CN108303624A - A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis - Google Patents
A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis Download PDFInfo
- Publication number
- CN108303624A CN108303624A CN201810099032.XA CN201810099032A CN108303624A CN 108303624 A CN108303624 A CN 108303624A CN 201810099032 A CN201810099032 A CN 201810099032A CN 108303624 A CN108303624 A CN 108303624A
- Authority
- CN
- China
- Prior art keywords
- feature
- switch cabinet
- detection
- sound
- voice signal
- 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
- G01R31/1209—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 using acoustic measurements
Abstract
The invention discloses a kind of method for detection of partial discharge of switch cabinet based on voice signal analysis, are related to, include the following steps：S1, acquisition high-tension switch cabinet insulation breakdown electric discharge sound training sample；S2, training sample is pre-processed, calculates its sound spectrograph, and extract significant characteristics；S3, deep learning model of the structure based on sound spectrograph are simultaneously trained；S4, electric discharge sound to be measured similarly pre-process with step S2, the grader that step S3 is generated is sent into after extraction characteristic parameter and completes final detection identification.Trained grader can identify electric discharge sound when high-tension switch cabinet insulation breakdown, to carry out real-time fault detection, the working condition of real-time display high-tension switch cabinet.The present invention has the advantages of accuracy of detection is high, strong robustness.
Description
Technical field
The present invention relates to a kind of method for detection of partial discharge of switch cabinet, more particularly to utilize voice signal detection switch cabinet office
The method of portion's electric discharge.
Background technology
Switchgear is one of capital equipment of distribution, is the most key electrical during power generation conveying and use
Equipment, while being also a kind of higher equipment of rate of breakdown.Characteristics of Partial Discharge in switchgear can accurately reflect switchgear
Equipment different phase damaged condition, using this characteristic to switchgear state of insulation carry out online hotline maintenance be it is capable it
Effective approach.
General maintenance behave often expends a large amount of human and material resources, financial resources, and malfunction elimination is not in time, cannot eliminate
Potential security risk, often causes serious accident.And safely and effectively live detection can not only be substantially reduced and be set
The possibility of standby damage, extends the service life of switchgear, and entire power supply system can also be made to possess guarantee steady in a long-term.
In recent years, the detection of partial discharge of switchgear is always the research hotspot and difficult point of lot of domestic and foreign scholar, some
Detection method is also come into being, such as arc light detecting, ultrasound examination, ultra-high-frequency detection and audio sound event detection etc..It is high
The cabinet that compresses switch is run in the environment of closing, therefore, the normal operation of switchgear is not interfered with for service work, numerous scholars are
It is more prone to use Noninvasive testing.The process of shelf depreciation be usually associated with emit light and heat, corona sound even breakdown sound,
This process can carry out the maintenance and maintenance of switchgear using the method for audio event detection according to the range of audible sound.
Meanwhile according to the propagating characteristic of voice signal, the source of trouble can also be accurately positioned using audio event detection.But current
The feature of research is mainly MFCC features.MFCC is one of common feature of speech recognition, but is used for non-speech recognition, real
Border effect also needs to verify.
Invention content
The technical problem to be solved in the present invention：For the above problem of the prior art, it is efficient to provide a kind of discharge examination,
The good method for detection of partial discharge of switch cabinet based on voice signal analysis of robustness.
In order to solve the above-mentioned technical problem, the technical solution adopted by the present invention is：
A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis, includes the following steps：
S1, acquisition high-tension switch cabinet insulation breakdown electric discharge sound training sample；
S2, training sample is pre-processed, calculates its sound spectrograph, and extract significant characteristics；
S3, deep learning model of the structure based on sound spectrograph are simultaneously trained；
S4, electric discharge sound to be measured similarly pre-process with step S2, step S3 lifes are sent into after extracting characteristic parameter
At grader complete final detection identification.
