CN109100627A - A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode - Google Patents
A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode Download PDFInfo
- Publication number
- CN109100627A CN109100627A CN201811288775.8A CN201811288775A CN109100627A CN 109100627 A CN109100627 A CN 109100627A CN 201811288775 A CN201811288775 A CN 201811288775A CN 109100627 A CN109100627 A CN 109100627A
- Authority
- CN
- China
- Prior art keywords
- discharge pulse
- shelf depreciation
- acquisition device
- feature
- high speed
- 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
Classifications
-
- 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
- G01R31/1227—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 of components, parts or materials
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode, it is related to measuring electric variable field, the following steps are included: passing through high speed acquisition device, partial discharge pulse's signal that various typical fault defect models generate is acquired and is stored, PRPD spectrum data library is established.Shelf depreciation grayscale image is converted by PRPD spectrum data, building depth residual error network carries out self-adaptive feature extraction to shelf depreciation grayscale image, while depth excavates the feature mode of different shelf depreciation types in electric discharge grayscale image, realizes the identification of shelf depreciation mode；Beneficial effects of the present invention: identification while with higher accuracy rate and robustness of the present invention to local electric discharge type.Entire identification process does not need artificial constructed feature, belongs to the end-to-end mode identification method of data driven type.
Description
Technical field
The present invention relates to technology of sharing fields, more specifically refer to a kind of power equipment part based on end-to-end mode
Discharge fault diagnostic method.
Background technique
Power equipment in modern power systems is producing, and manufactures, installation, there are various insulation defects in operation, causes
The phenomenon that different degrees of internal field's distortion occurs, then causes apparatus local discharge (partial discharge, PD),
Therefore local discharge signal is detected in time, grasps the insulation status of power equipment in real time for maintenance electric system health
Stable operation is particularly important.
Mainly there is time-based analytical model (time resolved partial for local discharge signal at present
Discharge, TRPD) and analytical model (phase resolved partial discharge, PRPD) two based on phase
Kind.Wherein PRPD analysis method mainly pass through excavate partial discharge pulse operating frequency phase, discharge capacity and discharge time three it
Between internal relation and construct corresponding discharge characteristic " fingerprint ", carry out PD Pattern Recognition.Mould is analyzed compared to TRPD
Formula, PRPD analytical model method have stability preferable, and the lesser advantage of data volume obtains broader applications in practice.
The key for influencing PD Pattern Recognition performance is effective extraction of discharge characteristic " fingerprint ", and current method is main
Feature is recognized using domain knowledge is corresponding with mathematical tool engineer by expert, is such as constructed by PRPD spectrum data all
Such as degree of skewness, steepness, the manual features such as discharge capacity factor, which are re-fed into classifier, classifies, but these manual features exist
Subjectivity is strong, uncertain big, relies on expertise, the bad disadvantage of recognition effect.In recent years, deep learning (Deep
Learning) one research boom has been started in area of pattern recognition.Deep learning model mainly passes through simple non-linear
Layer heap is folded successively to be completed to be abstracted initial data, and study obtains the more abstract expression of initial data, feature learning process
Excessive artificial participation is not needed, there is powerful character representation learning ability, is badly in need of deep learning field is widely applied
Convolutional neural networks (Convolutional Neural Network, CNN) and its expansion depth residual error network (Residual
Network, ResNet) it is applied to GIS partial discharge area of pattern recognition, improve the performance of GIS partial discharge pattern-recognition.
In addition, the capturing technology of existing local discharge signal is: complete signal being acquired and stored, then passes through software again
Processing identifies from complete signal and extracts partial discharge pulse therein, then judges failure by analyzing partial discharge pulse
Type.This technology has the disadvantage in that (1) data volume is big, causes equipment to need to store and transmit a large amount of data, influences to set
Standby working efficiency；(2) frequency in shelf depreciation can not be acquired in the pulse of GHz.Therefore need to design a kind of shelf depreciation
The feature of signal triggers capturing technology.
Summary of the invention
A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode provided by the invention, purpose exist
In solution the above-mentioned problems in the prior art.
The technical solution adopted by the invention is as follows:
A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode, comprising the following steps:
(1), by high speed acquisition device, the feature triggering capturing technology based on local discharge signal is to various typical fault defects
Partial discharge pulse's signal that model generates is acquired and stores, and establishes PRPD spectrum data library；
(2), by grayscale image constructing module, the grayscale image of shelf depreciation type is constructed using PRPD spectrum data；
(3), it by feature extraction and categorization module, is extracted using the shelf depreciation type identification network based on depth residual error network
The identification feature of grayscale image, and Classification and Identification is carried out to identification feature；
(4), using grayscale image constructing module and trained feature extraction and categorization module as the classifier of detecting instrument, to existing
The PRPD spectrum data of field shelf depreciation carries out feature extraction and Analysis on Fault Diagnosis.
