CN110082082A - A kind of GIS state identification method based on vibration signal Principal Component Analysis - Google Patents
A kind of GIS state identification method based on vibration signal Principal Component Analysis Download PDFInfo
- Publication number
- CN110082082A CN110082082A CN201910350848.XA CN201910350848A CN110082082A CN 110082082 A CN110082082 A CN 110082082A CN 201910350848 A CN201910350848 A CN 201910350848A CN 110082082 A CN110082082 A CN 110082082A
- Authority
- CN
- China
- Prior art keywords
- gis
- vibration signal
- principal component
- feature
- energy
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
-
- 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
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a kind of GIS state identification methods based on vibration signal Principal Component Analysis, extract the compound characteristics vector of 14 characteristic quantities composition, 14 dimension of GIS vibration signal under normal and different faults state；Then compression dimensionality reduction is carried out into principal component feature vector to feature vector with Principal Component Analysis；Then the corresponding decision function between vibration signal principal component feature vector and GIS state is obtained by the two stage training of depth confidence network, classified using its GIS vibration signal principal component feature vector for treating state recognition, determines the GIS state according to classifying.Compound characteristics vector is optimized using Principal Component Analysis, the raw information of compound characteristics vector is not only remained, but also reduce feature vector dimension, improves the working efficiency of classifier, effectively improve the accuracy and speed of GIS state recognition.
Description
Technical field
The present invention relates to Cubicle Gas-Insulated Switchgear (GIS) status recognition technique fields, and in particular to a kind of
GIS state identification method based on vibration signal Principal Component Analysis.
Background technique
Cubicle Gas-Insulated Switchgear (Gas Insulated Switchgear, GIS) was come out in last century 60
Age, because its occupied area is small, high reliablity, high safety, installation period are short, the skilful maintenance workload of fortune small the advantages that due to it is at full speed
Development, is worldwide used widely in each grade transformer substation.But due to its complicated full-closed structure, GIS
Once breaking down all to have a wide range of influence and be difficult to standard and going positioning, rapid rush-repair.
Therefore, in order to ensure that the safe and stable operation of power grid, the reliability of GIS is just particularly important, therefore the event to GIS
Barrier state carries out effectively identification and has become the task of top priority.The means for GIS equipment operational condition monitoring have ultrasonic wave local at present
Discharge examination, superfrequency detection, optical analysis and chemical analysis etc., but although these methods have obtained researcher or scene
Staff's is widely recognized as, but is all only applicable to GIS internal discharge fault detection, it is difficult to identify the mechanical breakdown in GIS.
And vibration signals spectrograph is characterized in that the important indicator for characterizing mechanical features is based on due to its sensibility to mechanical breakdown in GIS
The GIS state recognition of vibration signal has become recent study hot spot.
And the single characteristics quantity that the existing GIS state recognition based on vibration signal is often directed to vibration signal carries out GIS
State recognition, but the machine performance of GIS and characteristic quantity are not one-to-one relationship, and different machine performances may cause
Same Feature change, therefore erroneous judgement often will cause according to the GIS state recognition of single characteristics quantity.It is quasi- in order to improve detection
True rate, it is necessary to various features amount composition compound characteristics vector is acquired, it is right in conjunction with the message complementary sense relationship between different characteristic amount
GIS state makes more reliable judgement, but the increase of characteristic quantity can aggravate the work load of computer, reduce calculating speed with
Precision.
Summary of the invention
The technical problems to be solved by the present invention are: the existing GIS state recognition based on vibration signal is in different characteristic
In the selection of amount, in conjunction with the message complementary sense relationship between different characteristic amount, more reliable judgement is made to GIS state, but special
The problem of increase of sign amount can aggravate the work load of computer, reduce calculating speed and precision.Therefore, it is carrying out based on vibration
When the GIS state recognition of signal, it is particularly important to choose suitable characteristic quantity.The present invention provides a kind of bases to solve the above problems
In the GIS state identification method of vibration signal Principal Component Analysis, while retaining GIS vibration signal characteristics as much as possible again
Have higher state recognition efficiency, can high efficiency, accurately realize the GIS state judgement of comprehensive various features amount.