Preferably, the calculating step of the significant characteristics of sound spectrograph is in step S2：
S21, voice signal is converted into sound spectrograph by Fast Fourier Transform (FFT)；
S22, the significance feature for calculating each channel；
The significance for calculating each feature by Core-Periphery difference operator based on sound spectrograph, obtains different channels and different rulers
Significance feature under degree indicates intensity I variations, orientation OθWhat feature, frequency F variations, time-frequency T features and tone P changed
Notable figure.Specific formula is as follows：
M (c, s)=| M (c) Θ M (s) |, M ∈ { I, Oθ,F,T,P} (1)
Wherein, center scale c ∈ { 2,3,4 }, s=c+ δ, and δ ∈ { 3,4 }.The difference meter of center scale c and surrounding scale s
Calculation is indicated with symbol Θ.
S23, extraction eigenmatrix；
Every width characteristic pattern is divided into s rows t row, total s × t sub-regions substitute the sub-district with the median of every sub-regions
Image is further normalized to the eigenmatrix of s × t by domain, to be described with the image characteristic matrix of a low resolution
Entire language spectrum.
S24, Feature Dimension Reduction simultaneously reconstruct, and obtain main feature vector；
The vector for being 1 × st by the corresponding eigenmatrix remodeling of every width characteristic pattern, and by these vectorial constitutive characteristic vectors
Matrix.Calculating within class scatter matrix is：
Wherein, i is class number, and j is sample number, miFor the mean value of the i-th class sample, NiFor number of samples.C classification
Overall within class scatter matrix be：
Wherein, PiFor the prior probability of the i-th class.Inter _ class relationship matrix is defined as：
Wherein, m is the mean vector of all samples.
Thus obtained transformation matrix passes through the u of calculatingiIt can be obtained in original feature space dimensionality reduction to lower dimensional space
Main feature vector.
Preferably, it is 6 that s values, which are 5, t values,.
Preferably, the training process of the deep learning model of sound spectrograph specifically comprises the following steps in step S3：
S31, initialization deep learning network structure, determine the sample of training data, frequency of training L, it is seen that layer and hidden layer
Neuromere count m and n, learning rate ε, initialize hidden layer and visual layers bias term aiAnd bjAnd two layers neuron node it
Between weights Wij, wherein i and j respectively represents the number of plies of hidden layer and visual layers；
S32, partial derivative Δ W of the likelihood function to bias term and weights is found outij, Δ ai, Δ bj, and to bias term and weights
It is updated；
Δbj≈pθ(hj=1 | v0)-pθ(hj=1 | vk) (8)
Wherein, pθThe probability of representation parameter θ, h and v respectively represent hidden layer and visual layers, and k represents iterations.
S33, step S31 and step S32 is repeated, until reaching frequency of training L.
The present invention is by adopting the above-described technical solution, have the advantages that：
(1) partial discharge of switchgear detection model analyze based on voice signal of present invention structure, different from electricity or
Ultrasonic signal is the conventional method of research object, and model is first using voice signal as research object；Then with the sound spectrograph of sound
Characterized by essential characteristic, and combine conspicuousness model extraction notable feature；Electric discharge letter is finally identified using deep learning network
Number, detection efficiency of the invention is high；
(2) the method for the present invention detection efficiency is high, and robustness is high, and the environmental suitability of method is strong.
Description of the drawings
Attached drawing is used for providing the preferred understanding to the present invention, and a part for constitution instruction, the reality with the present invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings：
Fig. 1 is the partial discharge of switchgear algorithmic system structure chart of present invention method；
Fig. 2 is the deep learning network structure in the embodiment of the present invention；
Fig. 3 is the method robustness comparing result in the embodiment of the present invention.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that the embodiment of the description
It is a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The every other embodiment that member is obtained without creative efforts, shall fall within the protection scope of the present invention.
As shown in Figure 1, the implementation steps for the method for detection of partial discharge of switch cabinet that the present embodiment is analyzed based on voice signal
Including：1) acquisition high-tension switch cabinet insulation breakdown electric discharge sound training sample；2) training sample is pre-processed, calculates its language spectrum
Figure, and extract significant characteristics；3) structure based on deep learning network and is trained；4) electric discharge sound to be measured is carried out and is walked
It is sent into the grader that step 3) generates after rapid 2) same pretreatment, characteristic parameter extraction and completes final detection identification.
The calculating step of the significant characteristics of sound spectrograph is in step 2)：
S21, voice signal is converted into sound spectrograph by Fast Fourier Transform (FFT)；
S22, the significance feature for calculating each channel；
The significance for calculating each feature by Core-Periphery difference operator based on sound spectrograph, obtains different channels and different rulers
Significance feature under degree indicates intensity I variations, orientation OθWhat feature, frequency F variations, time-frequency T features and tone P changed
Notable figure.Specific formula is as follows：
M (c, s)=| M (c) Θ M (s) |, M ∈ { I, Oθ,F,T,P} (9)
Wherein, center scale c ∈ { 2,3,4 }, s=c+ δ, and δ ∈ { 3,4 }.The difference meter of center scale c and surrounding scale s
Calculation is indicated with symbol Θ.
S23, extraction eigenmatrix；
Every width characteristic pattern is divided into s rows t row, total s × t sub-regions substitute the sub-district with the median of every sub-regions
Image is further normalized to the eigenmatrix of s × t by domain, to be described with the image characteristic matrix of a low resolution
Entire language spectrum；Here, it is 6 that s values, which are 5, t values,.
S24, Feature Dimension Reduction simultaneously reconstruct, and obtain main feature vector；
The vector for being 1 × st by the corresponding eigenmatrix remodeling of every width characteristic pattern, and by these vectorial constitutive characteristic vectors
Matrix.Calculating within class scatter matrix is：
Wherein, i is class number, and j is sample number, miFor the mean value of the i-th class sample, NiFor number of samples.C classification
Overall within class scatter matrix be：
Wherein, PiFor the prior probability of the i-th class.Inter _ class relationship matrix is defined as：
Wherein, m is the mean vector of all samples.
Thus obtained transformation matrix passes through the u of calculatingiIt can be obtained in original feature space dimensionality reduction to lower dimensional space
Main feature vector.
Fig. 2 show the deep learning model for the method for detection of partial discharge of switch cabinet structure analyzed based on voice signal.
When known to visual layers, the neurode in hidden layer h is conditional sampling, i.e.,：P (h | v)=p (h1|v)p(h2|v)…p(hn
| it is conditional sampling between the neurode of visual layers v, i.e., when known to hidden layer h equally v)：P (v | h)=p (v1|h)
p(v2|h)…p(vm| h), since all hidden layer h nodes and visual layers v nodes are satisfied by ANALOGY OF BOLTZMANN DISTRIBUTION, when from
When visual layers v input, hidden layer h can be obtained by the probability distribution p of visual layers v to hidden layer h (h | v), and according to from implying
Layer h to visual layers v probability distribution p (vh) and visual layers v1 can be obtained, by network parameter is adjusted so that from imply
When the visual layers v1 that layer h is obtained is identical as original visual layers v, hidden layer h can be indicated as the another of visual layers v, i.e. conduct
The feature of visual layers v input datas.
The energy function of deep learning network is represented by：
M is visual layers number of nodes, and n is node in hidden layer, WijFor i-th of node of visible layer and j-th of node of hidden layer
Between weights, θ={ Wij,ai,biBe deep learning network parameter, aiAnd biIt is visible layer and the bias term of hidden layer, root
According to ANALOGY OF BOLTZMANN DISTRIBUTION and deep learning network energy function, the joint probability distribution that can be obtained between visual layers and hidden layer is：
Understand that the marginal probability distribution of visible layer neuron is：
The target for learning deep learning network is to seek parameter θ, and according to given training data, θ can be by seeking maximum likelihood
Estimation obtains：
Obtain optimal parameter θ*, can be to likelihood functionIt is asked most using stochastic gradient rise method
Big value.
Log-likelihood function is about the gradient of θ：
Wherein<X>PFor the mathematic expectaion of the X on probability distribution P, Pθ(h|vi) be visible layer sample be viWhen hidden layer
Probability distribution, likelihood function is to weight Wij, it is seen that the bias term a of layer and hidden layeriAnd biDerivation be：
Wherein ε is learning rate, since each neuron in deep learning network only takes 0 or 1 two states, so v
It is only 0 or 1 with h, can be obtained：
It can similarly obtain：
When due to known to visual layers, the neurode in hidden layer h is conditional sampling, i.e.,：P (h | v)=p (h1|v)p
(h2|v)…p(hn| v), decomposition can obtain：
The p for including in the latter of all kinds of partial derivatives of likelihood functionθ(v, h) indicates the Joint Distribution of visual layers and hidden layer,
Since there are Zθ, so being difficult to be found out with calculating, approximation is generally obtained with to sdpecific dispersion algorithm by gibbs sampler.Algorithm
Using the original state of visual layers as training sample, K gibbs sampler is carried out, can be led in the gibbs sampler of kth
Cross pθ(h|vk-1) obtain hk-1, pass through pθ(v|hk-1) obtain vk, then use vkTo likelihood function to the derivative of weights and bias term
It is estimated：
Δbj≈pθ(hj=1 | v0)-pθ(hj=1 | vk) (26)
The training process that deep learning network can be obtained is：
1) structure for initializing deep learning network, determines the sample of training data, frequency of training L, it is seen that layer and hidden layer
Neuromere count m and n, learning rate ε, initialize hidden layer and visual layers bias term aiAnd bjAnd two layers neuron node it
Between weights Wij；
2) partial derivative Δ W of the likelihood function to bias term and weights is found outij, Δ ai, Δ bj, and to bias term and weights into
Row update；
1) and 2) 3) two step is repeated, until reaching frequency of training L.
In order to which the performance for the method for detection of partial discharge of switch cabinet analyzed based on voice signal the present embodiment is compared,
It is respectively provided with three kinds of workplaces power plant, substation and the factory floor of switchgear.Test data is superimposed for initial data
The noise of different signal-to-noise ratio (0dB, 5dB, 10dB, 15dB, 20dB, 25dB, 30dB).The main Types of noise include white noise,
Pink noise and factory noise.From the figure 3, it may be seen that influence of the pink noise to algorithm is minimum, white noise influences maximum.From trend
It sees, algorithm detects error rate and gradually rise with the reduction of signal-to-noise ratio.When signal-to-noise ratio reaches 25dB or higher, algorithm is not
It can ensure within 20% with the error rate under environment, illustrate that algorithm is good to the robustness of different noise circumstances.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, although with reference to aforementioned reality
Applying example, invention is explained in detail, for those skilled in the art, still can be to previous embodiment
Recorded technical solution is modified or equivalent replacement of some of the technical features.All spirit in the present invention
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (4)
1. a kind of method for detection of partial discharge of switch cabinet based on voice signal analysis, includes the following steps：
S1, acquisition high-tension switch cabinet insulation breakdown electric discharge sound training sample；
S2, training sample is pre-processed, calculates its sound spectrograph, and extract significant characteristics；
S3, deep learning model of the structure based on sound spectrograph are simultaneously trained；
S4, electric discharge sound to be measured similarly pre-process with step S2, be sent into what step S3 was generated after extracting characteristic parameter
Grader completes final detection identification.
2. a kind of method for detection of partial discharge of switch cabinet based on voice signal analysis according to claim 1, feature
It is, the calculating step of the significant characteristics of sound spectrograph is in step S2：
S21, voice signal is converted into sound spectrograph by Fast Fourier Transform (FFT)；
S22, the significance feature for calculating each channel；
The significance for being calculated each feature by Core-Periphery difference operator based on sound spectrograph, is obtained under different channels and different scale
Significance feature, that is, indicate intensity I variation, orientation OθFeature, frequency F variations, time-frequency T features and tone P change notable
Figure.Specific formula is as follows：
M (c, s)=| M (c) Θ M (s) |, M ∈ { I, Oθ,F,T,P} (1)
Wherein, center scale c ∈ { 2,3,4 }, s=c+ δ, and δ ∈ { 3,4 }.The difference of center scale c and surrounding scale s, which calculates, to be used
Symbol Θ is indicated.
S23, extraction eigenmatrix；
Every width characteristic pattern is divided into s rows t row, total s × t sub-regions substitute the subregion with the median of every sub-regions, will
Image is further normalized to the eigenmatrix of s × t, to describe entire language with the image characteristic matrix of a low resolution
Spectrum.
S24, Feature Dimension Reduction simultaneously reconstruct, and obtain main feature vector；
The vector for being 1 × st by the corresponding eigenmatrix remodeling of every width characteristic pattern, and by these vectorial constitutive characteristic vector matrixs.
Calculating within class scatter matrix is：
Wherein, i is class number, and j is sample number, miFor the mean value of the i-th class sample, NiFor number of samples.C classification it is total
Body within class scatter matrix is：
Wherein, PiFor the prior probability of the i-th class.Inter _ class relationship matrix is defined as：
Wherein, m is the mean vector of all samples.
Thus obtained transformation matrix passes through the u of calculatingiMain feature can be obtained by original feature space dimensionality reduction to lower dimensional space
Vector.
3. a kind of method for detection of partial discharge of switch cabinet based on voice signal analysis according to claim 2, feature
It is, s values are that 5, t values are 6.
4. a kind of method for detection of partial discharge of switch cabinet based on voice signal analysis according to claim 1, feature
It is, the training process of the deep learning model of sound spectrograph specifically comprises the following steps in step S3：
S31, initialization deep learning network structure, determine the sample of training data, frequency of training L, it is seen that the god of layer and hidden layer
Through number of nodes m and n, learning rate ε initializes the bias term a of hidden layer and visual layersiAnd bjAnd between two layers of neuron node
Weights Wij, wherein i and j respectively represents the number of plies of hidden layer and visual layers；
S32, partial derivative Δ W of the likelihood function to bias term and weights is found outij, Δ ai, Δ bj, and bias term and weights are carried out
Update；
Δbj≈pθ(hj=1 | v0)-pθ(hj=1 | vk) (8)
Wherein, pθThe probability of representation parameter θ, h and v respectively represent hidden layer and visual layers, and k represents iterations.
S33, step S31 and step S32 is repeated, until reaching frequency of training L.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810099032.XA CN108303624A (en) | 2018-01-31 | 2018-01-31 | A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810099032.XA CN108303624A (en) | 2018-01-31 | 2018-01-31 | A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108303624A true CN108303624A (en) | 2018-07-20 |
Family
ID=62850659
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810099032.XA Pending CN108303624A (en) | 2018-01-31 | 2018-01-31 | A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108303624A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109752633A (en) * | 2019-01-28 | 2019-05-14 | 国网山东省电力公司日照供电公司 | The method and system that a kind of pair of transformer station partial discharge signals are positioned |
| CN109814013A (en) * | 2019-01-28 | 2019-05-28 | 上海交通大学 | A kind of pair of shelf depreciation carries out the method and system of accurate pointing |
| CN110244204A (en) * | 2019-06-27 | 2019-09-17 | 国网湖南省电力有限公司 | A kind of switchgear method for diagnosing faults, system and the medium of multiple characteristic values |
| CN110334936A (en) * | 2019-06-28 | 2019-10-15 | 阿里巴巴集团控股有限公司 | A kind of construction method, device and the equipment of credit qualification Rating Model |
| CN112946442A (en) * | 2021-04-12 | 2021-06-11 | 厦门理工学院 | Switch cabinet partial discharge detection method, terminal equipment and storage medium |
| CN113297922A (en) * | 2021-04-30 | 2021-08-24 | 广西电网有限责任公司电力科学研究院 | High-voltage switch cabinet fault diagnosis method and device and storage medium |