Further, the particular content of the step (2) is:
PRRD spectrum data is inputted to grayscale image constructing module, using partial discharge pulse signal amplitude q as the longitudinal axis, is divided into M
Section, operating frequency phase φ are horizontal axis, are divided into N number of section, q- φ plane is divided into M × N number of section；Count each section
Partial discharge pulse's number, construct q- φ-n map；To partial discharge pulse's numberInto
Row normalizes i.e.:；Wherein,For normalization after pulse number,For actual pulse number,For the maximum impulse number of the q- φ-n map；Set the gray value of each point pixel:,
MakeMaxima and minima be corresponding in turn to the minimal gray grade and maximum gray scale of gray level image, to construct
The visualization grayscale image of shelf depreciation type.
Further, the shelf depreciation type identification network is by input layer, several convolutional layers, maximum value pond layer, the overall situation
Average pond layer, full articulamentum and output layer are constituted；Wherein, input data is shelf depreciation grayscale image；Convolutional layer, maximum value pond
The combination for changing layer realizes network characterization extracted in self-adaptive function, and the average pond layer of the overall situation realizes Feature Compression, full articulamentum
Then undertake classifying to input feature vector for task.
Further, the shelf depreciation type identification network by input layer, the first convolutional layer, the first maximum value pond layer,
First residual error structure, the second residual error structure, the second maximum value pond layer, third residual error structure, the 4th residual error structure, the overall situation are average
Pond layer, full articulamentum, output layer are constituted；The input data of the input layer is shelf depreciation grayscale image；The average pond of the overall situation
Layer will the corresponding feature point group of upper one layer all characteristic patterns at feature vector is exported, which is that the network is played a game
The feature of portion's electric discharge grayscale image extracted in self-adaptive；Full articulamentum then undertakes classifying to input feature vector for task.
Further, further include step (5): the PRPD spectrum data by continuing to build up various typical fault defect models,
The PRPD spectrum data of collected shelf depreciation when on-site test faulty equipment is constantly collected simultaneously, these data are constantly tired
Product is into PDPR spectrum data library and is trained, optimizes to the classifier.
Further, in the step (1), the method that high speed acquisition device acquires partial discharge pulse's signal is specifically included
Following steps:
(S1), the sample rate and local discharge sensor type matching of high speed acquisition device are set, and adjust high speed acquisition device
For low activation threshold value, partial discharge pulse's signal of continuous sampling different types of faults defect, while guaranteeing that sampling duration can cover
Cover the discharge pulse signal of multiple frequency cycles；
(S2), collected partial discharge pulse's waveform is unfolded, the list of careful analysis different types of faults defect shelf depreciation
The Time-distribution of multiple pulses on the temporal signatures and time series of a pulse, and quantify these two types of characteristic parameters；
(S3), the pulse temporal feature according to obtained in step (S2) and burst length distribution characteristics design high speed acquisition device
To the triggering collection parameter of local discharge pulse signal；
(S4), according to the triggering collection parameter of partial discharge pulse's signal in step (S3), the triggering of configuration high-speed acquisition device
Acquisition parameter is implemented to acquire to local discharge pulse information
(S5), according to the acquisition in step (S4) to local discharge pulse signal, partial discharge pulse's signal of acquisition is stored.
Further, the step (S1) comprising the following specific steps
(S1.1), the sample rate and local discharge sensor type matching of acquisition device are set, and adjustment high speed acquisition device is low
Threshold values trigger condition；
(S1.2), it using partial discharge pulse's signal of high speed acquisition device continuous sampling different types of faults defect, protects simultaneously
Card sampling duration can cover the discharge pulse signal of multiple frequency cycles, the return step (S1.1) if acquiring failure, if acquisition
Successfully then follow the steps (S2)；
Further, the step (S3) comprising the following specific steps
(S3.1), the single pulse waveforms of partial discharge pulse's signal of analytical procedure (S2) different types of faults defect collected
Feature and multiple-pulse Time-distribution, and quantization characteristic parameter value；
(S3.2), the touching according to the design high speed acquisition device of characteristic ginseng value obtained by step (S3.1) to local discharge pulse signal
Send out acquisition parameter；
Further, the step (S4) comprising the following specific steps
(S4.1), the triggering collection parameter of local discharge pulse signal is matched according to the high speed acquisition device that step (S3.2) is designed
Set the triggering collection parameter of high speed acquisition device；
(S4.2), using high speed acquisition device, partial discharge pulse's signal is acquired, the return step (S4.1) if acquiring failure,
(S5) is successfully thened follow the steps if capturing.
Further, further comprising the steps of:
(S6), corresponding upper computer software is developed, is realized to the control of high speed acquisition device and data flow interaction；
(S7), partial discharge pulse's sequence is showed and is reconstructed by upper computer software, further analyze the potential valence of data
Value.
By the above-mentioned description of this invention it is found that being compared with existing technology, the present invention has the advantages that
PRPD spectrum data is converted shelf depreciation grayscale image by the present invention, constructs depth residual error network to shelf depreciation grayscale image
Self-adaptive feature extraction is carried out, while depth excavates the feature mode of different shelf depreciation types in electric discharge grayscale image, realization office
The identification of portion's discharge mode.Identification while with higher accuracy rate and robustness of the present invention to local electric discharge type.Entirely
Identification process does not need artificial constructed feature, belongs to the end-to-end mode identification method of data driven type.
