, computer software or combination of the two. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the technical solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
Those skilled in the art will appreciate that aspects of the method and system for diagnosing transformer anomalies based on combined acoustic and vibration signal analysis provided by the present invention may be embodied as a system, method, or program product. Accordingly, various aspects of the present disclosure may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" system.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (9)
1. A method for diagnosing transformer abnormality, the method comprising:
step one, obtaining a learning multi-mode feature representation from transformer operation voiceprint and vibration data by using contrast learning, and obtaining feature vectors of two types of data by using extracted short-time features through a self-encoder; namely a voiceprint signal and a vibration signal;
secondly, performing feature fusion on the learned voiceprint signals and vibration signals by using a self-encoder, and obtaining voiceprint vibration fusion feature vectors through the extracted short-time features by using the self-encoder;
and thirdly, inputting the voiceprint vibration fusion feature vector into an LSTM classification network, and outputting to obtain an abnormal classification result of the transformer.
2. The transformer abnormality diagnostic method according to claim 1,
in step one, a certain short-time characteristic is recorded as
；
Extracting a section of voiceprint and vibration data of the transformer operation, dividing the voiceprint and vibration data into a plurality of overlapped vibrations, and expressing the extracted voiceprint and vibration data by short-time characteristics as
Where (N) is the sequence number of the subframe;
and visualizing the short-time features into a spectrogram, wherein the spectrogram reflects the frequency change condition of the voiceprint vibration signal in continuous time, and the spectrogram features are extracted by utilizing a CNN convolutional neural network.
3. The transformer abnormality diagnostic method according to claim 2,
in the first step, the reasonable feature representation is learned from the running voiceprint and vibration data of the transformer by using contrast learning;
comparing the running voiceprint and vibration data of the transformer with positive samples and negative samples in a feature space respectively, and calculating feature representation of the samples;
for any sample x, the goal of contrast learning is to learn an encoder f such that:
Wherein
Is a positive sample similar to x and,
is a negative sample dissimilar to x, score is a metric function to measure the similarity between samples;
if the vector inner product is used to calculate the similarity of two samples, the loss function of contrast learning can be expressed as:
Where the corresponding sample x has 1 sample and N-1 negative samples.
4. The transformer abnormality diagnostic method according to claim 1 or 2,
in the second step, the first step is carried out,
inputting the extracted voiceprint signal spectrogram into an automatic supervision learning model, and outputting the voiceprint signal spectrogram into an initial voiceprint characteristic diagram;
inputting the extracted vibration signal spectrogram into an automatic supervision learning model, and outputting the spectrogram as an initial vibration characteristic diagram;
training an auto-encoder by adopting a reconstruction loss function based on pixel errors;
and respectively inputting the initial voiceprint characteristic diagram and the initial vibration characteristic diagram into a self-encoder F, and outputting the initial voiceprint characteristic diagram and the initial vibration characteristic diagram as fusion characteristic vectors to realize the characteristic fusion of the voiceprint signal characteristic and the vibration signal characteristic.
5. The transformer abnormality diagnostic method according to claim 4,
the initial feature map is the output of the CNN model before the last layer.
6. The transformer abnormality diagnostic method according to claim 1 or 2,
the third step also comprises: dividing the voiceprint signal characteristics and the vibration signal characteristics into N frames, and obtaining N characteristic vectors through characteristic fusion;
setting the size of the feature vector as P, and inputting data into a matrix of N x P;
and inputting the feature fusion into an LSTM network classifier, and outputting the result as a transformer abnormal category identification result.
7. The transformer abnormality diagnostic method according to claim 1 or 2,
the first step also comprises the following steps:
constructing a normal state database by collecting the state data of the transformer in a preset period;
acquiring first vocal print data by acquiring a first sound signal of a transformer in normal operation in a preset period, and performing denoising processing and AD conversion; carrying out audio frequency spectrum analysis on the first voiceprint data, extracting first audio features and storing the first audio features in a first audio database; acquiring a second sound signal of the transformer operation in real time, and carrying out denoising treatment and AD conversion to obtain second acoustic data; carrying out audio frequency spectrum analysis on the second voiceprint data, and extracting a second audio characteristic; and identifying and judging whether a second audio characteristic exists in the first audio database through a neural network: if not, an abnormal alarm signal is sent out.
8. A transformer abnormality diagnosis system, characterized in that the system employs the transformer abnormality diagnosis method according to any one of claims 1 to 7;
the system comprises: the device comprises an extraction module, a comparison learning module, a self-encoder, an LSTM classification network module and an output display module;
the extraction module is used for extracting a section of voiceprint and vibration data of the operation of the transformer;
the comparison learning module is used for obtaining a learning multi-mode feature representation from the transformer operation voiceprint and vibration data and outputting the extracted short-time feature to the self-encoder;
processing the running voiceprint and vibration data of the transformer by using an auto-encoder to obtain the characteristic vectors of the two types of data; performing feature fusion on the voiceprint signal and the vibration signal to obtain a voiceprint vibration fusion feature vector;
and the LSTM classification network module processes the voiceprint vibration fusion characteristic vector and outputs an abnormal classification result of the transformer through the output display module.
9. A diagnostic device for implementing a transformer abnormality diagnostic method is characterized by comprising:
a memory for storing a computer program and a transformer abnormality diagnosis method;
a processor for executing the computer program and the transformer abnormality diagnosis method to realize the steps of the transformer abnormality diagnosis method according to any one of claims 1 to 7;
the display screen is used for displaying the diagnosis process and the diagnosed transformer abnormity classification result;
and the communication module is used for transmitting the diagnostic process information and the diagnosed transformer abnormity classification result information to the server.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210842002.XA CN115270860A (en) | 2022-07-18 | 2022-07-18 | Transformer abnormity diagnosis method, system and diagnosis equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210842002.XA CN115270860A (en) | 2022-07-18 | 2022-07-18 | Transformer abnormity diagnosis method, system and diagnosis equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115270860A true CN115270860A (en) | 2022-11-01 |
Family
ID=83766982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210842002.XA Pending CN115270860A (en) | 2022-07-18 | 2022-07-18 | Transformer abnormity diagnosis method, system and diagnosis equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115270860A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115711669A (en) * | 2022-11-07 | 2023-02-24 | 石家庄宜中机电技术有限公司 | Method, device, terminal and storage medium for diagnosing faults of shaft equipment |
| CN115762558A (en) * | 2022-11-18 | 2023-03-07 | 沃克斯迅达电梯有限公司 | Performance detection system and method for escalator production |
| CN117056814A (en) * | 2023-10-11 | 2023-11-14 | 国网山东省电力公司日照供电公司 | Transformer voiceprint vibration fault diagnosis method |
-
2022
- 2022-07-18 CN CN202210842002.XA patent/CN115270860A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115711669A (en) * | 2022-11-07 | 2023-02-24 | 石家庄宜中机电技术有限公司 | Method, device, terminal and storage medium for diagnosing faults of shaft equipment |
| CN115762558A (en) * | 2022-11-18 | 2023-03-07 | 沃克斯迅达电梯有限公司 | Performance detection system and method for escalator production |
| CN117056814A (en) * | 2023-10-11 | 2023-11-14 | 国网山东省电力公司日照供电公司 | Transformer voiceprint vibration fault diagnosis method |
| CN117056814B (en) * | 2023-10-11 | 2024-01-05 | 国网山东省电力公司日照供电公司 | Transformer voiceprint vibration fault diagnosis method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115270860A (en) | Transformer abnormity diagnosis method, system and diagnosis equipment | |
| Georgoulas et al. | Principal Component Analysis of the start-up transient and Hidden Markov Modeling for broken rotor bar fault diagnosis in asynchronous machines | |
| Scanlon et al. | Residual life prediction of rotating machines using acoustic noise signals | |
| CN114386537B (en) | Lithium battery fault diagnosis method and device based on Catboost and electronic equipment | |
| US20200402221A1 (en) | Inspection system, image discrimination system, discrimination system, discriminator generation system, and learning data generation device | |
| WO2023138337A1 (en) | Motor fault detection method and apparatus | |
| CN114298141A (en) | Power generation equipment abnormity detection method and system based on deep learning | |
| CN114120974A (en) | Fan blade fault diagnosis method based on deep learning | |
| CN111582315B (en) | Sample data processing method and device and electronic equipment | |
| CN116012681A (en) | Method and system for diagnosing motor faults of pipeline robot based on sound vibration signal fusion | |
| CN113986643A (en) | Method, electronic device and computer program product for analyzing log file | |
| Hong et al. | Supervised-learning-based intelligent fault diagnosis for mechanical equipment | |
| Vitor et al. | Induction motor short circuit diagnosis and interpretation under voltage unbalance and load variation conditions | |
| Grandhi et al. | Machine-learning based fault diagnosis of electrical motors using acoustic signals | |
| Boztas | Comparison of acoustic signal–based fault detection of mechanical faults in induction motors using image classification models | |
| CN110321425B (en) | Method and device for judging defect type of power grid | |
| CN114255373B (en) | Sequence anomaly detection method, device, electronic equipment and readable medium | |
| Senanayaka et al. | Online fault diagnosis system for electric powertrains using advanced signal processing and machine learning | |
| CN110489262A (en) | Automatic positioning method, system, medium and the electronic equipment of system fault module | |
| CN115862681B (en) | Rotor rub-impact fault acoustic signal diagnosis method and device and electronic equipment | |
| Soltau et al. | Detecting Speech Abnormalities With a Perceiver-Based Sequence Classifier that Leverages a Universal Speech Model | |
| US11650184B2 (en) | System and method for monitoring rotating equipment | |
| CN115600131B (en) | Automatic test system based on graphical test platform | |
| Koshy et al. | Wavelet based bearing fault prognosis using machine learning in cloud platform | |
| CN114780940B (en) | Cross-system data sharing interaction project operation monitoring and early warning method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |