CN119066521A - Fault diagnosis method and device for switch cabinet, electronic equipment and storage medium - Google Patents
Fault diagnosis method and device for switch cabinet, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN119066521A CN119066521A CN202411186140.2A CN202411186140A CN119066521A CN 119066521 A CN119066521 A CN 119066521A CN 202411186140 A CN202411186140 A CN 202411186140A CN 119066521 A CN119066521 A CN 119066521A
- Authority
- CN
- China
- Prior art keywords
- time domain
- data
- frequency domain
- domain data
- switch cabinet
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 76
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000003860 storage Methods 0.000 title claims abstract description 17
- 239000013598 vector Substances 0.000 claims abstract description 232
- 238000012545 processing Methods 0.000 claims abstract description 84
- 230000004927 fusion Effects 0.000 claims abstract description 79
- 238000000605 extraction Methods 0.000 claims abstract description 37
- 238000006243 chemical reaction Methods 0.000 claims abstract description 16
- 238000012544 monitoring process Methods 0.000 claims abstract description 14
- 239000011159 matrix material Substances 0.000 claims description 86
- 238000013507 mapping Methods 0.000 claims description 52
- 238000012360 testing method Methods 0.000 claims description 49
- 238000012549 training Methods 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 21
- 238000009826 distribution Methods 0.000 claims description 19
- 238000001914 filtration Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 abstract description 8
- 208000028659 discharge Diseases 0.000 description 39
- 238000013136 deep learning model Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 239000000725 suspension Substances 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000007847 structural defect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a fault diagnosis method and device of a switch cabinet, electronic equipment and a storage medium, and relates to the technical field of power system monitoring and fault diagnosis, wherein the method comprises the steps of collecting time domain data of the switch cabinet, and performing frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet; the method comprises the steps of inputting time domain data and frequency domain data into a feature processing model, carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, inputting the time-frequency fusion feature vector into a fault recognition model, outputting the fault type of the switch cabinet through the fault recognition model, and generating a fault diagnosis result of the switch cabinet based on the fault type. The invention solves the technical problems of low diagnosis accuracy in the related art of performing fault diagnosis on the switch cabinet based on the big data technology.
Description
Technical Field
The invention relates to the technical field of power system monitoring and fault diagnosis, in particular to a fault diagnosis method and device of a switch cabinet, electronic equipment and a storage medium.
Background
In an electrical power system, a switchgear plays a vital role in distribution and control of electrical energy as an important electrical device. However, due to long-term operation and environmental factors, a partial discharge fault may occur in the switch cabinet, and such a fault may not only affect the safe and stable operation of the power system, but also may cause serious consequences such as equipment damage and power failure. Therefore, the diagnosis and classification of the partial discharge faults of the switch cabinet are important to the reliability and safety of the power system.
In the related art, the method for diagnosing the partial discharge faults of the switch cabinet mainly relies on experience judgment and manual analysis, has the problems of long diagnosis period, low accuracy and the like, and in an actual working scene, the classification of the partial discharge faults is based on a big data technology, and the method needs to manually analyze massive data by workers to diagnose and classify the partial discharge faults of the switch cabinet, so that the execution efficiency is low, and the current operation requirement of a power system is difficult to meet.
In view of the above problems, no effective solution has been proposed at present.
Disclosure of Invention
The embodiment of the invention provides a fault diagnosis method and device of a switch cabinet, electronic equipment and a storage medium, which at least solve the technical problems of low diagnosis accuracy in the related art of performing fault diagnosis on the switch cabinet based on a big data technology.
According to one aspect of the embodiment of the invention, a fault diagnosis method of a switch cabinet is provided, which comprises the steps of collecting time domain data of the switch cabinet, carrying out frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet, inputting the time domain data and the frequency domain data into a feature processing model, carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, wherein the feature processing model is constructed in advance, inputting the time-frequency fusion feature vector into a fault recognition model, outputting the fault type of the switch cabinet through the fault recognition model, and generating a fault diagnosis result of the switch cabinet based on the fault type.
Optionally, the steps of inputting the time domain data and the frequency domain data into a feature processing model, and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model comprise inputting the time domain data into a time domain encoder of the feature processing model, carrying out embedding mapping on the time domain data through the time domain encoder to obtain a time domain feature vector, inputting the frequency domain data into a frequency domain encoder of the feature processing model, carrying out embedding mapping on the frequency domain data through the frequency domain encoder to obtain a frequency domain feature vector, inputting the time domain feature vector and the frequency domain feature vector into a fusion encoder of the feature processing model, and carrying out fusion on the time domain feature vector and the frequency domain feature vector through the fusion encoder to obtain the time-frequency fusion feature vector.
Optionally, the step of performing embedding mapping on the time domain data through the time domain encoder to obtain time domain feature vectors comprises the steps of performing embedding mapping on the time domain data through the time domain encoder to obtain time domain embedded vectors, dividing the time domain embedded vectors to obtain H attention heads corresponding to the time domain embedded vectors, wherein H is a positive integer, calculating query data, key data and value data of the attention heads corresponding to each time domain embedded vector through a query projection matrix, a key projection matrix and a value projection matrix, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes used for weight distribution, calculating attention weights of the attention heads based on the query data and the key data of the attention heads corresponding to each time domain embedded vector, calculating output vectors of the attention heads based on the attention weights and the value data, and splicing the output vectors of the attention heads corresponding to the H time domain embedded vectors to obtain the feature vectors.
The method comprises the steps of obtaining frequency domain feature vectors, wherein the frequency domain feature vectors are obtained by carrying out embedding mapping on frequency domain data through a frequency domain encoder, obtaining attention heads corresponding to K frequency domain embedding vectors through dividing the frequency domain embedding vectors, wherein K is a positive integer, calculating query data, key data and value data of the attention heads through a query projection matrix, a key projection matrix and a value projection matrix for attention heads corresponding to the frequency domain embedding vectors, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes used for carrying out weight distribution, calculating attention weights of the attention heads based on the query data and the key data of the attention heads for attention heads corresponding to the frequency domain embedding vectors, calculating output vectors of the attention heads based on the attention weights and the value data, and splicing the output vectors of the attention heads corresponding to the K frequency domain embedding vectors to obtain the feature vectors.
The method comprises the steps of receiving a pulse collection device into a switch cabinet, setting a trigger threshold for the pulse collection device, monitoring a current signal of the switch cabinet, recording a capturing time point when the amplitude of the current signal is larger than the trigger threshold, and collecting N time domain pulse signals from the switch cabinet based on a preset pulse duration by taking the capturing time point as a starting point to obtain time domain data, wherein N is a positive integer.
Optionally, the step of performing frequency domain conversion on the time domain data comprises the steps of processing the time domain data based on a Fourier transform algorithm to obtain initial frequency domain data, and performing high-pass filtering on the initial frequency domain data according to a preset cut-off frequency to obtain frequency domain data of the switch cabinet.
The fault recognition model is obtained through pre-training, the step of training the fault recognition model comprises the steps of collecting historical time domain data and fault type information of the switch cabinet in a historical time period, obtaining historical frequency domain data based on the historical time domain data, carrying out feature extraction and feature fusion on the historical time domain data and the historical frequency domain data to obtain historical time-frequency feature vectors, constructing a sample set based on the historical time-frequency feature vectors and the fault type information, dividing the sample set based on a preset dividing proportion to obtain a training set and a test set, training an initial fault recognition model through the test set to obtain a trained fault recognition model, testing the trained fault recognition model based on the test set to obtain a test result, and obtaining the final fault recognition model under the condition that the test result indicates that the fault recognition model passes the test.
According to another aspect of the embodiment of the invention, a fault diagnosis device of a switch cabinet is provided, which comprises an acquisition unit, an extraction unit and a generation unit, wherein the acquisition unit is used for acquiring time domain data of the switch cabinet and performing frequency domain conversion on the time domain data to obtain time domain data and frequency domain data of the switch cabinet, the extraction unit is used for inputting the time domain data and the frequency domain data into a feature processing model and performing feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, the feature processing model is constructed in advance, the output unit is used for inputting the time-frequency fusion feature vector into a fault recognition model, outputting a fault type of the switch cabinet through the fault recognition model, and the generation unit is used for generating a fault diagnosis result of the switch cabinet based on the fault type.
Optionally, the extraction unit comprises a first mapping module, a second mapping module and a first fusion module, wherein the first mapping module is used for inputting the time domain data to a time domain encoder of the feature processing model, embedding and mapping the time domain data through the time domain encoder to obtain a time domain feature vector, the second mapping module is used for inputting the frequency domain data to a frequency domain encoder of the feature processing model, embedding and mapping the frequency domain data through the frequency domain encoder to obtain a frequency domain feature vector, and the first fusion module is used for inputting the time domain feature vector and the frequency domain feature vector to a fusion encoder of the feature processing model, and fusing the time domain feature vector and the frequency domain feature vector through the fusion encoder to obtain the time-frequency fusion feature vector.
The first mapping module comprises a first mapping submodule, a first calculating submodule and a first splicing submodule, wherein the first mapping submodule is used for carrying out embedding mapping on the time domain data through the time domain encoder to obtain time domain embedded vectors, the first dividing submodule is used for dividing the time domain embedded vectors to obtain attention heads corresponding to H time domain embedded vectors, H is a positive integer, the first calculating submodule is used for calculating query data, key data and value data of the attention heads corresponding to each time domain embedded vector through a query projection matrix, a key projection matrix and a value projection matrix, the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes used for carrying out weight distribution, the second calculating submodule is used for calculating attention weights of the attention heads based on the query data and the key data of the attention heads corresponding to each time domain embedded vector and calculating output vectors of the attention heads based on the attention weights and the value data of the attention heads, and the first splicing submodule is used for splicing the attention heads corresponding to the H time domain embedded vectors to obtain the feature vectors.
The second mapping module comprises a second mapping submodule, a second dividing submodule, a third calculating submodule and a second splicing submodule, wherein the second mapping submodule is used for carrying out embedding mapping on the frequency domain data through the frequency domain encoder to obtain frequency domain embedded vectors, the second dividing submodule is used for dividing the frequency domain embedded vectors to obtain attention heads corresponding to K frequency domain embedded vectors, K is a positive integer, the third calculating submodule is used for calculating query data, key data and value data of the attention heads through a query projection matrix, a key projection matrix and a value projection matrix for the attention heads corresponding to each frequency domain embedded vector, the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes used for carrying out weight distribution, the fourth calculating submodule is used for calculating attention weights of the attention heads based on the query data and the key data of the attention heads and calculating output vectors of the attention heads based on the attention weights and the value data of the attention heads for each frequency domain embedded vector, and the second splicing submodule is used for splicing the attention heads corresponding to the frequency domain embedded vectors.
Optionally, the acquisition unit comprises a first setting module, a first monitoring module and a first acquisition module, wherein the first setting module is used for connecting a pulse acquisition device into the switch cabinet and setting a trigger threshold for the pulse acquisition device, the first monitoring module is used for monitoring a current signal of the switch cabinet, and recording a capturing time point when the amplitude of the current signal is larger than the trigger threshold, and the first acquisition module is used for acquiring N time domain pulse signals from the switch cabinet based on a preset pulse duration by taking the capturing time point as a starting point to obtain the time domain data, wherein N is a positive integer.
Optionally, the acquisition unit further comprises a first processing module for processing the time domain data based on a Fourier transform algorithm to obtain initial frequency domain data, and a first filtering module for performing high-pass filtering on the initial frequency domain data according to a preset cut-off frequency to obtain frequency domain data of the switch cabinet.
Optionally, the fault diagnosis device of the switch cabinet further comprises a second acquisition module, a first extraction module, a first construction module, a first training module and a second test module, wherein the second acquisition module is used for acquiring historical time domain data and fault type information of the switch cabinet in a historical time period and acquiring historical frequency domain data based on the historical time domain data, the first extraction module is used for carrying out feature extraction and feature fusion on the historical time-frequency data and the historical frequency domain data to obtain a historical time-frequency feature vector, the first construction module is used for constructing a sample set based on the historical time-frequency feature vector and the fault type information and dividing the sample set based on a preset dividing ratio to obtain a training set and a test set, the first training module is used for training an initial fault recognition model through the test set to obtain a trained fault recognition model, and the first test module is used for carrying out test on the trained fault recognition model based on the test set to obtain a test result, and the final fault recognition model is obtained under the condition that the test result indicates that the fault recognition model passes the test.
According to another aspect of the embodiment of the present invention, there is further provided a computer readable storage medium, where the computer readable storage medium includes a stored computer program, and when the computer program runs, the device where the computer readable storage medium is controlled to execute the fault diagnosis method of any one of the switch cabinets.
According to another aspect of the embodiment of the present invention, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, where the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the fault diagnosis method of any one of the switch cabinets described above.
In the application, firstly, time domain data of a switch cabinet are acquired, frequency domain conversion is carried out on the time domain data to obtain time domain data and frequency domain data of the switch cabinet, the time domain data and the frequency domain data are input into a feature processing model, feature extraction and feature fusion are carried out on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, wherein the feature processing model is constructed in advance, then the time-frequency fusion feature vector is input into a fault identification model, a fault type of the switch cabinet is output through the fault identification model, and finally a fault diagnosis result of the switch cabinet is generated based on the fault type.
According to the application, time domain data of the switch cabinet are acquired, frequency domain data are acquired through the time domain data, time-frequency characteristics are extracted based on the time domain data and the frequency domain data, the time-frequency characteristics can describe time domain and frequency domain characteristics of signals more accurately, so that the diagnosis accuracy and reliability of partial discharge faults are improved, meanwhile, a pre-trained fault recognition model is adopted to analyze the time-frequency characteristics, further the type of the partial discharge faults is recognized, thereby realizing accurate positioning of the faults, improving the accuracy of the diagnosis of the partial discharge faults of the switch cabinet, and further solving the technical problems of fault diagnosis and lower diagnosis accuracy of the switch cabinet based on a big data technology in related technologies.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:
FIG. 1 is a flow chart of an alternative switchgear fault diagnosis method according to an embodiment of the invention;
FIG. 2 is a schematic diagram of an alternative switchgear fault diagnosis process according to an embodiment of the invention;
FIG. 3 is a schematic diagram of an alternative switchgear fault diagnosis apparatus according to an embodiment of the invention;
Fig. 4 is a block diagram of a hardware structure of an electronic device (or a mobile device) of a fault diagnosis method of a switchgear according to an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
To facilitate an understanding of the invention by those skilled in the art, some terms or nouns involved in the various embodiments of the invention are explained below:
Transformer, a deep learning model based on self-attention mechanism (self-attention), is used for processing sequence data and is excellent in the field of natural language processing.
It should be noted that, the fault diagnosis method and the device thereof of the switch cabinet in the application can be used in the technical field of power system monitoring and fault diagnosis, and also can be used in any field except the technical field of power system monitoring and fault diagnosis, and the application field of the fault diagnosis method and the device thereof of the switch cabinet in the application is not limited in the case of diagnosing the partial discharge fault of the switch cabinet based on artificial intelligence.
The following embodiments of the present invention are applicable to various fault diagnosis systems/applications/apparatuses of electric apparatuses, and particularly to fault diagnosis systems/applications/apparatuses of switch cabinets. The invention relates to a time-frequency characteristic switch cabinet fault diagnosis method based on artificial intelligence, which is used for carrying out characteristic extraction and analysis of a switch cabinet pulse data sequence through a pre-constructed characteristic processing model and a fault diagnosis model so as to realize real-time monitoring and accurate fault classification of equipment fault conditions. Through the deep learning technology, the invention can improve the efficiency and accuracy of the operation and maintenance of the power system, and has important significance for guaranteeing the stable operation of the power grid and improving the reliability of power supply.
The present invention will be described in detail with reference to the following examples.
Example 1
According to an embodiment of the present invention, there is provided an embodiment of a fault diagnosis method for a switchgear, it being noted that the steps shown in the flowcharts of the drawings may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in an order different from that herein.
Fig. 1 is a flowchart of an alternative fault diagnosis method for a switchgear according to an embodiment of the present invention, as shown in fig. 1, the method including the steps of:
Step S101, acquiring time domain data of a switch cabinet, and performing frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet;
Step S102, inputting time domain data and frequency domain data into a feature processing model, and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, wherein the feature processing model is constructed in advance;
step S103, inputting the time-frequency fusion feature vector into a fault identification model, and outputting the fault type of the switch cabinet through the fault identification model;
and step S104, generating a fault diagnosis result of the switch cabinet based on the fault type.
Through the steps, firstly, time domain data of the switch cabinet are collected, frequency domain conversion is carried out on the time domain data to obtain time domain data and frequency domain data of the switch cabinet, the time domain data and the frequency domain data are input into a feature processing model, feature extraction and feature fusion are carried out on the time domain data and the frequency domain data based on the feature processing model to obtain time-frequency fusion feature vectors of the switch cabinet, wherein the feature processing model is built in advance, then the time-frequency fusion feature vectors are input into a fault recognition model, fault types of the switch cabinet are output through the fault recognition model, and finally fault diagnosis results of the switch cabinet are generated based on the fault types.
In the embodiment, time domain data of the switch cabinet are acquired, frequency domain data are acquired through the time domain data, time-frequency characteristics are extracted based on the time domain data and the frequency domain data, the time-frequency characteristics can describe time domain and frequency domain characteristics of signals more accurately, and therefore accuracy and reliability of diagnosis of partial discharge faults are improved.
Embodiments of the present invention will be described in detail with reference to the following steps.
The embodiment of the invention provides a method for diagnosing a partial discharge fault of a switch cabinet based on artificial intelligence and deep learning technology, which can be applied to the switch cabinet or other power equipment such as a transformer, wherein the switch cabinet is an important electrical equipment in a power system and is used for distributing electric energy and controlling the power equipment, and the partial discharge fault possibly occurs along with the factors such as the lengthening of the operation time of the switch cabinet and the influence of external environment, so that the fault not only can influence the safe and stable operation of the power system, but also can cause serious consequences such as equipment damage and power failure, and therefore, the partial discharge fault diagnosis of the switch cabinet is very important for the reliability and the safety of the power system.
The embodiment of the invention provides a partial discharge diagnosis method based on a time-frequency characteristic and a deep learning model. Firstly, pulse data are acquired by utilizing three-phase coupling current of a switch cabinet, time-frequency characteristics are extracted based on the pulse data, then a deep learning model is introduced, and partial discharge faults are diagnosed and classified based on analysis of the time-frequency characteristics so as to judge the type of multi-source signal partial discharge of the switch cabinet. The method not only can fully utilize the information of the time-frequency characteristics, but also can utilize the strong analysis capability of the deep learning model, thereby realizing the accurate classification and diagnosis of the partial discharge faults of the switch cabinet.
Step S101, acquiring time domain data of the switch cabinet, and performing frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet.
It should be noted that, in the embodiment of the present invention, the switch cabinet is monitored in real time, because the local discharge is generally accompanied by the generation of high-frequency current, when the current of the switch cabinet is detected to be greater than the preset threshold, the time point is taken as the starting point to collect the time domain data of the switch cabinet, so as to obtain the time domain pulse information, the time domain pulse signal is used as the time domain data of the switch cabinet to analyze, and meanwhile, the time domain data is converted, so as to obtain the frequency domain data, and the characteristics of the signal can be reflected from two dimensions of the time domain and the frequency domain.
The method comprises the steps of receiving a pulse collection device into a switch cabinet, setting a trigger threshold for the pulse collection device, monitoring a current signal of the switch cabinet, recording a capturing time point when the amplitude of the current signal is larger than the trigger threshold, and collecting N time domain pulse signals from the switch cabinet based on a preset pulse duration by taking the capturing time point as a starting point to obtain time domain data, wherein N is a positive integer.
When the time domain signal of the switch cabinet is collected, the pulse collecting device is connected into the switch cabinet, a trigger threshold is preset, pulse data are collected in a triggering mode, namely the current signal of the switch cabinet is monitored in real time, when the amplitude of the current signal is larger than the trigger threshold, the time point is recorded as a starting point, the current signal is sampled from the time point, and therefore time domain data of the switch cabinet are obtained.
Specifically, a high-frequency pulse connecting wire is used for accessing an interface triggering type acquisition time domain pulse of an A/B/C three-phase of a live display of a switch cabinet as an original data source, the triggering type acquisition time domain pulse is used for detecting the amplitude value in a current time domain signal in real time, if the amplitude value is larger than a preset triggering threshold value, time domain pulse capturing is started, the time domain pulse capturing is used for capturing a time domain signal with the triggering time satisfying t-t 0 less than or equal to N from the triggering time t 0 as a time domain pulse signal according to the set pulse time length N, and the acquired time domain pulse signal is x= [ x 1,x2,...,xN ].
Optionally, the step of performing frequency domain conversion on the time domain data comprises the steps of processing the time domain data based on a Fourier transform algorithm to obtain initial frequency domain data, and performing high-pass filtering on the initial frequency domain data according to a preset cut-off frequency to obtain frequency domain data of the switch cabinet.
After the switch cabinet is subjected to time domain data acquisition, the time domain data is subjected to frequency domain conversion, the discrete time domain data can be processed based on discrete Fourier transform or fast Fourier transform to obtain initial frequency domain data, and then the initial frequency domain data is subjected to high-pass filtering according to preset cut-off frequency to remove noise or interference in a specific frequency range, so that the quality and accuracy of signals are improved, and the processed frequency domain data of the switch cabinet are obtained.
Specifically, for a discrete time domain signal X, the expression of its discrete Fourier transform X [ K ] is as follows:
Where X [ K ] is a frequency domain signal, X i is a time domain signal, K is a frequency index, and j is an imaginary unit.
Step S102, inputting the time domain data and the frequency domain data into a feature processing model, and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet.
It should be noted that, the invention introduces a deep learning model to perform feature processing on time domain data and frequency domain data, so as to extract features in the time domain data and the frequency domain data, where the feature processing model is pre-constructed and includes a time domain encoder for encoding the time domain data, a frequency domain encoder for encoding the frequency domain data, and a fusion encoder for fusing the time domain feature vector and the frequency domain feature vector output by the time domain encoder and the frequency domain encoder to obtain a time-frequency fusion feature vector.
Optionally, the steps of inputting the time domain data and the frequency domain data into the feature processing model, and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model comprise the steps of inputting the time domain data into a time domain coder of the feature processing model, carrying out embedding mapping on the time domain data through the time domain coder to obtain a time domain feature vector, inputting the frequency domain data into a frequency domain coder of the feature processing model, carrying out embedding mapping on the frequency domain data through the frequency domain coder to obtain a frequency domain feature vector, inputting the time domain feature vector and the frequency domain feature vector into a fusion coder of the feature processing model, and carrying out fusion on the time domain feature vector and the frequency domain feature vector through the fusion coder to obtain a time-frequency fusion feature vector.
When the feature extraction is carried out, the time domain data is input into a time domain coder of a feature processing model, the time domain data is subjected to embedding mapping based on the embedding information of the time domain sequence to obtain a time domain feature vector, meanwhile, the frequency domain data is subjected to embedding mapping through a frequency domain coder of the feature processing model to obtain a frequency domain feature vector, finally, the time domain feature vector and the frequency domain feature vector are input into a fusion coder, the fusion coder fuses the time domain feature vector and the frequency domain feature vector to obtain a fused time-frequency fusion feature vector, and the feature extraction and the vectorization of the time domain data and the frequency domain data are completed.
Specifically, the embedded information of the time domain data may include pulse data of each time point in the time domain data, position information of each time point in the time domain data, absolute time information of each time point, period information of each time point and phase information of each time point, and the embedded information of the frequency domain data may include pulse data of each point of the frequency domain data and position information of the time domain data.
For the position embedding of data, assuming that an input sequence X exists at the time t, the pulse data sampling interval is tau, the dimension of the device state vector after coding representation is set as N, and the definition of several position codes can be expressed as follows:
Where i is the actual position of each pulse data in the time domain data, PE i is the position vector of the pulse data in the time domain data, Representing the t-th element in this position vector, N is the dimension of the timing, w k＝1/100002k/N, t=0, 1, 2.
The method comprises the steps of obtaining time domain feature vectors, wherein the time domain feature vectors comprise the steps of carrying out embedding mapping on time domain data through a time domain encoder to obtain time domain embedding vectors, dividing the time domain embedding vectors to obtain attention heads corresponding to H time domain embedding vectors, wherein H is a positive integer, calculating query data, key data and value data of the attention heads through a query projection matrix, a key projection matrix and a value projection matrix for each time domain embedding vector, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes for carrying out weight distribution, calculating attention weights of the attention heads based on the query data and the key data of the attention heads for each time domain embedding vector, calculating output vectors of the attention heads based on the attention weights and the value data, and splicing the output vectors of the attention heads corresponding to the H time domain embedding vectors to obtain the time domain feature vectors.
It should be noted that, all three encoders in the feature processing model adopt a multi-head attention mechanism based on dynamic weight distribution, and the performance of the model can be optimized by dynamically adjusting the weight of each head, so that the accuracy of data processing is improved.
Specifically, when constructing a time domain feature vector, firstly, embedding and mapping time domain data through embedding information to obtain a time domain embedded vector, then, dividing the time domain embedded vector to obtain attention heads corresponding to a plurality of time domain embedded vectors, for each attention head of the time domain embedded vector, calculating query data, key data and value data of each attention head through a pre-constructed query projection matrix, a key projection matrix and a value projection matrix, calculating attention weights of the attention heads according to the query data and the key data, obtaining output vectors of the attention heads through attention weight adjustment value data, finally, splicing the output vectors of the attention heads to obtain the time domain feature vector, wherein different attention heads possibly have different contributions to a final task, calculating the attention weights according to the needs of the task, and distributing different weights for each attention head.
The method comprises the steps of obtaining frequency domain feature vectors, wherein the frequency domain feature vectors comprise the steps of carrying out embedding mapping on frequency domain data through a frequency domain encoder to obtain frequency domain embedded vectors, dividing the frequency domain embedded vectors to obtain attention heads corresponding to K frequency domain embedded vectors, K is a positive integer, calculating query data, key data and value data of the attention heads through a query projection matrix, a key projection matrix and a value projection matrix for each attention head corresponding to the frequency domain embedded vectors, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes used for carrying out weight distribution, calculating attention weights of the attention heads based on the query data and the key data of the attention heads for each attention head corresponding to the frequency domain embedded vectors, calculating output vectors of the attention heads based on the attention weights and the value data, and splicing the output vectors of the attention heads corresponding to the K frequency domain embedded vectors to obtain the frequency domain feature vectors.
Similarly, when constructing the frequency domain feature vector, firstly, carrying out embedding mapping on the frequency domain data through embedding information to obtain frequency domain embedded vectors, then dividing the frequency domain embedded vectors to obtain attention heads corresponding to a plurality of frequency domain embedded vectors, calculating query data, key data and value data of each attention head of the frequency domain embedded vectors through a pre-constructed query projection matrix, a key projection matrix and a value projection matrix, calculating attention weights of the attention heads according to the query data and the key data, obtaining output vectors of the attention heads through attention weight adjustment value data, finally, splicing the output vectors of the attention heads to obtain the frequency domain feature vector, calculating the attention weights of different attention heads according to the task requirements, and distributing different weights to the attention heads.