In addition, the present invention in acquisition pulse signal, analyzes partial discharge pulse's letter of different typical fault defects first
Number, analyze the various features parameter value of pulse；Recycle the touching of these characteristic ginseng values design high speed acquisition device signal acquisition
Parameter is sent out, to capture the partial discharge pulse of useful (i.e. satisfaction joint trigger policy), only to reduce the data processing of equipment
Amount improves working efficiency.Meanwhile the trigger parameter established by various features combining parameter values can guarantee high speed acquisition device
Frequency is captured in the pulse of GHz, detection omission or distorted signals is avoided, improves the accuracy judged fault type.
Detailed description of the invention
Fig. 1 is the schematic diagram of the shelf depreciation type identification network based on depth residual error network.
Fig. 2 is discharge time of four kinds of electric discharge types under different cycles.
Fig. 3 is type identification accuracy rate of four kinds of electric discharge types under different cycles.
Fig. 4 is the feature visualization figure that the present invention extracts.Wherein 1,2,3,4 respectively corresponds and indicate that point discharge, particle are put
Electricity, bubble-discharge, suspended discharge.
Fig. 5 is the feature visualization figure that PRPD spectrum data manually extracts.Wherein 1,2,3,4 respectively correspond indicate tip put
Electricity, particle electric discharge, bubble-discharge, suspended discharge.
Specific embodiment
Illustrate a specific embodiment of the invention with reference to the accompanying drawings.In order to fully understand the present invention, it is described below and is permitted
More details, but to those skilled in the art, the present invention can also be realized without these details.
A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode, comprising the following steps:
(1), by high speed acquisition device, the feature triggering capturing technology typical events various to laboratory based on local discharge signal
Partial discharge pulse's signal that barrier defect model generates is acquired and stores, and establishes PRPD spectrum data library.
Specifically, high speed acquisition device includes sequentially connected local discharge sensor, signal conditioning module, high speed acquisition
Device and the computer for being equipped with control software.Wherein, high speed acquisition device is high speed acquisition oscillograph.By typical fault Defect Modes
Type is placed in the designated position of shelf depreciation simulator, and local discharge sensor is placed in shelf depreciation simulator and is detected
Position, and guarantee the shell reliable ground of shelf depreciation simulator and each detection device；Using calibration square wave to shelf depreciation
Simulator is calibrated；By the sampling parameter of the control software set high speed acquisition device of computer, channel parameters and signal
Improve parameter；Shelf depreciation simulator is boosted to typical fault defect to specified gear by the gain of adjustment signal conditioning module
The firing potential of model, the acquisition of high speed acquisition device commencing signal and preservation.
(2), by grayscale image constructing module, the grayscale image of shelf depreciation type is constructed using PRPD spectrum data.
Specifically, the specific method of grayscale image constructing module construction grayscale image is:
(2.1) PRRD spectrum data is inputted to grayscale image constructing module, using partial discharge pulse signal amplitude q as the longitudinal axis, divided
For M section, operating frequency phase φ is horizontal axis, is divided into N number of section, q- φ plane is divided into M × N number of section；
(2.2) partial discharge pulse's number in each section is counted, construct q- φ-n map；
(2.3) to partial discharge pulse's numberIt is normalized i.e.:；Wherein,To return
Pulse number after one change,For actual pulse number,For the maximum impulse number of the q- φ-n map；
(2.4) gray value of each point pixel is set:, makeMaxima and minima
It is corresponding in turn to the minimal gray grade and maximum gray scale of gray level image, to construct the visualization gray scale of shelf depreciation type
Figure.
(3), by feature extraction and categorization module, the shelf depreciation type identification network based on depth residual error network is utilized
The identification feature of grayscale image is extracted, and Classification and Identification is carried out to identification feature.
Such as Fig. 1, specifically, entire shelf depreciation type identification network is by input layer, several convolutional layers, maximum value pond
Layer, global average pond layer, full articulamentum and output layer are constituted.Input data is shelf depreciation grayscale image, wherein convolutional layer, pond
The combination for changing layer realizes network characterization extracted in self-adaptive function, and the average pond layer of the overall situation realizes Feature Compression, full articulamentum
Classifying to input feature vector for task is then undertaken, entire identification process does not need artificial constructed feature, belongs to data driven type
End-to-end mode identification method.It is situated between below to the specific structure and parameter setting of the shelf depreciation type identification network
It continues.
Shelf depreciation type identification network is by input layer, the first convolutional layer, the first maximum value pond layer, the first residual error knot
Structure, the second residual error structure, the second maximum value pond layer, third residual error structure, the 4th residual error structure, global average pond layer, entirely
Articulamentum, output layer are constituted.
1. input layer: input data is the shelf depreciation grayscale image of size M × N.
2. residual error structure: every two convolutional layer constitutes a residual error structure, the first, second residual error structure by identical mapping
Convolutional layer output characteristic pattern quantity be 16, convolutional layer output characteristic pattern quantity is 64 in third, the 4th residual error structure.
3. maximum value pond layer: present networks use two maximum value pond layers, be located at first layer convolutional layer later with
And between second, third residual error structure, it is intended to being operated by pondization by the size reduction of characteristic pattern is original 1/4, so that
Characteristic information is more concentrated.
4. the average pond layer of the overall situation: in order to reduce parameter amount, the lift scheme speed of service, present networks use global average pond
Change layer to compress network output characteristic pattern, the average pond layer of the overall situation schemes average pond by the way that input feature vector figure is carried out whole
Change, obtain a characteristic point, finally will the corresponding feature point group of upper one layer all characteristic patterns at exporting feature vector.The output is special
Levying vector is feature of the network to shelf depreciation grayscale image extracted in self-adaptive.
5. activation primitive: output layer uses softmax function as activation primitive, and Relu function is used after remaining convolutional layer
As activation primitive.The unilateral of Relu function inhibits operation so that neuron has sparse activity, Ke Yiyou in neural network
Effect promotes network training speed and ability in feature extraction.
(4), using grayscale image constructing module and trained feature extraction and categorization module as the classifier of detecting instrument,
Feature extraction and Analysis on Fault Diagnosis are carried out to the PRPD spectrum data of live shelf depreciation.
(5), by continuing to build up the PRPD spectrum data of various typical fault defect models, while scene inspection is constantly collected
The PRPD spectrum data of collected shelf depreciation when surveying faulty equipment constantly accumulates these data into PDPR database and right
Above-mentioned classifier is trained, optimizes.
Above-mentioned, in step (1), the method that high speed acquisition device acquires partial discharge pulse's signal is based on shelf depreciation
The feature of signal triggers capturing technology, specifically includes the following steps:
(S1), the sample rate and local discharge sensor type matching of high speed acquisition device are set, and adjust high speed acquisition device
For low activation threshold value, partial discharge pulse's signal of continuous sampling different types of faults defect, while guaranteeing that sampling duration can cover
Cover the discharge pulse signal of multiple frequency cycles；
(S2), collected partial discharge pulse's waveform is unfolded, the list of careful analysis different types of faults defect shelf depreciation
The Time-distribution of multiple pulses on the temporal signatures and time series of a pulse, and quantify these two types of characteristic parameters；
(S3), the pulse temporal feature according to obtained in step (S2) and burst length distribution characteristics design high speed acquisition device
To the triggering collection parameter of local discharge pulse signal；
(S4), according to the triggering collection parameter of partial discharge pulse's signal in step (S3), the triggering of configuration high-speed acquisition device
Acquisition parameter is implemented to acquire to local discharge pulse information
(S5), according to the acquisition in step (S4) to local discharge pulse signal, partial discharge pulse's signal of acquisition is stored.
(S6), corresponding upper computer software is developed, is realized to the control of high speed acquisition device and data flow interaction；
(S7), partial discharge pulse's sequence is showed and is reconstructed by upper computer software, further analyze the potential valence of data
Value.
Specifically, above-mentioned steps (S1) comprising the following specific steps
(S1.1), the sample rate and local discharge sensor type matching of acquisition device are set, and adjustment high speed acquisition device is low
Threshold values trigger condition；
(S1.2), it using partial discharge pulse's signal of high speed acquisition device continuous sampling different types of faults defect, protects simultaneously
Card sampling duration can cover the discharge pulse signal of multiple frequency cycles, the return step (S1.1) if acquiring failure, if acquisition
Successfully then follow the steps (S2)；
Specifically, above-mentioned steps (S3) comprising the following specific steps
(S3.1), the single pulse waveforms of partial discharge pulse's signal of analytical procedure (S2) different types of faults defect collected
Feature and multiple-pulse Time-distribution, and quantization characteristic parameter value；
(S3.2), the touching according to the design high speed acquisition device of characteristic ginseng value obtained by step (S3.1) to local discharge pulse signal
Send out acquisition parameter；
Specifically, above-mentioned steps (S4) comprising the following specific steps
(S4.1), the triggering collection parameter of local discharge pulse signal is matched according to the high speed acquisition device that step (S3.2) is designed
Set the triggering collection parameter of high speed acquisition device；
(S4.2), using high speed acquisition device, partial discharge pulse's signal is acquired, the return step (S4.1) if acquiring failure,
(S5) is successfully thened follow the steps if capturing.
Experiment test
One, grayscale image constructing module and feature extraction and categorization module
Referring to Fig.1, according to the specific method of above-mentioned steps (2) and step (3) make grayscale image constructing module and feature extraction and
Categorization module.
Two, Partial Discharge Data acquisition platform
2.1 experimental facilities and data acquisition
The high speed acquisition device for designing Partial Discharge Data acquisition, using a variety of partial discharge phenomenons of equipment simulating and carries out data
Acquisition.High speed acquisition device be sequentially connected pass shelf depreciation sensor, signal conditioning module, high speed acquisition device (i.e. oscillograph) with
And the computer of control software is installed.
Discharge Simulation device is calibrated using calibration square wave.By tested four kinds of typical defect models (point discharge,
Particle electric discharge, bubble-discharge and suspended discharge) merging designated position, and guarantee detection device and ontology shell reliable ground.If
Determine the sampling parameter of high speed acquisition device, channel parameters, signal condition parameter；The gain of adjustment signal conditioning module to specified gear,
Shelf depreciation simulator is boosted into typical defect model firing potential, simulation point discharge, particle electric discharge, air gap are put
The electric discharge scene of four kinds of electricity, suspended discharge properties.The operation of entire platform is carried out by being equipped with the controlling terminal of control software
Manipulation, local discharge signal, which is improved by signal conditioning module after generating and is sent into high speed acquisition device, samples four kinds of typical defect moulds
The signal of type and preservation, establish corresponding PRPD spectrum data library.