The present invention is achieved through the following technical solutions:
A kind of GIS state identification method based on vibration signal Principal Component Analysis, which comprises
Step 1: multiple groups GIS is acquired by the vibration acceleration sensor installed on GIS normally and under malfunction
Vibration signal；
Step 2: collected GIS vibration signal being handled, extracts GIS vibration signal time domain, frequency domain and energy respectively
Measure feature, and construct GIS vibration signal compound characteristics vector；
Step 3: processing being optimized to GIS vibration signal compound characteristics vector using Principal Component Analysis, obtains GIS vibration
Dynamic signal principal component feature vector；
Step 4: building depth confidence network model, by the GIS vibration signal principal component feature under multiple groups known state to
It measures the training sample as depth confidence Network Recognition model and GIS is obtained by the two stage training of depth confidence network model
Corresponding decision function between vibration signal principal component feature vector and GIS state；
Step 5: GIS vibration signal of the acquisition to state recognition carries out the calculating and analysis of principal component feature vector, uses
Corresponding decision function between GIS vibration signal principal component feature vector and GIS state carries out Classification and Identification to it, identifies this
GIS state.
Further, the GIS vibration signal temporal signatures extracted in step 2 include GIS vibration signal peak-to-peak value, are averaged
Value, the degree of bias and kurtosis；
Frequency domain character includes principal component frequency, 100Hz accounting, 50Hz odd times frequency multiplication accounting；
GIS vibration signal is decomposed as multiple modal components IMF using set empirical mode decomposition algorithm, calculates each IMF
Energy, taking in original energy accounting is more than the energy of 90% preceding 7 modal components IMF as GIS vibration signal energy
Measure feature.
Further, above-mentioned 14 dimensional feature amount (the 14 dimensional feature amounts are as follows: GIS vibration signal extracted according to GIS vibration signal
Peak-to-peak value, average value, the degree of bias and kurtosis, principal component frequency, 100Hz accounting, 50Hz odd times frequency multiplication accounting, in original energy
Accounting is more than the energy of 90% preceding 7 modal components IMF, and successively carries out the number of characteristic quantity in the order described above) building
GIS vibration signal compound characteristics vector α=(k1,k2... k14)T, in which: α is GIS vibration signal compound characteristics vector, k1For
First dimensional feature amount peak-to-peak value, k2For the second dimensional feature amount average value, other and so on, subscript T is (k1,k2... k14)
Transposition.
Further, the step 3 specifically includes:
Step 3.1: surveyed m group GIS vibration signal is constituted to the matrix of m × 14
X=(α1,α2... αm)T=(x1,x2... x14)
Step 3.2: according to the covariance matrix C of following formula calculating matrix X:
In formula: cov (x, y) indicates the covariance of two groups of data；
Step 3.3: calculating the eigenvalue λ of covariance matrix Ci(i=1,2......14) with corresponding eigenvectors matrix
E arranges obtained characteristic value in descending order, and will respectively arrange rearrange in eigenvectors matrix E in this order, obtains transition square
Battle array T, takes each column vector in T to be characterized the factor；
Step 3.4: calculating each characterization factor contribution rate, choose 2 characterization factors that wherein the sum of contribution rate is more than 95%
Form transformation matrix, each characterization factor contribution rate calculation formula are as follows:
In formula: KrIndicate r-th of characterization factor contribution rate, λrIndicate the corresponding characteristic value of r-th of characterization factor, λjIt indicates
The corresponding characteristic value of j-th of characterization factor,It indicates to λjFrom λ1To λmSummation；
Step 3.5: the eigenmatrix Y of m × 2 being obtained by matrix operation Y=X × U, each column of matrix Y are one group
The principal component feature vector of GIS vibration signal.
Further, GIS vibration signal is decomposed as multiple modal components IMF using set empirical mode decomposition algorithm,
Calculate the calculation formula of each IMF energy are as follows:
In formula: RjIndicate the summation of energy in each modal components IMF, nIMFIt indicates to be wrapped in n-th of modal components IMF
The total amount of data contained,Indicate the energy value of each GIS vibration signal data point；
It is calculated to simplify, its energy feature, above-mentioned each IMF of calculating is characterized using 2 norms of each modal components IMF
The calculation formula of energy can be reduced to following calculation formula:
In formula: vnFor each modal components IMF energy.