| CN113608082A (en) * | 2021-07-30 | 2021-11-05 | 环宇集团（南京）有限公司 | Ring main unit discharge state identification method based on audio signal |
| CN113642714A (en) * | 2021-08-27 | 2021-11-12 | 国网湖南省电力有限公司 | Insulator pollution discharge state identification method and system based on small sample learning |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104637497A (en) * | 2015-01-16 | 2015-05-20 | 南京工程学院 | Speech spectrum characteristic extracting method facing speech emotion identification |
| CN104794534A (en) * | 2015-04-16 | 2015-07-22 | 国网山东省电力公司临沂供电公司 | Power grid security situation predicting method based on improved deep learning model |
| CN106295717A (en) * | 2016-08-30 | 2017-01-04 | 南京理工大学 | A kind of western musical instrument sorting technique based on rarefaction representation and machine learning |
| CN106710589A (en) * | 2016-12-28 | 2017-05-24 | 百度在线网络技术（北京）有限公司 | Artificial intelligence-based speech feature extraction method and device |
| CN106709820A (en) * | 2017-01-11 | 2017-05-24 | 中国南方电网有限责任公司电网技术研究中心 | Electrical power system load prediction method and device based on depth belief network |
| CN106782505A (en) * | 2017-02-21 | 2017-05-31 | 南京工程学院 | A kind of method based on electric discharge voice recognition high-tension switch cabinet state |
| CN107392364A (en) * | 2017-07-12 | 2017-11-24 | 河海大学 | The short-term load forecasting method of variation mode decomposition and depth belief network |
-
2018
- 2018-01-31 CN CN201810099032.XA patent/CN108303624A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104637497A (en) * | 2015-01-16 | 2015-05-20 | 南京工程学院 | Speech spectrum characteristic extracting method facing speech emotion identification |
| CN104794534A (en) * | 2015-04-16 | 2015-07-22 | 国网山东省电力公司临沂供电公司 | Power grid security situation predicting method based on improved deep learning model |
| CN106295717A (en) * | 2016-08-30 | 2017-01-04 | 南京理工大学 | A kind of western musical instrument sorting technique based on rarefaction representation and machine learning |
| CN106710589A (en) * | 2016-12-28 | 2017-05-24 | 百度在线网络技术（北京）有限公司 | Artificial intelligence-based speech feature extraction method and device |
| CN106709820A (en) * | 2017-01-11 | 2017-05-24 | 中国南方电网有限责任公司电网技术研究中心 | Electrical power system load prediction method and device based on depth belief network |
| CN106782505A (en) * | 2017-02-21 | 2017-05-31 | 南京工程学院 | A kind of method based on electric discharge voice recognition high-tension switch cabinet state |
| CN107392364A (en) * | 2017-07-12 | 2017-11-24 | 河海大学 | The short-term load forecasting method of variation mode decomposition and depth belief network |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109752633A (en) * | 2019-01-28 | 2019-05-14 | 国网山东省电力公司日照供电公司 | The method and system that a kind of pair of transformer station partial discharge signals are positioned |
| CN109814013A (en) * | 2019-01-28 | 2019-05-28 | 上海交通大学 | A kind of pair of shelf depreciation carries out the method and system of accurate pointing |
| CN109752633B (en) * | 2019-01-28 | 2020-12-04 | 国网山东省电力公司日照供电公司 | Method and system for positioning partial discharge signal of transformer substation |
| CN110244204A (en) * | 2019-06-27 | 2019-09-17 | 国网湖南省电力有限公司 | A kind of switchgear method for diagnosing faults, system and the medium of multiple characteristic values |
| CN110334936A (en) * | 2019-06-28 | 2019-10-15 | 阿里巴巴集团控股有限公司 | A kind of construction method, device and the equipment of credit qualification Rating Model |
| CN110334936B (en) * | 2019-06-28 | 2023-09-29 | 创新先进技术有限公司 | Method, device and equipment for constructing credit qualification scoring model |
| CN112946442B (en) * | 2021-04-12 | 2022-06-28 | 厦门理工学院 | Switch cabinet partial discharge detection method, terminal equipment and storage medium |
| CN112946442A (en) * | 2021-04-12 | 2021-06-11 | 厦门理工学院 | Switch cabinet partial discharge detection method, terminal equipment and storage medium |
| CN113297922A (en) * | 2021-04-30 | 2021-08-24 | 广西电网有限责任公司电力科学研究院 | High-voltage switch cabinet fault diagnosis method and device and storage medium |
| CN113608082A (en) * | 2021-07-30 | 2021-11-05 | 环宇集团（南京）有限公司 | Ring main unit discharge state identification method based on audio signal |
| CN113608082B (en) * | 2021-07-30 | 2024-03-22 | 环宇集团（南京）有限公司 | Ring main unit discharge state identification method based on audio signals |
| CN113642714A (en) * | 2021-08-27 | 2021-11-12 | 国网湖南省电力有限公司 | Insulator pollution discharge state identification method and system based on small sample learning |
| CN113642714B (en) * | 2021-08-27 | 2024-02-09 | 国网湖南省电力有限公司 | Insulator pollution discharge state identification method and system based on small sample learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108303624A (en) | A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis | |
| Deokar et al. | Integrated DWT–FFT approach for detection and classification of power quality disturbances | |
| US11480497B2 (en) | Anomalous sound detection training apparatus, acoustic feature extraction apparatus, anomalous sound sampling apparatus, and methods and programs for the same | |
| CN108761287B (en) | Transformer partial discharge type identification method | |
| US20200271714A1 (en) | Method and system of partial discharge recognition for diagnosing electrical networks | |
| Deb et al. | Short time modified Hilbert transform-aided sparse representation for sensing of overhead line insulator contamination | |
| Wijayakulasooriya et al. | Electric power quality disturbance classification using self-adapting artificial neural networks | |
| Hubana et al. | Approach for identification and classification of HIFs in medium voltage distribution networks | |
| Biswal et al. | Nonstationary power signal time series data classification using LVQ classifier | |
| WO2022088643A1 (en) | Fault diagnosis method and apparatus for buried transformer substation, and electronic device | |
| Su et al. | Classification of partial discharge events in GILBS using probabilistic neural networks and the fuzzy c-means clustering approach | |
| Al-geelani et al. | A review on hybrid wavelet regrouping particle swarm optimization neural networks for characterization of partial discharge acoustic signals | |
| Jena et al. | Variational mode decomposition‐based power system disturbance assessment to enhance WA situational awareness and post‐mortem analysis | |
| Li et al. | Partial discharge detection of insulated conductors based on CNN-LSTM of attention mechanisms | |
| Pan et al. | Learning approach based DC arc fault location classification in DC microgrids | |
| CN112581940A (en) | Discharging sound detection method based on edge calculation and neural network | |
| Orellana et al. | Danger assessment of the partial discharges temporal evolution on a polluted insulator using UHF measurement and deep learning | |
| KR20200116425A (en) | RNN based partial discharge detection apparatus and method in gas-insulated switchgear | |
| Duangpummet et al. | A robust method for blindly estimating speech transmission index using convolutional neural network with temporal amplitude envelope | |
| Dang et al. | Cochlear filter cepstral coefficients of acoustic signals for mechanical faults identification of power transformer | |
| Niazazari et al. | Event cause analysis in distribution networks using synchro waveform measurements | |
| Hao et al. | Insulator fouling monitoring based on acoustic signal and one-dimensional convolutional neural network | |
| Liu et al. | Underdetermined blind source separation algorithm of 220kV substation noise based on SCA | |
| Kannan et al. | A random forest model based pollution severity classification scheme of high voltage transmission line insulators | |
| Muniraj et al. | Adaptive neurofuzzy inference system-based pollution severity prediction of polymeric insulators in power transmission lines |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180720 |