2.2 data processing
The PRPD spectrum data of the corresponding four classes shelf depreciation of collected four kinds of typical defect models is rejected using 3 δ principles
Amplitude exceptional value, and the noise for introducing multiple types and phase carries out data increasing expansion, finally obtains 2280 PRPD samples, every class
570, sample.
Three, influence of the shelf depreciation number to experimental result
1824 samples are randomly selected from sample data as Experiment Training collection (456, every class sample), remaining 456 samples
As test set (114, every class sample).Test influence of the different discharge times to local electric discharge type identification network performance.It is logical
The cycle quantity of adjustment single sample data is crossed to control discharge time, by counting to sample data, obtains four kinds
Mean discharge number of the electric discharge type under the limitation of different cycles, such as Fig. 2.Classifier is counted under different cycles simultaneously to four kinds
The recognition accuracy of electric discharge type, such as Fig. 3.
Cycle can be obtained by Fig. 2 to be inversely proportional with mean discharge number.Office is effectively increased by the promotion that Fig. 3 can obtain cycle
Portion's electric discharge type identifies network to local electric discharge type recognition accuracy.Wherein particle electric discharge is with the identification of bubble-discharge by cycle
Number is affected, and the identification of point discharge and particle suspended discharge is influenced smaller by cycle.In summary in order to guarantee part
This experiment of accuracy rate of discharge mode identification uses the Partial Discharge Data sample of 50 cycles.
Four, shelf depreciation type identification result
In order to measure the shelf depreciation type identification network model performance of this experiment proposition, by the network and BP neural network, branch
Hold vector machine (SVM), the conventional machines learning method such as boosted tree carry out performance comparison, conventional machines learn input feature vector be based on
The degree of skewness of PRPD spectrum data construction, steepness, factor of discharging, phase degree of asymmetry, five kinds of phase related coefficient artificial special
Sign.Simultaneously to show influence of the introducing of residual error network for convolutional neural networks classification performance, it is permanent that this experiment tests removal
The convolutional neural networks classifier performance of equal mappings.Table 1 is each classifier for four kinds of shelf depreciation type identification accuracys rate
Comparison
| Classification | Point discharge | Particle electric discharge | Bubble-discharge | Suspended discharge |
| BP neural network | 76.14% | 52.11% | 58.25% | 55.09% |
| SVM (Gaussian kernel) | 94.38% | 73.68% | 68.07% | 68.28% |
| Boosted tree | 94.91% | 67.02% | 66.84% | 52.10% |
| Convolutional neural networks | 88.42% | 100% | 85.26% | 99.82% |
| Shelf depreciation type identification network of the invention | 96.67% | 100% | 86.67% | 99.82% |
Table 1: each classifier is for four kinds of shelf depreciation type identification accuracys rate
Obtaining the classifier of the invention constructed by table 1 has than traditional classifier better performance, except the identification of bubble-discharge is accurate
Rate is slightly below other than 90%, and the recognition accuracy of other three kinds of electric discharge types reaches 95% or more.This reflects this hair from side
The bright essential inner link that can more excavate initial data and goal task, at the same this experiment by feature that network extracts be based on
The manual features of PRPD spectrum data building carry out t-SNE dimension reduction and visualization processing, and processing result is as shown in Figure 4 and Figure 5.
The results show that the present invention has higher Clustering Effect using the feature that neural network is extracted, net is demonstrated again
The feature of network extracted in self-adaptive has better identification.Simultaneously the experimental results showed that the introducing of residual error network structure is for tip
The recognition performance of electric discharge has biggish promotion, and the recognition performance of other three classes is promoted smaller.This is because tip is put
The electric discharge grayscale image of electricity can significantly be distinguished with other electric discharge types by bottom visual signature, introduce residual error network knot
This low-level image feature can be transmitted in deeper network layer by structure by identical mapping.In addition, based on depth residual error network
Identification of the shelf depreciation type identification network model for the recognition accuracy of bubble-discharge compared to other electric discharge types is in
Apparent reduced levels, this is because data type of PRPD itself is in the presence of the shortcomings that can not embodying bubble-discharge main feature.
To sum up, for PD Pattern Recognition problem, PRPD spectrum data is converted to shelf depreciation gray scale by the present invention
Figure, and the shelf depreciation type identification network based on depth residual error network is constructed, it is experimentally confirmed and introduces residual error structure, it is real
Existing each layer information flow of network can effectively improve convolutional neural networks for the recognition capability of shelf depreciation.It is proposed by the present invention
Shelf depreciation type identification network based on residual error network effectively can carry out feature extraction to electric discharge grayscale image, compared to base
In the conventional machines learning method of manual features, have recognition accuracy higher, the preferable advantage of robustness.
The above is only a specific embodiment of the present invention, but the design concept of the present invention is not limited to this, all to utilize this
Design makes a non-material change to the present invention, and should all belong to behavior that violates the scope of protection of the present invention.
Claims (10)
1. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode, which is characterized in that including following step
It is rapid:
(1), by high speed acquisition device, the feature triggering capturing technology based on local discharge signal is to various typical fault defects
Partial discharge pulse's signal that model generates is acquired and stores, and establishes PRPD spectrum data library；
(2), by grayscale image constructing module, the grayscale image of shelf depreciation type is constructed using PRPD spectrum data；
(3), it by feature extraction and categorization module, is extracted using the shelf depreciation type identification network based on depth residual error network
The identification feature of grayscale image, and Classification and Identification is carried out to identification feature；
(4), using grayscale image constructing module and trained feature extraction and categorization module as the classifier of detecting instrument, to existing
The PRPD spectrum data of field shelf depreciation carries out feature extraction and Analysis on Fault Diagnosis.
2. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as described in claim 1, special
Sign is that the particular content of the step (2) is:
PRRD spectrum data is inputted to grayscale image constructing module, using partial discharge pulse signal amplitude q as the longitudinal axis, is divided into M
Section, operating frequency phase φ are horizontal axis, are divided into N number of section, q- φ plane is divided into M × N number of section；Count each section
Partial discharge pulse's number, construct q- φ-n map；To partial discharge pulse's numberInto
Row normalizes i.e.:；Wherein,For normalization after pulse number,For actual pulse
Number,For the maximum impulse number of the q- φ-n map；Set the gray value of each point pixel:,
MakeMaxima and minima be corresponding in turn to the minimal gray grade and maximum gray scale of gray level image, to construct
The visualization grayscale image of shelf depreciation type.
3. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as claimed in claim 1 or 2,
Be characterized in that: the shelf depreciation type identification network is by input layer, several convolutional layers, maximum value pond layer, global average pond
Change layer, full articulamentum and output layer to constitute；Wherein, input data is shelf depreciation grayscale image；Convolutional layer, maximum value pond layer
Combination realizes network characterization extracted in self-adaptive function, and the average pond layer of the overall situation realizes Feature Compression, and full articulamentum then undertakes
Classifying to input feature vector for task.
4. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as claimed in claim, feature
Be: the shelf depreciation type identification network is by input layer, the first convolutional layer, the first maximum value pond layer, the first residual error knot
Structure, the second residual error structure, the second maximum value pond layer, third residual error structure, the 4th residual error structure, global average pond layer, entirely
Articulamentum, output layer are constituted；The input data of the input layer is shelf depreciation grayscale image；The average pond layer of the overall situation is by upper one layer
For the corresponding feature point group of all characteristic patterns at output feature vector, which is the network to shelf depreciation gray scale
The feature of figure extracted in self-adaptive；Full articulamentum then undertakes classifying to input feature vector for task.
5. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as described in claim 1, special
Sign is, further includes step (5): the PRPD spectrum data by continuing to build up various typical fault defect models, while constantly
The PRPD spectrum data for collecting collected shelf depreciation when on-site test faulty equipment, constantly accumulates these data into PDPR
Spectrum data library is simultaneously trained the classifier, optimizes.
6. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as described in claim 1, special
Sign is: in the step (1), the method that high speed acquisition device acquires partial discharge pulse's signal, specifically includes the following steps:
(S1), the sample rate and local discharge sensor type matching of high speed acquisition device are set, and adjust high speed acquisition device
For low activation threshold value, partial discharge pulse's signal of continuous sampling different types of faults defect, while guaranteeing that sampling duration can cover
Cover the discharge pulse signal of multiple frequency cycles；
(S2), collected partial discharge pulse's waveform is unfolded, the list of careful analysis different types of faults defect shelf depreciation
The Time-distribution of multiple pulses on the temporal signatures and time series of a pulse, and quantify these two types of characteristic parameters；
(S3), the pulse temporal feature according to obtained in step (S2) and burst length distribution characteristics design high speed acquisition device
To the triggering collection parameter of local discharge pulse signal；
(S4), according to the triggering collection parameter of partial discharge pulse's signal in step (S3), the triggering of configuration high-speed acquisition device
Acquisition parameter is implemented to acquire to local discharge pulse information
(S5), according to the acquisition in step (S4) to local discharge pulse signal, partial discharge pulse's signal of acquisition is stored.
7. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as claimed in claim 6, special
Sign is, the step (S1) comprising the following specific steps
(S1.1), the sample rate and local discharge sensor type matching of acquisition device are set, and adjustment high speed acquisition device is low
Threshold values trigger condition；
(S1.2), it using partial discharge pulse's signal of high speed acquisition device continuous sampling different types of faults defect, protects simultaneously
Card sampling duration can cover the discharge pulse signal of multiple frequency cycles, the return step (S1.1) if acquiring failure, if acquisition
Successfully then follow the steps (S2).
8. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as claimed in claim 6, special
Sign is, the step (S3) comprising the following specific steps
(S3.1), the single pulse waveforms of partial discharge pulse's signal of analytical procedure (S2) different types of faults defect collected
Feature and multiple-pulse Time-distribution, and quantization characteristic parameter value；
(S3.2), the touching according to the design high speed acquisition device of characteristic ginseng value obtained by step (S3.1) to local discharge pulse signal
Send out acquisition parameter.
9. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as claimed in claim 6, special
Sign is, the step (S4) comprising the following specific steps
(S4.1), the triggering collection parameter of local discharge pulse signal is matched according to the high speed acquisition device that step (S3.2) is designed
Set the triggering collection parameter of high speed acquisition device；
(S4.2), using high speed acquisition device, partial discharge pulse's signal is acquired, the return step (S4.1) if acquiring failure,
(S5) is successfully thened follow the steps if capturing.
10. a kind of power equipment partial discharges fault diagnostic method based on end-to-end mode as claimed in claim 6, special
Sign is, further comprising the steps of:
(S6), corresponding upper computer software is developed, is realized to the control of high speed acquisition device and data flow interaction；
(S7), partial discharge pulse's sequence is showed and is reconstructed by upper computer software, further analyze the potential valence of data
Value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811288775.8A CN109100627A (en) | 2018-10-31 | 2018-10-31 | A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811288775.8A CN109100627A (en) | 2018-10-31 | 2018-10-31 | A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109100627A true CN109100627A (en) | 2018-12-28 |
Family
ID=64869913
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811288775.8A Pending CN109100627A (en) | 2018-10-31 | 2018-10-31 | A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109100627A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110672988A (en) * | 2019-08-29 | 2020-01-10 | 国网江西省电力有限公司电力科学研究院 | Partial discharge mode identification method based on hierarchical diagnosis |
| CN111461298A (en) * | 2020-03-26 | 2020-07-28 | 广西电网有限责任公司电力科学研究院 | Convolutional neural network and method for circuit breaker fault identification |
| CN111724289A (en) * | 2020-06-24 | 2020-09-29 | 山东建筑大学 | Environmental protection equipment identification method and system based on time sequence |
| CN111810124A (en) * | 2020-06-24 | 2020-10-23 | 中国石油大学(华东) | Oil pumping well fault diagnosis method based on characteristic re-calibration residual convolution neural network model |
| CN112147470A (en) * | 2020-09-03 | 2020-12-29 | 上海交通大学 | GIL partial discharge source positioning method and system |
| CN112149549A (en) * | 2020-09-18 | 2020-12-29 | 国网山东省电力公司泰安供电公司 | GIS partial discharge type identification method based on depth residual error network |
| CN112305388A (en) * | 2020-10-30 | 2021-02-02 | 华能澜沧江水电股份有限公司 | On-line monitoring and diagnosing method for partial discharge fault of generator stator winding insulation |
| CN112686093A (en) * | 2020-12-02 | 2021-04-20 | 重庆邮电大学 | Fusion partial discharge type identification method based on DS evidence theory |
| CN112763871A (en) * | 2020-12-30 | 2021-05-07 | 珠海华网科技有限责任公司 | Partial discharge classification identification method |
| CN114295157A (en) * | 2021-11-30 | 2022-04-08 | 国网北京市电力公司 | Mountain fire hidden danger early warning method and device, storage medium and electronic equipment |
| CN117949794A (en) * | 2024-03-27 | 2024-04-30 | 阳谷新太平洋电缆有限公司 | Cable partial discharge fault detection method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090027921A (en) * | 2007-09-13 | 2009-03-18 | 현대중공업 주식회사 | Automatic gis partial discharge identification method |
| CN106556781A (en) * | 2016-11-10 | 2017-04-05 | 华乘电气科技（上海）股份有限公司 | Shelf depreciation defect image diagnostic method and system based on deep learning |
| CN108108723A (en) * | 2018-01-19 | 2018-06-01 | 深圳市恩钛控股有限公司 | A kind of face feature extraction method based on deep learning |
| CN108573225A (en) * | 2018-03-30 | 2018-09-25 | 国网天津市电力公司电力科学研究院 | A kind of local discharge signal mode identification method and system |
| CN108627743A (en) * | 2018-05-15 | 2018-10-09 | 国网江苏省电力有限公司电力科学研究院 | A kind of lossless method for catching of shelf depreciation nanosecond narrow pulse sequence |
| CN108693448A (en) * | 2018-03-28 | 2018-10-23 | 西安博源电气有限公司 | One kind being applied to power equipment PD Pattern Recognition system |
-
2018
- 2018-10-31 CN CN201811288775.8A patent/CN109100627A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090027921A (en) * | 2007-09-13 | 2009-03-18 | 현대중공업 주식회사 | Automatic gis partial discharge identification method |
| CN106556781A (en) * | 2016-11-10 | 2017-04-05 | 华乘电气科技（上海）股份有限公司 | Shelf depreciation defect image diagnostic method and system based on deep learning |
| CN108108723A (en) * | 2018-01-19 | 2018-06-01 | 深圳市恩钛控股有限公司 | A kind of face feature extraction method based on deep learning |
| CN108693448A (en) * | 2018-03-28 | 2018-10-23 | 西安博源电气有限公司 | One kind being applied to power equipment PD Pattern Recognition system |
| CN108573225A (en) * | 2018-03-30 | 2018-09-25 | 国网天津市电力公司电力科学研究院 | A kind of local discharge signal mode identification method and system |
| CN108627743A (en) * | 2018-05-15 | 2018-10-09 | 国网江苏省电力有限公司电力科学研究院 | A kind of lossless method for catching of shelf depreciation nanosecond narrow pulse sequence |
Non-Patent Citations (2)
| Title |
|---|