Further, data are trained using depth confidence network (DBN) in step 4, by multiple unsupervised
Being limited Boltzmann machine (RBM) and one has deep neural network made of counterpropagation network (BP) stacking of supervision, passes through
The two stage training of depth confidence network model is respectively to high-rise unsupervised pre-training and by high-rise by low layer to low layer
Have supervision finely tune.
Wherein: the first stage is to train each limited Boltzmann machine RBM unsupervisedly using greedy algorithm, works as lower layer
After the completion of RBM training, the input as upper layer RBM is output it, successively successively training, to learn the feature of higher not
The disconnected training parameter for updating every layer, since greedy algorithm cannot be such that the learning parameter between each layer is optimal, therefore needs to carry out
The fine tuning of second stage；Second stage takes the mode training the last layer BP network of supervision, the mistake that the first stage is generated
Poor reverse transfer finely tunes the parameter between RBM layers each to each layer of following RBM, and according to transmission result, makes entire depth
The parameter of confidence network model is optimal.By the adjustment of two stages learning parameter, the input feature vector of data is just abstracted into
The feature of higher order, to obtain better classifying quality.
The present invention has the advantage that and the utility model has the advantages that
1, it is odd to extract GIS vibration signal peak-to-peak value, average value, degree of bias principal component frequency, 100Hz accounting, 50Hz by the present invention
Accounting is more than the energy of 90% preceding 7 IMF components as vibration signal characteristics in secondary frequency multiplication accounting and original energy
Amount, contains time domain, frequency domain and the energy feature of vibration signal, the characteristic of reflection GIS vibration signal that can be more complete；
2, present invention employs depth confidence networks (DBN), are divided by machine learning principle GIS vibration signal
Class has higher accuracy rate and faster convergence rate；
3, the present invention optimizes compound characteristics vector using Principal Component Analysis, not only remains compound characteristics vector
Raw information, and reduce feature vector dimension, improve the working efficiency of classifier, effectively improve the knowledge of GIS state
Other accuracy and speed.
Detailed description of the invention
Attached drawing described herein is used to provide to further understand the embodiment of the present invention, constitutes one of the application
Point, do not constitute the restriction to the embodiment of the present invention.In the accompanying drawings:
Fig. 1 is flow diagram of the invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below with reference to embodiment and attached drawing, to this
Invention is described in further detail, and exemplary embodiment of the invention and its explanation for explaining only the invention, are not made
For limitation of the invention.
Embodiment
As shown in Figure 1, a kind of GIS state identification method based on vibration signal Principal Component Analysis, which comprises
Step 1: vibration acceleration sensor being installed on GIS, acquisition multiple groups GIS normally and under malfunction vibrates letter
Number；
Step 2: GIS vibration signal collected to step 1 is handled, and extracts GIS vibration signal time domain, frequency domain respectively
And energy feature, and construct GIS vibration signal compound characteristics vector；
Step 2.1: extracting GIS vibration signal temporal signatures, including GIS vibration signal peak-to-peak value, average value, the degree of bias and high and steep
Degree；
(1) peak-to-peak value
Peak-to-peak value can characterize GIS oscillation intensity, calculation formula are as follows:
xpp=xmax-xmin (1)
In formula: xmaxIndicate GIS vibration signal maximum value, xminIndicate GIS vibration signal minimum value, xppIndicate GIS vibration
Signal strength.
(2) average value
Average value can characterize the offset of relative equilibrium position, calculation formula are as follows:
In formula: inIndicate that GIS vibration signal average value, n indicate GIS vibration signal data total amount, xiIndicate each data
Point.
(3) degree of bias
The degree of bias can characterize the degree of asymmetry of GIS vibration signal relative mean values, calculation formula are as follows:
In formula: Skew (x) indicates the signal degree of bias, and σ indicates data variance,Indicate that GIS vibration signal average value, n indicate
GIS vibration signal data total amount, xiIndicate each data point.
(4) kurtosis
Kurtosis is to describe the numerical statistic amount of signal waveform spike degree, can be with the distribution character of characterize data, calculation formula
Are as follows:
In formula: K indicates GIS vibration signal kurtosis, and E (t) expression seeks mathematic expectaion to data,Indicate GIS vibration signal
Average value, σ indicate data variance.
Peak-to-peak value, average value, the degree of bias and the kurtosis of collected GIS vibration signal are calculated according to above-mentioned each formula.
Step 2.2: extracting GIS vibration signal temporal signatures, including principal component frequency, 100Hz accounting, 50Hz odd times frequency multiplication
Accounting；
(1) principal component frequency
Principal component frequency is frequency at amplitude maximum.GIS operate normally when, vibration concentrate at 100Hz substantially, it is main at
Crossover rate is 100Hz.When GIS breaks down, other frequency contents may substantially increase and then 100Hz is replaced to become master
Component frequency.
(2) 100Hz accounting
When GIS is operated normally, fundamental frequency 100Hz, other frequency component amplitudes almost be can be ignored.When faulty
When, the amplitude of other frequencys increases accordingly, and 100Hz accounting can change.
(3) 50Hz odd times frequency multiplication accounting
When GIS is operated normally, 50Hz odd times harmonic is not present in vibration signals spectrograph, when faulty generation,
It will appear respective component, corresponding change will also occur for accounting.
Step 2.3: being decomposed GIS vibration signal for multiple modal components IMF, meter using set empirical mode decomposition algorithm
Calculate the calculation formula of each IMF energy are as follows:
In formula: RjIndicate the summation of energy in each modal components IMF, nIMFIt indicates to be wrapped in n-th of modal components IMF
The total amount of data contained,Indicate the energy value of each GIS vibration signal data point；
It is calculated to simplify, its energy feature, above-mentioned each IMF of calculating is characterized using 2 norms of each modal components IMF
The calculation formula of energy can be reduced to following calculation formula:
Formula (6) indicates on the basis of formula (5), seeks 2 norms to the energy of each IMF component.
Based on experience value, taking in original energy accounting is more than the energy conduct of 90% preceding 7 modal components IMF
GIS vibration signal energy feature.
Step 2.4: by above-mentioned 14 characteristic quantities (14 dimensional feature amounts are as follows: GIS vibration signal peak-to-peak value, average value, the degree of bias and
Kurtosis, principal component frequency, 100Hz accounting, 50Hz odd times frequency multiplication accounting, accounting is more than first 7 of 90% in original energy
The energy of modal components IMF, and the number of characteristic quantity is successively carried out in the order described above) successively constitute vibration signal compound characteristics
Vector α=(k1,k2... k14)T, in which: α is GIS vibration signal compound characteristics vector, k1For the first dimensional feature amount peak-to-peak value,
k2For the second dimensional feature amount average value, other and so on, subscript T is (k1,k2... k14) transposition.
Step 3: place is optimized to the GIS vibration signal compound characteristics vector that step 2 constructs using Principal Component Analysis
Reason, compression dimensionality reduction obtain GIS vibration signal principal component feature vector；
Step 3 specifically includes:
Step 3.1: being by the matrix that surveyed m group GIS vibration signal constitutes m × 14
X=(α1,α2... αm)T=(x1,x2... x14)
Step 3.2: according to the covariance matrix C of following formula calculating matrix X:
In formula: cov (x, y) indicates the covariance of two groups of data；
Step 3.3: calculating the eigenvalue λ of covariance matrix Ci(i=1,2......14) with corresponding eigenvectors matrix
E arranges obtained characteristic value in descending order, and will respectively arrange rearrange in eigenvectors matrix E in this order, obtains transition square
Battle array T, takes each column vector in T to be characterized the factor；
Step 3.4: calculating each characterization factor contribution rate, based on experience value, choose wherein the sum of contribution rate is more than 95% 2
A characterization factor forms transformation matrix, each characterization factor contribution rate calculation formula are as follows:
In formula: KrIndicate r-th of characterization factor contribution rate, λrIndicate the corresponding characteristic value of r-th of characterization factor, λjIt indicates
The corresponding characteristic value of j-th of characterization factor,It indicates to λjFrom λ1To λmSummation.
Step 3.5: the eigenmatrix Y of m × 2 being obtained by matrix operation Y=X × U, each column of matrix Y are one group
The principal component feature vector of GIS vibration signal.
Step 4: the GIS vibration signal principal component feature vector obtained according to step 3, for constructing depth confidence network mould
Type, using the GIS vibration signal principal component feature vector under multiple groups known state as the training of depth confidence Network Recognition model
Sample obtains GIS vibration signal principal component feature vector and GIS state by the two stage training of depth confidence network model
Between correspondence decision function；
Depth confidence network used in it is trained data, by multiple unsupervised limited Boltzmann machines
(RBM) and one has deep neural network made of counterpropagation network (BP) stacking of supervision, passes through depth confidence network mould
The two stage training of type is respectively to have supervision to finely tune to low layer by low layer to high-rise unsupervised pre-training and by high-rise.
First stage is to train each limited Boltzmann machine RBM unsupervisedly using greedy algorithm, when lower layer RBM is instructed
After the completion of white silk, output it the input as upper layer RBM, successively successively training, thus learn higher feature and constantly more
New every layer of training parameter, since greedy algorithm cannot be such that the learning parameter between each layer is optimal, therefore needs to carry out second
The fine tuning in stage；
Second stage takes the mode training the last layer BP network of supervision, and the error that the first stage generates reversely is passed
Each layer of following RBM is transported to, and the parameter between RBM layers each is finely tuned according to transmission result, makes entire depth confidence network
The parameter of model is optimal.
By the adjustment of two stages learning parameter, the input feature vector of data is just abstracted into the feature of higher order, thus
To better classifying quality.
Step 5: GIS vibration signal of the acquisition to state recognition carries out the calculating and analysis of principal component feature vector, uses
Corresponding decision function between GIS vibration signal principal component feature vector and GIS state carries out Classification and Identification to it, identifies this
GIS state.
Working principle is: based on the existing GIS state recognition based on vibration signal in the selection of different characteristic amount, knot
The message complementary sense relationship between different characteristic amount is closed, more reliable judgement is made to GIS state, but the increase of characteristic quantity can add
The work load of re-computation machine, reduces calculating speed and the problem of precision, it is proposed by the present invention it is a kind of be based on vibration signal it is main at
The GIS state identification method of point analytic approach, firstly, be extracted GIS vibration signal peak-to-peak value, average value, degree of bias principal component frequency,
Accounting is more than the energy of 90% preceding 7 IMF components in 100Hz accounting, 50Hz odd times frequency multiplication accounting and original energy
As vibr ation signals, construct GIS vibration signal compound characteristics vector, this 14 dimensional feature amount contain vibration signal when
Domain, frequency domain and energy feature, the characteristic of reflection GIS vibration signal that can be more complete；Secondly, using Principal Component Analysis pair
Compound characteristics vector optimizes, and not only remains the raw information of compound characteristics vector, but also reduces feature vector dimension,
The working efficiency for improving classifier in this way effectively improves the accuracy and speed of GIS state recognition；Then, depth is used
Confidence network classifies to GIS vibration signal by machine learning principle, there is higher accuracy rate and faster convergence speed
Degree；Finally, GIS vibration signal of the acquisition to state recognition, carries out the calculating and analysis of principal component feature vector, shakes with GIS
Corresponding decision function between dynamic signal principal component feature vector and GIS state carries out Classification and Identification to it, identifies the GIS shape
State, can high efficiency, accurately realize the GIS state judgement of comprehensive various features amount.
Above-described specific embodiment has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that being not intended to limit the present invention the foregoing is merely a specific embodiment of the invention
Protection scope, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all include
Within protection scope of the present invention.
Claims (6)
1. a kind of GIS state identification method based on vibration signal Principal Component Analysis, it is characterised in that: the described method includes:
Step 1: acquiring multiple groups GIS by the vibration acceleration sensor installed on GIS and normally and under malfunction vibrate
Signal；
Step 2: collected GIS vibration signal being handled, it is special to extract GIS vibration signal time domain, frequency domain and energy respectively
Sign, and construct GIS vibration signal compound characteristics vector；
Step 3: processing being optimized to GIS vibration signal compound characteristics vector using Principal Component Analysis, obtains GIS vibration letter
Number principal component feature vector；
Step 4: building depth confidence network model makees the GIS vibration signal principal component feature vector under multiple groups known state
GIS vibration signal is obtained by the two stage training of depth confidence network model for the training sample of depth confidence network model
Corresponding decision function between principal component feature vector and GIS state；
Step 5: GIS vibration signal of the acquisition to state recognition carries out the calculating and analysis of principal component feature vector, with GIS
Corresponding decision function between vibration signal principal component feature vector and GIS state carries out Classification and Identification to it, identifies the GIS
State.
2. a kind of GIS state identification method based on vibration signal Principal Component Analysis according to claim 1, feature
Be: the GIS vibration signal temporal signatures extracted in step 2 include GIS vibration signal peak-to-peak value, average value, the degree of bias and kurtosis；
Frequency domain character includes principal component frequency, 100Hz accounting, 50Hz odd times frequency multiplication accounting；It will using set empirical mode decomposition algorithm
It is multiple modal components IMF that GIS vibration signal, which decomposes, calculates each IMF energy, taking the accounting in original energy is more than 90%
Preceding 7 modal components IMF energy as GIS vibration signal energy feature.
3. a kind of GIS state identification method based on vibration signal Principal Component Analysis according to claim 2, feature
It is: constructs GIS vibration signal compound characteristics vector α=(k according to the 14 dimensional feature amounts that GIS vibration signal extracts1,k2,
...k14)T, wherein α is GIS vibration signal compound characteristics vector, k1For the first dimensional feature amount, k2For the second dimensional feature amount, successively
Go down k14For the tenth four-dimensional characteristic quantity, subscript T is (k1,k2... k14) transposition.
4. a kind of GIS state identification method based on vibration signal Principal Component Analysis according to claim 3, feature
Be: the step 3 specifically includes:
Step 3.1: surveyed m group GIS vibration signal is constituted to the matrix of m × 14
X=(α1,α2... αm)T=(x1,x2... x14)
Step 3.2: according to the covariance matrix C of following formula calculating matrix X:
In formula: cov (x, y) indicates the covariance of two groups of data；
Step 3.3: calculating the eigenvalue λ of covariance matrix Ci(i=1,2......14) and corresponding eigenvectors matrix E, will
Obtained characteristic value arranges in descending order, and will respectively arrange rearrange in eigenvectors matrix E in this order, obtains transition matrix T,
Each column vector in T is taken to be characterized the factor；
Step 3.4: calculating each characterization factor contribution rate, choose 2 characterization factors composition that wherein the sum of contribution rate is more than 95%
Transformation matrix, each characterization factor contribution rate calculation formula are as follows:
In formula: KrIndicate r-th of characterization factor contribution rate, λrIndicate the corresponding characteristic value of r-th of characterization factor, λjIt indicates j-th
The corresponding characteristic value of characterization factor,It indicates to λjFrom λ1To λmSummation；
Step 3.5: the eigenmatrix Y of m × 2 being obtained by matrix operation Y=X × U, each column of matrix Y are one group of GIS vibration
The principal component feature vector of dynamic signal.