Specifically, for the h head of the multi-head attention, a query projection matrix (Q), a key projection matrix (K), and a value projection matrix (V) are defined as:
Wherein, The query projection matrix, the key projection matrix and the value projection matrix are respectively used for weight distribution, and the input sequence is mapped to the corresponding query space, key space and value space through the projection matrices.
For each attention head, the calculation formula of the attention weight is expressed as follows:
Where Attention h(Qh,Kh,Vh) represents the Attention weight of the Attention header, d k is the dimension of the key vector. A weight learning network is introduced to dynamically adjust the output of each header.
Assuming that W o is a learnable weight matrix of the weight learning network, the output z of the weight learning network can be expressed as:
Z＝ReLU(WO·[Attention1(Q1,K1,V1),Attention2(Q2,K2,V2),...,Attentionh(Qh,Kh,Vh)]) Final multi-head attention output Y:
Y=σ(Wy·z+b)
Where σ is the activation function, W y is another learnable weight matrix, and b is the bias term.
Step S103, inputting the time-frequency fusion feature vector into a fault identification model, and outputting the fault type of the switch cabinet through the fault identification model.
And finally, classifying the time-frequency fusion feature vectors through a fault identification model, and outputting the fault type of the switch cabinet.
The fault recognition model is obtained by pre-training, and comprises the steps of collecting historical time domain data and fault type information of a switch cabinet in a historical time period, acquiring historical frequency domain data based on the historical time domain data, carrying out feature extraction and feature fusion on the historical time domain data and the historical frequency domain data to obtain historical time-frequency feature vectors, constructing a sample set based on the historical time-frequency feature vectors and the fault type information, dividing the sample set based on a preset dividing proportion to obtain a training set and a testing set, training an initial fault recognition model through the testing set to obtain a trained fault recognition model, testing the trained fault recognition model based on the testing set to obtain a test result, and obtaining a final fault recognition model under the condition that the test result indicates that the fault recognition model passes the test.
It should be noted that the fault recognition model is pre-constructed and used for recognizing and classifying the partial discharge faults of the switch cabinet, the partial discharge faults can be divided into various types of partial discharge such as internal discharge, tip corona discharge, surface discharge, suspension potential discharge and the like, and the reasons for generating the different types of partial discharge are different, wherein the internal discharge is the internal discharge caused by the internal defect of the insulating medium due to the air gap, impurities and the like in the insulating medium introduced by the manufacturing process, the tip corona discharge is the tip corona discharge caused by burrs on the surface of the metal exposed in the air, the surface discharge is the surface discharge caused by the surface pollutants of the insulating medium, and the suspension potential discharge is the suspension potential discharge caused by the structural design defect, the structural defect in the transportation and the operation process.
When the fault recognition model is constructed, the historical time domain data and fault type information of the switch cabinet in the historical time period are used as sample data, the sample data are subjected to feature extraction and feature fusion to obtain historical time-frequency feature vectors, a training set and a testing set are constructed according to the historical time-frequency feature vectors and the fault type information, the initial depth model is subjected to iterative training through the training set, the trained model is subjected to testing through the testing set, and the final fault recognition model is obtained after the testing is passed.
The classification is realized by the position of the sample feature in the feature space, the probability distribution can be calculated by constructing a self-attention layer, and the cross entropy loss can be minimized by a post-processing network, so that the fault recognition model shows higher accuracy in classification tasks.
Wherein the self-attention layer can be realized by a self-attention mechanism in a transducer. Assuming z is the output of the self-attention layer, the cross entropy loss function is defined as:
Where y is the true tag distribution, Is the probability distribution of model predictions. The goal of the optimization layer is to continuously adjust z to minimize H.
And step S104, generating a fault diagnosis result of the switch cabinet based on the fault type.
After the type of the partial discharge fault of the switch cabinet is identified, a fault diagnosis result can be generated according to the type of the partial discharge fault and a fault resolution strategy of historical working experience, and a fault maintenance person can perform fault maintenance on the switch cabinet according to the fault diagnosis result so as to ensure safe and stable operation of a power system.
The following detailed description is directed to alternative embodiments.
Fig. 2 is a schematic diagram of a fault diagnosis flow of an optional switch cabinet according to an embodiment of the present invention, as shown in fig. 2, when fault diagnosis is performed on the switch cabinet, first, three-phase pulse currents of switch cabinet a/B/C are collected to obtain time domain data, frequency domain conversion is performed on the time domain data to obtain frequency domain data, the time domain data and the frequency domain data are input into two different channels, time domain embedding mapping is performed through time domain embedding information, coding is performed through a time domain coder, thus a time domain feature vector is extracted, meanwhile, embedding mapping is performed through frequency domain embedding information, coding is performed through a frequency domain coder, thus a frequency domain feature vector is extracted, then, a time-frequency fusion feature vector is obtained by fusing the time domain feature vector and the frequency domain feature vector through a fusion coder, finally, fault identification and classification are performed through a fault identification model, and a partial discharge fault type is output.
The embodiment of the invention can accurately extract the time-frequency characteristics from the time sequence data, remarkably improve the accuracy and efficiency of the fault diagnosis of the switch cabinet, has strong model robustness, can adapt to various power equipment and fault scenes, and simultaneously enhances the adaptability to data changes. The introduction of the multi-head attention mechanism of position coding and dynamic weight distribution improves the understanding of the model on the time sequence data structure and enhances the generalization capability and the interpretability of the model. In addition, the method for carrying out fault identification through the model supports real-time monitoring and rapid fault early warning, effectively reduces equipment shutdown time, reduces maintenance cost, prolongs equipment service life, and provides powerful technical support for intelligent and sustainable development of an electric power system.
The following describes in detail another embodiment.
Example two
The fault diagnosis device for a switch cabinet provided in this embodiment includes a plurality of implementation units, where each implementation unit corresponds to each implementation step in the first embodiment, and specific implementation and beneficial effects of each implementation unit may refer to the foregoing method embodiment and will not be described herein.
Fig. 3 is a schematic view of an alternative fault diagnosis apparatus of a switchgear according to an embodiment of the present invention, which may include, as shown in fig. 3, an acquisition unit 31, an extraction unit 32, an output unit 33, a generation unit 34, wherein,
The acquisition unit 31 is configured to acquire time domain data of the switch cabinet, and perform frequency domain conversion on the time domain data to obtain time domain data and frequency domain data of the switch cabinet;
The extracting unit 32 is configured to input time domain data and frequency domain data into a feature processing model, and perform feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, where the feature processing model is pre-constructed;
An output unit 33, configured to input the time-frequency fusion feature vector to a fault recognition model, and output a fault type of the switch cabinet through the fault recognition model;
and a generating unit 34 for generating a fault diagnosis result of the switch cabinet based on the fault type.
The fault diagnosis device of the switch cabinet acquires time domain data of the switch cabinet through the acquisition unit 31 and performs frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet, inputs the time domain data and the frequency domain data into the feature processing model through the extraction unit 32, performs feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, wherein the feature processing model is constructed in advance, inputs the time-frequency fusion feature vector into the fault recognition model through the output unit 33, outputs the fault type of the switch cabinet through the fault recognition model, and generates a fault diagnosis result of the switch cabinet based on the fault type through the generation unit 34.
In the embodiment, time domain data of the switch cabinet are acquired, frequency domain data are acquired through the time domain data, time-frequency characteristics are extracted based on the time domain data and the frequency domain data, the time-frequency characteristics can describe time domain and frequency domain characteristics of signals more accurately, and therefore accuracy and reliability of diagnosis of partial discharge faults are improved.
Optionally, the extracting unit 32 includes a first mapping module, configured to input time domain data to a time domain encoder of the feature processing model and perform embedding mapping on the time domain data by the time domain encoder to obtain a time domain feature vector, a second mapping module, configured to input frequency domain data to a frequency domain encoder of the feature processing model and perform embedding mapping on the frequency domain data by the frequency domain encoder to obtain a frequency domain feature vector, and a first fusion module, configured to input the time domain feature vector and the frequency domain feature vector to a fusion encoder of the feature processing model, and fuse the time domain feature vector and the frequency domain feature vector by the fusion encoder to obtain a time-frequency fusion feature vector.
The first mapping module comprises a first mapping sub-module, a first calculating sub-module and a first splicing sub-module, wherein the first mapping sub-module is used for carrying out embedding mapping on time domain data through a time domain coder to obtain time domain embedded vectors, the first dividing sub-module is used for dividing the time domain embedded vectors to obtain attention heads corresponding to H time domain embedded vectors, H is a positive integer, the first calculating sub-module is used for calculating query data, key data and value data of the attention heads corresponding to each time domain embedded vector through a query projection matrix, a key projection matrix and a value projection matrix, the query projection matrix, the key projection matrix and the value projection matrix are matrices which are constructed in advance and used for carrying out weight distribution, the second calculating sub-module is used for calculating the attention weight of the attention heads corresponding to each time domain embedded vector based on the query data and the key data of the attention heads and calculating the output vector of the attention heads based on the attention heads and the value data, and the first splicing sub-module is used for splicing the output vectors of the attention heads corresponding to the H time domain embedded vectors to obtain time domain feature vectors.
The second mapping module comprises a second mapping sub-module, a second dividing sub-module and a third splicing sub-module, wherein the second mapping sub-module is used for carrying out embedding mapping on frequency domain data through a frequency domain encoder to obtain frequency domain embedded vectors, the second dividing sub-module is used for dividing the frequency domain embedded vectors to obtain attention heads corresponding to K frequency domain embedded vectors, K is a positive integer, the third calculating sub-module is used for calculating query data, key data and value data of the attention heads corresponding to each frequency domain embedded vector through a query projection matrix, a key projection matrix and a value projection matrix, the query projection matrix, the key projection matrix and the value projection matrix are matrices which are constructed in advance and used for carrying out weight distribution, the fourth calculating sub-module is used for calculating attention weights of the attention heads corresponding to each frequency domain embedded vector based on the query data and the key data of the attention heads and calculating output vectors of the attention heads based on the attention weights and the value data, and the second splicing sub-module is used for splicing the output vectors of the attention heads corresponding to the K frequency domain embedded vectors to obtain frequency domain feature vectors.
Optionally, the acquisition unit 31 includes a first setting module, configured to access the pulse acquisition device to the switch cabinet and set a trigger threshold for the pulse acquisition device, a first monitoring module, configured to monitor a current signal of the switch cabinet, record a capturing time point when the amplitude of the current signal is greater than the trigger threshold, and acquire N time domain pulse signals from the switch cabinet based on a preset pulse duration with the capturing time point as a starting point, so as to obtain time domain data, where N is a positive integer.
Optionally, the acquisition unit 31 further includes a first processing module, configured to process the time domain data based on a fourier transform algorithm to obtain initial frequency domain data, and a first filtering module, configured to perform high-pass filtering on the initial frequency domain data according to a preset cut-off frequency to obtain frequency domain data of the switch cabinet.
The fault diagnosis device of the switch cabinet further comprises a second acquisition module, a first extraction module, a first construction module and a first training module, wherein the second acquisition module is used for acquiring historical time domain data and fault type information of the switch cabinet in a historical time period and acquiring historical frequency domain data based on the historical time domain data, the first extraction module is used for carrying out feature extraction and feature fusion on the historical time-frequency feature vector and the historical frequency domain data to obtain a historical time-frequency feature vector, the first construction module is used for constructing a sample set based on the historical time-frequency feature vector and the fault type information and dividing the sample set based on a preset dividing proportion to obtain a training set and a testing set, the first training module is used for training an initial fault recognition model through the testing set to obtain a trained fault recognition model, the first testing module is used for testing the trained fault recognition model based on the testing set to obtain a testing result, and the final fault recognition model is obtained under the condition that the testing result indicates that the fault recognition model passes the testing.
The fault diagnosis apparatus of the switch cabinet may further include a processor and a memory, wherein the acquisition unit 31, the extraction unit 32, the output unit 33, the generation unit 34, and the like are stored as program units in the memory, and the processor executes the program units stored in the memory to realize corresponding functions.
The processor includes a kernel, and the kernel fetches a corresponding program unit from the memory. The kernel can be provided with one or more than one, and the partial discharge faults of the switch cabinet can be diagnosed by adjusting the parameters of the kernel.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), which includes at least one memory chip.
According to another aspect of the embodiment of the present invention, there is also provided a computer readable storage medium, where the computer readable storage medium includes a stored computer program, and when the computer program runs, the device on which the computer readable storage medium is located is controlled to execute the fault diagnosis method of any one of the switch cabinets.
According to another aspect of the embodiment of the present invention, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, where the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the fault diagnosis method of any one of the switch cabinets described above.
According to another aspect of the embodiments of the present invention, there is also provided a computer program product, including a computer program, wherein the computer program when executed by a processor implements the fault diagnosis method of any one of the switch cabinets described above.
The application further provides a computer program product which is suitable for executing a program initialized with the following method steps when the computer program is executed on data processing equipment, wherein the computer program is used for collecting time domain data of a switch cabinet, carrying out frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet, inputting the time domain data and the frequency domain data into a feature processing model, carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain time-frequency fusion feature vectors of the switch cabinet, wherein the feature processing model is constructed in advance, inputting the time-frequency fusion feature vectors into a fault identification model, outputting the fault type of the switch cabinet through the fault identification model, and generating a fault diagnosis result of the switch cabinet based on the fault type.
The application also provides a computer program product which is suitable for executing a program initialized with the following steps of the method when being executed on data processing equipment, wherein the steps of inputting time domain data and frequency domain data into a feature processing model and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model comprise the steps of inputting the time domain data into a time domain coder of the feature processing model and carrying out embedded mapping on the time domain data through the time domain coder to obtain time domain feature vectors, inputting the frequency domain data into a frequency domain coder of the feature processing model and carrying out embedded mapping on the frequency domain data through the frequency domain coder to obtain frequency domain feature vectors, inputting the time domain feature vectors and the frequency domain feature vectors into a fusion coder of the feature processing model, and carrying out fusion on the time domain feature vectors and the frequency domain feature vectors through the fusion coder to obtain time-frequency fusion feature vectors.
The application further provides a computer program product which is suitable for executing a program initialized with the following method steps when the computer program is executed on data processing equipment, the step of obtaining time domain feature vectors comprises the steps of carrying out embedding mapping on time domain data through a time domain coder to obtain time domain embedded vectors, dividing the time domain embedded vectors to obtain attention heads corresponding to H time domain embedded vectors, wherein H is a positive integer, calculating query data, key data and value data of the attention heads through a query projection matrix, a key projection matrix and a value projection matrix for each attention head corresponding to the time domain embedded vectors, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrices used for carrying out weight distribution, calculating attention weights of the attention heads based on the query data and the key data of each attention head for each attention head corresponding to the time domain embedded vectors, calculating output vectors of the attention heads based on the attention weights and the value data, and splicing the output vectors of the attention heads corresponding to the H time domain embedded vectors to obtain the time domain feature vectors.
The application also provides a computer program product which is suitable for executing a program initialized with the following method steps when being executed on data processing equipment, the step of carrying out embedding mapping on frequency domain data through a frequency domain encoder to obtain frequency domain feature vectors comprises the steps of carrying out embedding mapping on the frequency domain data through the frequency domain encoder to obtain frequency domain embedded vectors, dividing the frequency domain embedded vectors to obtain attention heads corresponding to K frequency domain embedded vectors, wherein K is a positive integer, calculating query data, key data and value data of the attention heads through query projection matrix, key projection matrix and value projection matrix for each attention head corresponding to the frequency domain embedded vectors, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrices used for carrying out weight distribution, calculating attention weights of the attention heads based on the query data and the key data of the attention heads for each attention head corresponding to each frequency domain embedded vector, calculating output vectors of the attention heads based on the attention weights and the value data, and splicing the output vectors of the attention heads corresponding to the K frequency domain embedded vectors to obtain the frequency domain feature vectors.
The application also provides a computer program product which is suitable for executing a program initialized with the following method steps when the computer program product is executed on data processing equipment, wherein the step of collecting time domain signals of a switch cabinet comprises the steps of connecting a pulse collecting device into the switch cabinet and setting a trigger threshold for the pulse collecting device, monitoring current signals of the switch cabinet, recording a capturing time point under the condition that the amplitude of the current signals is larger than the trigger threshold, and collecting N time domain pulse signals from the switch cabinet based on the preset pulse duration by taking the capturing time point as a starting point to obtain time domain data, wherein N is a positive integer.
The application also provides a computer program product adapted to perform, when executed on a data processing apparatus, an initialization procedure having the method steps of processing time domain data based on a fourier transform algorithm to obtain initial frequency domain data, and high pass filtering the initial frequency domain data according to a preset cut-off frequency to obtain frequency domain data of a switch cabinet.
The application further provides a computer program product which is suitable for executing a program initialized with the following method steps when the computer program is executed on data processing equipment, wherein the fault recognition model is obtained through pre-training, the step of training the fault recognition model comprises the steps of collecting historical time domain data and fault type information of a switch cabinet in a historical time period and obtaining historical frequency domain data based on the historical time domain data, carrying out feature extraction and feature fusion on the historical time domain data and the historical frequency domain data to obtain historical time-frequency feature vectors, constructing a sample set based on the historical time-frequency feature vectors and the fault type information, dividing the sample set based on a preset dividing proportion to obtain a training set and a test set, training an initial fault recognition model through the test set to obtain a trained fault recognition model, testing the trained fault recognition model based on the test set to obtain a test result, and obtaining a final fault recognition model under the condition that the test result indicates that the fault recognition model passes the test.
Fig. 4 is a block diagram of a hardware structure of an electronic device (or a mobile device) of a fault diagnosis method of a switchgear according to an embodiment of the present invention. As shown in fig. 4, the electronic device may include one or more processors 402 (shown in fig. 4 as 402a, 402b,) 402n (the processor 402 may include, but is not limited to, a microprocessor MCU or a processing device such as a programmable logic device FPGA), a memory 404 for storing data. Among other things, a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a keyboard, a power supply, and/or a camera may be included. It will be appreciated by those of ordinary skill in the art that the configuration shown in fig. 4 is merely illustrative and is not intended to limit the configuration of the electronic device described above. For example, the electronic device may also include more or fewer components than shown in FIG. 4, or have a different configuration than shown in FIG. 4.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed technology may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, for example, may be a logic function division, and may be implemented in another manner, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. The storage medium includes a U disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, etc. which can store the program code.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (10)
1. A fault diagnosis method for a switchgear, comprising:
acquiring time domain data of a switch cabinet, and performing frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet;
Inputting the time domain data and the frequency domain data into a feature processing model, and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, wherein the feature processing model is constructed in advance;
inputting the time-frequency fusion feature vector to a fault identification model, and outputting the fault type of the switch cabinet through the fault identification model;
and generating a fault diagnosis result of the switch cabinet based on the fault type.
2. The fault diagnosis method according to claim 1, wherein the step of inputting the time domain data and the frequency domain data to a feature processing model and performing feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model comprises:
inputting the time domain data to a time domain coder of the feature processing model, and performing embedding mapping on the time domain data through the time domain coder to obtain a time domain feature vector;
Inputting the frequency domain data to a frequency domain coder of the feature processing model, and performing embedding mapping on the frequency domain data through the frequency domain coder to obtain a frequency domain feature vector;
And inputting the time domain feature vector and the frequency domain feature vector to a fusion encoder of the feature processing model, and fusing the time domain feature vector and the frequency domain feature vector through the fusion encoder to obtain the time-frequency fusion feature vector.
3. The fault diagnosis method according to claim 2, wherein the step of embedding and mapping the time domain data by the time domain encoder to obtain a time domain feature vector comprises:
performing embedding mapping on the time domain data through the time domain coder to obtain a time domain embedded vector;
Dividing the time domain embedded vectors to obtain H attention heads corresponding to the time domain embedded vectors, wherein H is a positive integer;
For the attention head corresponding to each time domain embedded vector, calculating query data, key data and value data of the attention head through a query projection matrix, a key projection matrix and a value projection matrix, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes for weight distribution;
for each attention head corresponding to the time-domain embedded vector, calculating an attention weight of the attention head based on query data and key data of the attention head, and calculating an output vector of the attention head based on the attention weight and the value data;
and splicing the output vectors of the attention heads corresponding to the H time domain embedded vectors to obtain the time domain feature vector.
4. The fault diagnosis method according to claim 2, wherein the step of embedding and mapping the frequency domain data by the frequency domain encoder to obtain a frequency domain feature vector comprises:
Performing embedding mapping on the frequency domain data through the frequency domain coder to obtain a frequency domain embedded vector;
Dividing the frequency domain embedded vectors to obtain attention heads corresponding to K frequency domain embedded vectors, wherein K is a positive integer;
for the attention head corresponding to each frequency domain embedded vector, calculating query data, key data and value data of the attention head through a query projection matrix, a key projection matrix and a value projection matrix, wherein the query projection matrix, the key projection matrix and the value projection matrix are pre-constructed matrixes for weight distribution;
For each attention head corresponding to the frequency domain embedded vector, calculating an attention weight of the attention head based on query data and key data of the attention head, and calculating an output vector of the attention head based on the attention weight and the value data;
And splicing the output vectors of the attention heads corresponding to the K frequency domain embedded vectors to obtain the frequency domain feature vector.
5. The fault diagnosis method according to claim 1, wherein the step of collecting the time domain signal of the switch cabinet comprises:
The pulse acquisition device is connected into the switch cabinet, and a trigger threshold value is set for the pulse acquisition device;
monitoring a current signal of the switch cabinet, and recording a capturing time point under the condition that the amplitude of the current signal is larger than a trigger threshold;
And collecting N time domain pulse signals from the switch cabinet based on preset pulse duration by taking the capturing time point as a starting point to obtain the time domain data, wherein N is a positive integer.
6. The fault diagnosis method according to claim 1, wherein the step of frequency-domain converting the time-domain data includes:
Processing the time domain data based on a Fourier transform algorithm to obtain initial frequency domain data;
and carrying out high-pass filtering on the initial frequency domain data according to a preset cut-off frequency to obtain the frequency domain data of the switch cabinet.
7. The method of claim 1, wherein the fault identification model is pre-trained, and wherein the step of training the fault identification model comprises:
acquiring historical time domain data and fault type information of the switch cabinet in a historical time period, and acquiring historical frequency domain data based on the historical time domain data;
performing feature extraction and feature fusion on the historical time domain data and the historical frequency domain data to obtain a historical time-frequency feature vector;
constructing a sample set based on the historical time-frequency characteristic vector and the fault type information, and dividing the sample set based on a preset dividing proportion to obtain a training set and a testing set;
training an initial fault recognition model through a test set to obtain a trained fault recognition model;
And testing the trained fault recognition model based on the test set to obtain a test result, and obtaining the final fault recognition model under the condition that the test result indicates that the fault recognition model passes the test.
8. A fault diagnosis apparatus of a switchgear, comprising:
the acquisition unit is used for acquiring time domain data of the switch cabinet and performing frequency domain conversion on the time domain data to obtain the time domain data and the frequency domain data of the switch cabinet;
The extraction unit is used for inputting the time domain data and the frequency domain data into a feature processing model, and carrying out feature extraction and feature fusion on the time domain data and the frequency domain data based on the feature processing model to obtain a time-frequency fusion feature vector of the switch cabinet, wherein the feature processing model is constructed in advance;
The output unit is used for inputting the time-frequency fusion feature vector to a fault identification model and outputting the fault type of the switch cabinet through the fault identification model;
and the generating unit is used for generating a fault diagnosis result of the switch cabinet based on the fault type.
9. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium is located to perform the fault diagnosis method of the switchgear of any one of claims 1-7.
10. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the fault diagnosis method of the switchgear of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411186140.2A CN119066521A (en) | 2024-08-27 | 2024-08-27 | Fault diagnosis method and device for switch cabinet, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411186140.2A CN119066521A (en) | 2024-08-27 | 2024-08-27 | Fault diagnosis method and device for switch cabinet, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN119066521A true CN119066521A (en) | 2024-12-03 |
Family
ID=93642363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411186140.2A Pending CN119066521A (en) | 2024-08-27 | 2024-08-27 | Fault diagnosis method and device for switch cabinet, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119066521A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119382347A (en) * | 2024-12-25 | 2025-01-28 | 山东寿光巨能电气有限公司 | A remote monitoring system for fault information of low voltage distribution cabinet |
-
2024
- 2024-08-27 CN CN202411186140.2A patent/CN119066521A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119382347A (en) * | 2024-12-25 | 2025-01-28 | 山东寿光巨能电气有限公司 | A remote monitoring system for fault information of low voltage distribution cabinet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3699614B1 (en) | Method and system of partial discharge recognition for diagnosing electrical networks | |
| Ma et al. | Smart transformer for smart grid—intelligent framework and techniques for power transformer asset management | |
| CN104655423B (en) | A kind of Fault Diagnosis of Roller Bearings based on time-frequency domain multi-dimensional vibration Fusion Features | |
| Liu et al. | Dual-channel convolutional network-based fault cause identification for active distribution system using realistic waveform measurements | |
| CN113884844B (en) | Transformer partial discharge type identification method and system | |
| CN117668751A (en) | High-low voltage power system fault diagnosis method and device | |
| Huang et al. | BIT-based intermittent fault diagnosis of analog circuits by improved deep forest classifier | |
| CN119066521A (en) | Fault diagnosis method and device for switch cabinet, electronic equipment and storage medium | |
| CN104713714B (en) | A kind of primary cut-out action characteristic analysis method clustered based on grid multi-density | |
| CN115015683B (en) | Cable production performance test method, device, equipment and storage medium | |
| CN118362831B (en) | Cable fault intelligent positioning system and method based on big data | |
| CN119691642A (en) | Fault detection method, device, equipment and storage medium for electrical equipment | |
| Shao et al. | Hybrid multi‐scale residual network for high‐voltage circuit breakers fault diagnosis | |
| Fahim et al. | An agreement based dynamic routing method for fault diagnosis in power network with enhanced noise immunity | |
| CN115130516A (en) | A GIS equipment state identification method and device based on spectral power spectrum entropy | |
| Abdul-hamza et al. | Electricity Theft Detection System Using Cloud Computing and Deep Learning Techniques. | |
| Vieira et al. | Embedding a neural classifier to detect faults in a three-phase induction motor | |
| CN119644996B (en) | Solid-state power controller fault diagnosis system and method based on data analysis | |
| Wang et al. | ECGAN: An Efficient Diagnostic Strategy for Hidden Deterioration in DC-DC Converters | |
| CN119805310B (en) | GIS bus electrical contact defect diagnosis method and related device integrating magnetic field-current characteristics | |
| Haroun et al. | Self-Organizing Map and feature selection for of IM broken rotor bars faults detection and diagnosis | |
| CN116643170B (en) | Motor shafting vibration testing method and device and computer equipment | |
| Qaisar et al. | Efficient Detection of Appliance Consumption Pattern by Using Level-CrossingSampling | |
| CN119884847B (en) | Secondary power failure detection and diagnosis method and device for high-temperature aging test equipment | |
| Luo et al. | A novel generative adversarial network based early fault diagnosis method for permanent magnet synchronous motor bearings |
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 |