| 丁清山: ""基于小波的GIS局部放电类型识别算法及其实现"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 刘兵 等: ""基于卷积神经网络的变压器局部放电模式识别"", 《高压电器》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110672988A (en) * | 2019-08-29 | 2020-01-10 | 国网江西省电力有限公司电力科学研究院 | Partial discharge mode identification method based on hierarchical diagnosis |
| CN111461298A (en) * | 2020-03-26 | 2020-07-28 | 广西电网有限责任公司电力科学研究院 | Convolutional neural network and method for circuit breaker fault identification |
| CN111724289A (en) * | 2020-06-24 | 2020-09-29 | 山东建筑大学 | Environmental protection equipment identification method and system based on time sequence |
| CN111810124A (en) * | 2020-06-24 | 2020-10-23 | 中国石油大学(华东) | Oil pumping well fault diagnosis method based on characteristic re-calibration residual convolution neural network model |
| CN111810124B (en) * | 2020-06-24 | 2023-09-22 | 中国石油大学(华东) | Oil pumping well fault diagnosis method based on characteristic recalibration residual convolutional neural network model |
| CN112147470B (en) * | 2020-09-03 | 2021-07-20 | 上海交通大学 | GIL partial discharge source positioning method and system |
| CN112147470A (en) * | 2020-09-03 | 2020-12-29 | 上海交通大学 | GIL partial discharge source positioning method and system |
| CN112149549A (en) * | 2020-09-18 | 2020-12-29 | 国网山东省电力公司泰安供电公司 | GIS partial discharge type identification method based on depth residual error network |
| CN112305388A (en) * | 2020-10-30 | 2021-02-02 | 华能澜沧江水电股份有限公司 | On-line monitoring and diagnosing method for partial discharge fault of generator stator winding insulation |
| CN112305388B (en) * | 2020-10-30 | 2024-03-08 | 华能澜沧江水电股份有限公司 | On-line monitoring and diagnosing method for insulation partial discharge faults of generator stator winding |
| CN112686093A (en) * | 2020-12-02 | 2021-04-20 | 重庆邮电大学 | Fusion partial discharge type identification method based on DS evidence theory |
| CN112763871A (en) * | 2020-12-30 | 2021-05-07 | 珠海华网科技有限责任公司 | Partial discharge classification identification method |
| CN114295157A (en) * | 2021-11-30 | 2022-04-08 | 国网北京市电力公司 | Mountain fire hidden danger early warning method and device, storage medium and electronic equipment |
| CN117949794A (en) * | 2024-03-27 | 2024-04-30 | 阳谷新太平洋电缆有限公司 | Cable partial discharge fault detection method |
| CN117949794B (en) * | 2024-03-27 | 2024-06-04 | 阳谷新太平洋电缆有限公司 | Cable partial discharge fault detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109100627A (en) | A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode | |
| CN109116203A (en) | Power equipment partial discharges fault diagnostic method based on convolutional neural networks | |
| CN108573225A (en) | A kind of local discharge signal mode identification method and system | |
| CN108509976A (en) | The identification device and method of animal | |
| CN110514924A (en) | The Winding in Power Transformer Fault Locating Method of depth convolutional neural networks fusion visual recognition | |
| CN110108992B (en) | Cable partial discharge fault identification method and system based on improved random forest algorithm | |
| CN105205396A (en) | Detecting system for Android malicious code based on deep learning and method thereof | |
| CN108597053A (en) | Shaft tower and channel targets identification based on image data and neural network and defect diagnostic method | |
| CN110166484A (en) | A kind of industrial control system intrusion detection method based on LSTM-Attention network | |
| CN105678332A (en) | Converter steel-making endpoint determination method and system based on flame image CNN recognizing and modeling process | |
| CN108562821B (en) | Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax | |
| CN107909118A (en) | A kind of power distribution network operating mode recording sorting technique based on deep neural network | |
| CN108809948A (en) | A kind of abnormal network connecting detection method based on deep learning | |
| CN108693448B (en) | Partial discharge mode recognition system applied to power equipment | |
| CN110261746B (en) | Cable defect detection method based on periodic attenuation characteristics of oscillating wave voltage | |
| CN109410184A (en) | Live streaming pornographic image detection method based on dense confrontation network semi-supervised learning | |
| CN107942210A (en) | The classification of transformer pulse electric current Partial Discharge and denoising method and system | |
| CN111126820A (en) | Electricity stealing prevention method and system | |
| CN108334902A (en) | A kind of track train equipment room smog fireproof monitoring method based on deep learning | |
| CN110020637A (en) | A kind of analog circuit intermittent fault diagnostic method based on more granularities cascade forest | |
| CN109061426A (en) | Partial discharge of transformer method for diagnosing faults and on-Line Monitor Device | |
| CN110415260A (en) | Smog image segmentation and recognition methods based on dictionary and BP neural network | |
| CN110493262A (en) | It is a kind of to improve the network attack detecting method classified and system | |
| CN117056678B (en) | Machine pump equipment operation fault diagnosis method and device based on small sample | |
| CN116466408B (en) | Artificial neural network superbedrock identification method based on aeromagnetic data |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181228 |
|
| RJ01 | Rejection of invention patent application after publication |