5. a kind of GIS state identification method based on vibration signal Principal Component Analysis according to claim 2, feature
It is: GIS vibration signal is decomposed as multiple modal components IMF using set empirical mode decomposition algorithm, calculates each IMF energy
Calculation formula are as follows:
In formula: RjIndicate the summation of energy in each modal components IMF, nIMFIndicate number included in n-th of modal components IMF
According to total amount,Indicate the energy value of each GIS vibration signal data point；
It is calculated to simplify, its energy feature, above-mentioned each IMF energy of calculating is characterized using 2 norms of each modal components IMF
Calculation formula can be reduced to following calculation formula:
6. a kind of GIS state identification method based on vibration signal Principal Component Analysis according to claim 1, feature
It is: by the two stage training of depth confidence network model respectively by low layer to high-rise unsupervised pre-training in step 4
And there is supervision to finely tune to low layer by high-rise, in which: the first stage be trained unsupervisedly using greedy algorithm each by
Boltzmann machine RBM is limited, after the completion of lower layer RBM training, outputs it the input as upper layer RBM, successively successively training, from
And learns the feature of higher and constantly update every layer of training parameter；Second stage takes the mode of supervision to train last
Layer BP network, the Backward error propagation that the first stage is generated to each layer of following RBM, and it is each according to transmission result fine tuning
Parameter between RBM layers is optimal the parameter of entire depth confidence network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910350848.XA CN110082082B (en) | 2019-04-28 | 2019-04-28 | GIS state identification method based on vibration signal principal component analysis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910350848.XA CN110082082B (en) | 2019-04-28 | 2019-04-28 | GIS state identification method based on vibration signal principal component analysis method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110082082A true CN110082082A (en) | 2019-08-02 |
| CN110082082B CN110082082B (en) | 2020-11-17 |
Family
ID=67417339
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910350848.XA Active CN110082082B (en) | 2019-04-28 | 2019-04-28 | GIS state identification method based on vibration signal principal component analysis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110082082B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111486043A (en) * | 2020-04-24 | 2020-08-04 | 华能四川水电有限公司 | Lower rack fault diagnosis method based on hydro-turbo generator set runout data |
| CN112633333A (en) * | 2020-12-11 | 2021-04-09 | 广州致新电力科技有限公司 | Method for identifying partial discharge defects |
| CN113191437A (en) * | 2021-05-07 | 2021-07-30 | 国网山西省电力公司电力科学研究院 | Transformer mechanical fault detection method based on vibration signal composite eigenvector |
| CN113378031A (en) * | 2021-06-24 | 2021-09-10 | 深圳海域信息技术有限公司 | WHOIS query method and system based on global network distributed processing |
| CN114279707A (en) * | 2021-12-17 | 2022-04-05 | 哈尔滨工业大学 | Large-scale rotating equipment spindle state feature extraction method based on multi-domain analysis and principal component analysis |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108663202A (en) * | 2018-05-03 | 2018-10-16 | 国家电网公司 | GIS mechanical failure diagnostic methods based on chaos cuckoo algorithm and system |
| CN109374270A (en) * | 2018-09-19 | 2019-02-22 | 国网甘肃省电力公司电力科学研究院 | A kind of analysis of GIS abnormal vibrations and mechanical fault diagnosis device and method |
| CN109447048A (en) * | 2018-12-25 | 2019-03-08 | 苏州闪驰数控系统集成有限公司 | A kind of artificial intelligence early warning system |
| CN109597342A (en) * | 2019-01-16 | 2019-04-09 | 郑州轻工业学院 | A kind of the sand dredger monitoring device and method of dynamic group net INTELLIGENT IDENTIFICATION |
| CN109635428A (en) * | 2018-12-11 | 2019-04-16 | 红相股份有限公司 | A kind of GIS mechanical failure diagnostic method based on the analysis of machine performance signal |
-
2019
- 2019-04-28 CN CN201910350848.XA patent/CN110082082B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108663202A (en) * | 2018-05-03 | 2018-10-16 | 国家电网公司 | GIS mechanical failure diagnostic methods based on chaos cuckoo algorithm and system |
| CN109374270A (en) * | 2018-09-19 | 2019-02-22 | 国网甘肃省电力公司电力科学研究院 | A kind of analysis of GIS abnormal vibrations and mechanical fault diagnosis device and method |
| CN109635428A (en) * | 2018-12-11 | 2019-04-16 | 红相股份有限公司 | A kind of GIS mechanical failure diagnostic method based on the analysis of machine performance signal |
| CN109447048A (en) * | 2018-12-25 | 2019-03-08 | 苏州闪驰数控系统集成有限公司 | A kind of artificial intelligence early warning system |
| CN109597342A (en) * | 2019-01-16 | 2019-04-09 | 郑州轻工业学院 | A kind of the sand dredger monitoring device and method of dynamic group net INTELLIGENT IDENTIFICATION |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111486043A (en) * | 2020-04-24 | 2020-08-04 | 华能四川水电有限公司 | Lower rack fault diagnosis method based on hydro-turbo generator set runout data |
| CN112633333A (en) * | 2020-12-11 | 2021-04-09 | 广州致新电力科技有限公司 | Method for identifying partial discharge defects |
| CN113191437A (en) * | 2021-05-07 | 2021-07-30 | 国网山西省电力公司电力科学研究院 | Transformer mechanical fault detection method based on vibration signal composite eigenvector |
| CN113378031A (en) * | 2021-06-24 | 2021-09-10 | 深圳海域信息技术有限公司 | WHOIS query method and system based on global network distributed processing |
| CN113378031B (en) * | 2021-06-24 | 2022-06-28 | 深圳海域信息技术有限公司 | WHOIS query method and system based on global network distributed processing |
| CN114279707A (en) * | 2021-12-17 | 2022-04-05 | 哈尔滨工业大学 | Large-scale rotating equipment spindle state feature extraction method based on multi-domain analysis and principal component analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110082082B (en) | 2020-11-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110082082A (en) | A kind of GIS state identification method based on vibration signal Principal Component Analysis | |
| Cui et al. | Enhance high impedance fault detection and location accuracy via $\mu $-PMUs | |
| Li et al. | Real-time faulted line localization and PMU placement in power systems through convolutional neural networks | |
| CN106682303B (en) | A kind of three-level inverter method for diagnosing faults based on empirical mode decomposition and decision tree RVM | |
| Abdel-Galil et al. | Power quality disturbance classification using the inductive inference approach | |
| Song et al. | Multiple event detection and recognition for large-scale power systems through cluster-based sparse coding | |
| CN105447502A (en) | Transient power disturbance identification method based on S conversion and improved SVM algorithm | |
| CN109633368A (en) | The method of duration power quality disturbances containing distributed power distribution network based on VMD and DFA | |
| CN109033612A (en) | A kind of Diagnosis Method of Transformer Faults based on vibration noise and BP neural network | |
| CN103136587A (en) | Power distribution network operating state classification recognition method based on support vector machine | |
| Li et al. | A power system disturbance classification method robust to PMU data quality issues | |
| Qu et al. | False data injection attack detection in power systems based on cyber-physical attack genes | |
| CN109858503A (en) | The traction converter failure diagnostic method of decision tree is promoted based on gradient | |
| CN109974835A (en) | A kind of vibration detection identification and space-time localization method and system based on fiber-optic signal feature | |
| Beg et al. | Cyber-physical anomaly detection in microgrids using time-frequency logic formalism | |
| Hong et al. | Towards accurate and efficient classification of power system contingencies and cyber-attacks using recurrent neural networks | |
| Kezunovic | Translational knowledge: From collecting data to making decisions in a smart grid | |
| Wang et al. | Gas-insulated switchgear insulation defect diagnosis via a novel domain adaptive graph convolutional network | |
| Zhang et al. | Online power system dynamic security assessment with incomplete PMU measurements: A robust white‐box model | |
| CN112327189A (en) | KNN algorithm-based energy storage battery health state comprehensive judgment method | |
| Gu et al. | Doubly-fed deep learning method for bad data identification in linear state estimation | |
| Li et al. | Fault diagnosis of ZD6 turnout system based on wavelet transform and GAPSO-FCM | |
| CN111521905A (en) | Power grid fault diagnosis method based on intelligent fusion of multi-dimensional data fault degrees | |
| CN116318878A (en) | Assessment method for security risk of power information network | |
| Han et al. | Study on earthquake prediction model based on traffic disaster 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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200617 Address after: 637000 Nanchong city in Sichuan province Fu Road No. 228 Applicant after: NANCHONG POWER SUPPLY COMPANY, STATE GRID SICHUAN ELECTRIC POWER Co.,Ltd. Address before: 637000 Nanchong city in Sichuan province Fu Road No. 228 Applicant before: NANCHONG POWER SUPPLY COMPANY, STATE GRID SICHUAN ELECTRIC POWER Co.,Ltd. Applicant before: SICHUAN University |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |