CN106324405A - Transformer fault diagnosis method based on improved principal component analysis - Google Patents
Transformer fault diagnosis method based on improved principal component analysis Download PDFInfo
- Publication number
- CN106324405A CN106324405A CN201610807639.XA CN201610807639A CN106324405A CN 106324405 A CN106324405 A CN 106324405A CN 201610807639 A CN201610807639 A CN 201610807639A CN 106324405 A CN106324405 A CN 106324405A
- Authority
- CN
- China
- Prior art keywords
- sample
- main constituent
- principal component
- transformer
- sigma
- 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
-
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Abstract
The invention discloses a transformer fault diagnosis method based on improved principal component analysis, which belongs to the technical field of transformer fault diagnosis. The method adopts the sum of absolute values of sample indexes to carry out standardized treatment on sample index values, and thus, the difference of each sample index value in magnitude can be eliminated, and the information difference features between samples can also be kept; and sample principal components are selected according to cumulative contribution rates of the principal components, Euclidean distances among the sample principal components are clustered, and a transformer fault type is judged. The method of the invention can effectively improve the accuracy for diagnosing inner hidden fault of the transformer.
Description
Technical field
The present invention relates to a kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis, belong to transformer fault and examine
Disconnected technical field.
Background technology
Power transformer is the visual plant of power system, and the reliability of himself has become as the base of power grid operation
Plinth.It is diagnosed to be the Hidden fault before transformer fault especially peril occurs, not only direct relation accurately and in time
To electric energy conveying reliability and the safety of Operation of Electric Systems, it is also possible to avoid its because of fault spread, cause shutdown with
And to system shock, burn out equipment and produce tremendous economic loss.Therefore, the accurate of latent transformer fault diagnosis is improved
Rate, has very important Research Significance.
Current Gases Dissolved in Transformer Oil analysis (DGA) method has become power system and has judged power transformer interior fault character
Main method.The most commonly used in DGA method is three ratio in judgement rules.Three ratio in judgement rule forms are simple, clear, make
With convenient, but application be frequently found the shortcomings, even some ratio such as understaffed code, encoded boundary be the most absolute at the scene and also looked for
Less than corresponding fault type, therefore the accuracy rate of its fault diagnosis need to improve.Along with the development of artificial intelligence technology, adopt
Grinding of transformer fault diagnosis is carried out with neutral net, fuzzy clustering, gray theory and evidential reasoning and other intelligent methods
Study carefully and become the theme that comparison is popular.But, although neural network possesses self-learning capability, but bigger to the dependency of sample;
Fuzzy theory is easier to ignore the dependency of sample space;Gray theory is also easily affected by artificial subjective factor;Evidential reasoning
Algorithm is the most complicated, and can have a certain degree of uncertainty and subjectivity in evaluation procedure.So, these artificial intelligence at present
Can method maturity in transformer fault diagnosis is applied need to improve.
Inside transformer Hidden fault comprises the factor of various complexity, and these factors many times comprise repetition
Information.These information are the most tediously long, are just less susceptible to correctly judge the law of development of fault.But grind in major part fault
During studying carefully, the characteristic information of fault always has certain dependency.Therefore, being correlated with between characteristic information variable how is eliminated
Property so that breakdown judge is more accurate, become an interesting problem.1933, the principal component analysis that Hotelling proposes
PCA (Principal Component Analysis) method is one of effective way realizing this purpose.The present invention based on
Transformator DGA is theoretical, proposes to use follow-on PCA directly to diagnose power transformer interior fault.By master
The contribution rate of accumulative total threshold value of composition, chooses main constituent number and corresponding characteristic vector, sets up the main one-tenth of raw sample data
Sub-model so that comprise the main constituent that information is the most overlapping and orthogonal, has the maximum comprehensive original sample variable information of energy
Ability.So it is easier to catch the principal contradiction of things, makes problem be simplified, failure classes severe particularly with running environment
For there is the Transformer Faults Analysis of the biggest ambiguity between type and malfunction characteristic quantity, put these nature of troubles in order and close
Cording has obvious advantage.
Summary of the invention
It is an object of the invention to deficiency of the prior art, it is provided that a kind of based on the transformator event improving principal component analysis
Barrier diagnostic method.
For solving above-mentioned technical problem, the technical solution used in the present invention is:
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis, comprises the steps:
1) transformer fault state is analyzed;
2) sample initial matrix standardization is carried out;
3) set up correlation matrix, calculate eigenvalue and characteristic vector;
4) calculate principal component contributor rate and contribution rate of accumulative total, choose sample main constituent;
5) distance between sample to be tested and state feature samples main constituent is calculated, it is judged that sample to be tested state belongs to.
Aforesaid analysis transformer fault state refers to, five kinds of crucial hydrocarbon gas H in being tested by oil chromatography2、CH4、
C2H2、C2H4、C2H6As characteristic gas, transformer fault is divided into normally, shelf depreciation, low-yield electric discharge, high-energy discharge,
Hot stall t 300 DEG C, 300 DEG C of t of hot stall 700 DEG C, hot stall t 700 DEG C and electric discharge and eight kinds of fault shapes of overheated mixed fault
State.
Aforesaid carrying out sample initial matrix standardization, particularly as follows: be provided with n sample, each sample has p item to refer to
Mark, obtains sample initial matrix X:
Wherein, xij, i=1,2 ... n, j=1,2 ... p, represent the jth item index of i-th sample,
xi=[xi1 xi2 … xip], i=1,2 ... n;
Use following formula to be standardized processing, obtain sample standardization data matrix,
Wherein,For xijStandardized data.
Aforesaid correlation matrix R is:
Wherein, rij, i, j=1,2 ..., p is the correlation coefficient of sample standardization data matrix;rij=rji;For xiMiddle unit
The average of element;For xjThe average of middle element;
Use Jacobi method | the λ I-R |=0 that solves characteristic equation, calculate the eigenvalue of R, and by eigenvalue order by size
Arrangement:
λ1≥λ2≥…≥λp>=0, λi, i=1,2 ... p is the eigenvalue of R;
Try to achieve the characteristic vector corresponding with eigenvalue: a simultaneously1、a2、…、ai、…、ap,
ai=[a1i a2i … api]T, i=1,2 ..., p.
Aforesaid calculating principal component contributor rate TiAnd contribution rate of accumulative total MC, formula is as follows:
Described selection sample main constituent refers to, when contribution rate of accumulative total MCWhen meeting following formula, front C main constituent is selected
Sample main constituent,
Wherein, 0 < MC≤ 1, threshold epsilon > 0, and be positive minimum real number.
Aforesaid threshold epsilon=0.005.
Aforesaid step 5) in, the computing formula of sample main constituent is:
FiFor sample xiMain constituent,
The main constituent F of state feature samples is obtained according to formula (7)MMain constituent with sample to be tested
FM=(F1 F2 … Fm)
Wherein, f is the sequence number of transformer fault classification, and m is selected sample number of principal components,
Choose state feature samples main constituent FMWith sample to be tested main constituentBetween Euclidean distance df, as sample to be tested
And the overall similarity between state feature samples,
Finding out the Euclidean distance of minimum, sample to be tested just belongs to the state feature samples institute corresponding to minimum Eustachian distance
The class malfunction belonged to.
Beneficial effect: the present invention uses variable absolute value sum to be standardized variable processing, and has both eliminated each index
Numerical value difference on the order of magnitude, maintains again the information gap feature between each variable;The variable of sample space contains DGA
5 kinds of characteristic gas and code of direct ratio, calculate the contribution rate of accumulative total threshold value of main constituent, choose pivot number, form main constituent
Model, simplifies the information characteristics that transformator is lengthy and tedious;Analyze the distance between sample main constituent, it is judged that latent transformer fault
State belongs to, and has obvious advantage for the Hidden fault character relation putting transformator in order complicated.
The inventive method can relatively accurately reflect the running status of transformator, has certain effectiveness, and necessarily
In degree, the noise in sample being had stronger denoising and immunocompetence, it is by calculating the main constituent of sample, it is achieved right
The feature extraction of complex information, eliminates the correlation properties between variable, thus improves the precision of cluster.
Accompanying drawing explanation
Fig. 1 is based on the Diagnosis Method of Transformer Faults flow chart improving principal component analysis;
Fig. 2 is that main constituent selects flow chart.
Detailed description of the invention
With embodiment, the present invention is further described below in conjunction with the accompanying drawings.
A kind of based on improvement principal component analysis the Diagnosis Method of Transformer Faults of the present invention, as it is shown in figure 1, include as follows
Step:
Step one: analyze transformer fault state;
Step 2: sample initial matrix standardization;
Step 3: set up correlation matrix, calculates eigenvalue and characteristic vector;
Step 4: calculate principal component contributor rate and contribution rate of accumulative total, choose sample main constituent;
Step 5: calculate the distance between sample to be tested and state feature samples main constituent, it is judged that sample to be tested state is returned
Belong to.
Specific implementation process is as follows:
High dimensional data is mainly projected to an orthogonal projector space of new characteristic vector by principal component analysis, real
The now feature extraction to complex process data, in order to eliminate the associate feature between variable, so that primary process specificity analysis
Complexity significantly simplified.Principal component model has given up part residual error, retains simultaneously and embodies the main of data characteristics
Gene, thus reach extraction system information, the purpose of scavenging system interference.This is for processing and analyze information redundancy, feature letter
The power transformer interior fault of number aliasing is highly effective.The present invention proposes a kind of transformer fault based on improvement principal component analysis and examines
Disconnected method.Below with 30 groups through the transformator DGA data instance of pendant-core examination its fault type known, provide its fault diagnosis side
Method.
1) transformer fault state is set up
Transformator DGA is present analysis power transformer interior fault most efficient method.Inventive algorithm is with DGA data as base
Plinth, selects H2、CH4、C2H2、C2H4、C2H6The content of these 5 kinds of gases, as characteristic quantity, diagnoses transformer insulated fault.Substantial amounts of
Research proves, the fault diagnosis of oil-filled electric equipment can not only depend on the constituent content of oil dissolved gas, should also depend on
The relative amount of gas.Based on the three-ratio method on the basis of thermokinetics, the relative amount of gas can be reflected, and in reality
It is widely accepted.So the present invention uses the coded combination of three-ratio method as input quantity, extract the state feature of sample.
" Gases Dissolved in Transformer Oil divides GB/T7252-2001 according to the existing promulgation of IEC60599 code and China
Analyse and judge directive/guide ", latent transformer malfunction is divided into 8 kinds of states (wherein comprising normal condition), each number of state indexes
It is followed successively by 1,2 ..., 8, it is shown in Table 1.
Table 1 transformer fault is classified
2) sample initial matrix standardization
Being provided with n sample, each sample has p item index (variable) X1、X2、…、Xp, obtain sample initial matrix X:
Wherein, xij, i=1,2 ... n, j=1,2 ... p, represent the jth item index of i-th sample,
xi=[xi1 xi2 … xip], i=1,2 ... n.
In order to eliminate impact and each index difference on the order of magnitude of dimension, need to be standardized X processing.Often
Rule standardisation process obtains divided by sample standard deviation generally by the difference of sample with sample average.But inventive algorithm
Sample initial matrix is due to containing dissimilar, varying number level, the data of different implication, so in course of standardization process
In, both to separate, and individually carries out.In order to retain the characteristic information of each index value in initial matrix to a greater degree, this
Invention does not use the standardized method of tradition PCA, but uses following formula to be standardized processing, and has both eliminated each
Index value difference on the order of magnitude, maintains again the information gap feature between each variable.
In formula,For xijStandardized data.
The present invention chooses 8 groups and has clear and definite conclusion through pendant-core examination, and malfunction type is single, multiplicity event does not occurs
The transformator DGA data being diagnosed before barrier, generation catastrophe failure are as state feature samples, as shown in table 2, wherein, and bag
Sample containing normal non-fault state;8th group of data contain electric discharge and overheating fault simultaneously, but its every kind trouble point (source) is all only
Have at 1;Gas content be finger pressure be 101.3kPa, temperature is when being 20 DEG C, the microlitre number of contained each gas component in every liter of oil,
Same afterwards.
Table 2 transformer state feature samples
Form sample initial matrix X according to table 2, see below formula.In matrix X, x9For sample to be tested, and x1、x2、…、x8For
State feature samples.
3) set up correlation matrix, calculate eigenvalue and characteristic vector,
Set up correlation matrix R, it may be assumed that
Wherein,
In formula, rij(i, j=1,2 ..., p) it is the correlation coefficient of sample standardization data matrix;rij=rji;For xiIn
The average of element;For xjThe average of middle element.Try to achieve characteristic vector by calculating correlation matrix R, and then try to achieve sample
Main constituent Fi, sample x can be deletediIndex little on result of calculation impact in index, simplifies and calculates, and therefore i value is 1,2 ...,
p。
Jacobi method is used to solve characteristic equation λ I-R |=0, calculate the eigenvalue of R: λ1≥λ2≥…≥λp>=0, and make
It sequentially arranges by size, λi, i=1,2 ... p is the eigenvalue of R.The most also can try to achieve the feature corresponding with eigenvalue to
Amount a1、a2、…、ai、…、ap。ai=[a1i a2i … api]TI=1,2 ..., p.
According to above formula, DGA data and code of direct ratio to X are standardized processing respectively, set up correlation matrix
R.Then Jacobi method is used to calculate eigenvalue λ and the eigenvectors matrix a of R.
λ=[0.2857 0.2111 0.0948 0.0184 0.0068 0.0030 0.0002 0]=[λ1 λ2 … λ8]
4) calculate principal component contributor rate and contribution rate of accumulative total, choose sample main constituent
According to main constituent contribution rate of accumulative total and threshold epsilon, choose sample main constituent.Calculate principal component contributor rate TiAnd front C
Contribution rate of accumulative total M of individual main constituentC。
Select main constituent primarily to distinguish principal character information and the secondary feature information that sample is comprised.Main special
Reference breath is reflected by main constituent subspace, available main constituent contribution rate of accumulative total MCRepresent, and secondary feature information spinner to comprise and makes an uproar
Sound, available 1-MCRepresent.When meeting following formula, it may be determined that front C main constituent, i.e. can main with reflected sample matrix X
Characteristic information.
In formula, 0 < MC≤1;Threshold epsilon > 0, and be positive minimum real number.
Main constituent contribution rate of accumulative total matrix M is calculated according to above formula.
M=[0.4608 0.8012 0.9541 0.9838 0.9950 0.9997 1.0000 1.0000]=[M1 M2
… M8]
5) distance between sample to be tested and state feature samples main constituent is calculated, it is judged that sample to be tested state belongs to
Sample xiMain constituent FiFor:
Can be obtained in sample space by above formula, the main constituent of state feature samples and sample to be tested is respectively
FM=(F1 F2 … Fm) and(f is the sequence number of transformer fault classification, and m is main
Composition number).Owing to Euclidean distance is modal distance metric, measurement be in hyperspace each point between distance,
And its effect in distance measure is extensively applied by every field, so the present invention utilizes above formula to calculate main constituent FMWith
Between Euclidean distance df, as the overall similarity between sample to be tested and state feature samples.Distance is the least, and both get over phase
Seemingly.
Find out the distance of minimum, so that it may judge which kind of malfunction sample to be tested belongs to and (include normal non-fault shape
State).
In actual applications, M is typically worked asCDuring > 0.995, the main constituent of selection can guarantee that enough precision.This
Bright take threshold epsilon=0.005.Work as MCWhen >=0.995, according to contribution rate of accumulative total matrix M, it may be determined that the number of its main constituent be 6 (because of
For M6=0.995).Owing to the number of its main constituent is 6, therefore only selected characteristic vector a1-a6.Obtain sample main constituent expression formula F
It is shown below.
The main constituent F of sample to be tested is calculated according to above formulaMWith other sample main constituentsDistance d:
D=[d1 d2 d3 d4 d5 d6 d7 d8]
=[1.4014 1.0098 0.7883 0.8382 1.4132 1.2094 1.0927 0.8159]
Understand, d3=0.7883 is minimum, i.e. sample to be tested x9Belong to the 3rd class malfunction, be low-yield putting as shown in Table 1
Electric fault, diagnostic result is consistent with actual phenomenon.And use David's triangulation method and IEC60599 three-ratio method to be diagnosed as respectively
, there is erroneous judgement in high-energy discharge and partial discharges fault.
In order to further illustrate accuracy and the reliability of the inventive method, choose the DGA of 30 groups of its malfunctions known
Data, as sample to be tested, carry out Analysis on Fault Diagnosis to it, are shown in Table 3, the fault type sequence number in table and the sequence number one in table 1
Cause, represent the fault type of each sample.
3 30 groups of transformer fault samples of table
Use the inventive method that this 30 example DGA data sample is carried out diagnostic analysis.It addition, in order to carry out diagnosis effect pair
Ratio, is respectively adopted David's triangulation method, IEC60599 three-ratio method, improvement three-ratio method, fuzzy C-mean algorithm (FCM) method in table 3
30 groups of transformer fault samples diagnose, and diagnostic result is as shown in table 4.In table 4, sample sequence number and number of state indexes and table 3
It is consistent.When the number of state indexes that every kind of diagnostic method is diagnosed to be is with when the number of state indexes of the 2nd row is consistent in table 4, illustrate that this is examined
Disconnected method diagnosis is accurate, otherwise diagnostic error.Last 2 row of table 4 be respectively every kind of diagnostic method Accurate Diagnosis total sample number with
And the accuracy rate of diagnosis.
As can be seen from Table 4: David's triangulation method, IEC60599 three-ratio method, improvement three-ratio method, FCM method and basis
Inventive method, the total sample number of Accurate Diagnosis is 22,21,10,21,26 respectively, and the accuracy rate of diagnosis of its correspondence is respectively
73.33%, 70.00%, 33.30%, 70.00%, 86.67%.It can thus be seen that the accuracy rate of the inventive method diagnosis is
Height, David's triangulation method, IEC60599 three-ratio method take second place, and improve three-ratio method minimum.This explanation the inventive method can be relatively
Adequately reflect the running status of transformator, there is certain effectiveness, and to a certain extent the noise in sample is had
Stronger denoising and immunocompetence, it is by calculating the main constituent of sample, it is achieved the feature extraction to complex information, eliminates and becomes
Correlation properties between amount, thus improve the precision of cluster.
The diagnostic result of 45 kinds of methods of table
It addition, diagnose by choosing different feature samples, find to use the inventive method, different feature samples
Little on sample evaluation result impact to be measured.Certainly, sample data accurately and accurately, more diagnosis can be produced impact.
The above is only the preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art
For Yuan, under the premise without departing from the principles of the invention, it is also possible to make some improvements and modifications, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (7)
1. a Diagnosis Method of Transformer Faults based on improvement principal component analysis, it is characterised in that comprise the steps:
1) transformer fault state is analyzed;
2) sample initial matrix standardization is carried out;
3) set up correlation matrix, calculate eigenvalue and characteristic vector;
4) calculate principal component contributor rate and contribution rate of accumulative total, choose sample main constituent;
5) distance between sample to be tested and state feature samples main constituent is calculated, it is judged that sample to be tested state belongs to.
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis the most according to claim 1, its feature exists
In: described analysis transformer fault state refers to, five kinds of crucial hydrocarbon gas H in being tested by oil chromatography2、CH4、C2H2、C2H4、
C2H6As characteristic gas, transformer fault is divided into normally, shelf depreciation, low-yield electric discharge, high-energy discharge, hot stall t
300 DEG C, 300 DEG C of t of hot stall 700 DEG C, hot stall t 700 DEG C and electric discharge and eight kinds of malfunctions of overheated mixed fault.
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis the most according to claim 1, its feature exists
In: described in carry out sample initial matrix standardization, particularly as follows: be provided with n sample, each sample has p item index, obtains sample
This initial matrix X:
Wherein, xij, i=1,2 ... n, j=1,2 ... p, represent the jth item index of i-th sample,
xi=[xi1 xi2 … xip], i=1,2 ... n;
Use following formula to be standardized processing, obtain sample standardization data matrix,
Wherein,For xijStandardized data.
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis the most according to claim 3, its feature exists
In: described correlation matrix R is:
Wherein, rij, i, j=1,2 ..., p is the correlation coefficient of sample standardization data matrix;rij=rji;For xiMiddle element
Average;For xjThe average of middle element;
Use Jacobi method | the λ I-R |=0 that solves characteristic equation, calculate the eigenvalue of R, and by eigenvalue order arrangement by size:
λ1≥λ2≥…≥λp>=0, λi, i=1,2 ... p is the eigenvalue of R;
Try to achieve the characteristic vector corresponding with eigenvalue: a simultaneously1、a2、…、ai、…、ap, ai=[a1i a2i … api]T, i=
1,2,…,p。
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis the most according to claim 4, its feature exists
In: described calculating principal component contributor rate TiAnd contribution rate of accumulative total MC, formula is as follows:
Described selection sample main constituent refers to, when contribution rate of accumulative total MCWhen meeting following formula, front C main constituent is selected sample
This main constituent,
Wherein, 0 < MC≤ 1, threshold epsilon > 0, and be positive minimum real number.
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis the most according to claim 5, its feature exists
In: described threshold epsilon=0.005.
A kind of Diagnosis Method of Transformer Faults based on improvement principal component analysis the most according to claim 4, its feature exists
In: step 5) in, the computing formula of sample main constituent is:
FiFor sample xiMain constituent,
The main constituent F of state feature samples is obtained according to formula (7)MMain constituent with sample to be tested
FM=(F1 F2 … Fm)
Wherein, f is the sequence number of transformer fault classification, and m is selected sample number of principal components,
Choose state feature samples main constituent FMWith sample to be tested main constituentBetween Euclidean distance df, as sample to be tested and shape
Overall similarity between state feature samples,
Finding out the Euclidean distance of minimum, sample to be tested just belongs to belonging to the state feature samples corresponding to minimum Eustachian distance
One class malfunction.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610807639.XA CN106324405A (en) | 2016-09-07 | 2016-09-07 | Transformer fault diagnosis method based on improved principal component analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610807639.XA CN106324405A (en) | 2016-09-07 | 2016-09-07 | Transformer fault diagnosis method based on improved principal component analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106324405A true CN106324405A (en) | 2017-01-11 |
Family
ID=57787622
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610807639.XA Pending CN106324405A (en) | 2016-09-07 | 2016-09-07 | Transformer fault diagnosis method based on improved principal component analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106324405A (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106950446A (en) * | 2017-03-22 | 2017-07-14 | 国网上海市电力公司 | The abnormal quick detection of power network based on PCA and localization method |
CN107228913A (en) * | 2017-06-09 | 2017-10-03 | 广西电网有限责任公司电力科学研究院 | A kind of condition diagnosing system of transformer fault type |
CN107677903A (en) * | 2017-09-14 | 2018-02-09 | 广西电网有限责任公司电力科学研究院 | A kind of clustering method of Transformer's Condition Monitoring data |
CN108388221A (en) * | 2018-03-21 | 2018-08-10 | 江南大学 | A kind of multilayered schema monitoring method of process flow industry process |
WO2018149084A1 (en) * | 2017-02-17 | 2018-08-23 | 平安科技(深圳)有限公司 | Method and device for identifying correlation variable, terminal, and storage medium |
CN108805415A (en) * | 2018-05-22 | 2018-11-13 | 国网江西省电力有限公司电力科学研究院 | The transformer body critical evaluation selecting index method excavated based on historical information |
CN109039280A (en) * | 2018-07-02 | 2018-12-18 | 福建农林大学 | Diagnosing failure of photovoltaic array method based on non-primary component data characteristics |
CN109164343A (en) * | 2018-08-30 | 2019-01-08 | 西华大学 | Quantify the Diagnosis Method of Transformer Faults with weighting KNN based on characteristic information |
CN109298099A (en) * | 2018-11-26 | 2019-02-01 | 国网河北省电力有限公司电力科学研究院 | Transformer oil chromatographic determination method, system and terminal device |
CN109459409A (en) * | 2017-09-06 | 2019-03-12 | 盐城工学院 | A kind of near-infrared exceptional spectrum recognition methods based on KNN |
CN110717552A (en) * | 2019-10-23 | 2020-01-21 | 智洋创新科技股份有限公司 | Method for determining visible mechanical continuous alarm of power transmission line channel |
CN112014692A (en) * | 2020-07-20 | 2020-12-01 | 国网安徽省电力有限公司电力科学研究院 | Partial discharge ultrahigh frequency signal blind source separation denoising method based on principal component analysis |
CN112147444A (en) * | 2020-09-25 | 2020-12-29 | 广东电网有限责任公司佛山供电局 | Power transformer working state monitoring method and system |
CN112801135A (en) * | 2020-12-31 | 2021-05-14 | 浙江浙能镇海发电有限责任公司 | Fault line selection method and device for power plant service power system based on characteristic quantity correlation |
CN114019282A (en) * | 2021-11-04 | 2022-02-08 | 云南电网有限责任公司电力科学研究院 | Transformer fault diagnosis method based on principal component analysis and random forest phase fusion |
CN114022114A (en) * | 2021-11-03 | 2022-02-08 | 广州智算信息技术有限公司 | Data governance platform based on telecommunication industry |
-
2016
- 2016-09-07 CN CN201610807639.XA patent/CN106324405A/en active Pending
Non-Patent Citations (1)
Title |
---|
杨廷方等: "基于改进型主成分分析的电力变压器潜伏性故障诊断", 《电力自动化设备》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018149084A1 (en) * | 2017-02-17 | 2018-08-23 | 平安科技(深圳)有限公司 | Method and device for identifying correlation variable, terminal, and storage medium |
CN106950446A (en) * | 2017-03-22 | 2017-07-14 | 国网上海市电力公司 | The abnormal quick detection of power network based on PCA and localization method |
CN107228913A (en) * | 2017-06-09 | 2017-10-03 | 广西电网有限责任公司电力科学研究院 | A kind of condition diagnosing system of transformer fault type |
CN107228913B (en) * | 2017-06-09 | 2023-08-08 | 广西电网有限责任公司电力科学研究院 | State diagnosis system for fault type of transformer |
CN109459409A (en) * | 2017-09-06 | 2019-03-12 | 盐城工学院 | A kind of near-infrared exceptional spectrum recognition methods based on KNN |
CN109459409B (en) * | 2017-09-06 | 2022-03-15 | 盐城工学院 | KNN-based near-infrared abnormal spectrum identification method |
CN107677903A (en) * | 2017-09-14 | 2018-02-09 | 广西电网有限责任公司电力科学研究院 | A kind of clustering method of Transformer's Condition Monitoring data |
CN108388221A (en) * | 2018-03-21 | 2018-08-10 | 江南大学 | A kind of multilayered schema monitoring method of process flow industry process |
CN108388221B (en) * | 2018-03-21 | 2019-06-21 | 江南大学 | A kind of multilayered schema monitoring method of process flow industry process |
CN108805415A (en) * | 2018-05-22 | 2018-11-13 | 国网江西省电力有限公司电力科学研究院 | The transformer body critical evaluation selecting index method excavated based on historical information |
CN109039280A (en) * | 2018-07-02 | 2018-12-18 | 福建农林大学 | Diagnosing failure of photovoltaic array method based on non-primary component data characteristics |
CN109164343B (en) * | 2018-08-30 | 2020-11-06 | 西华大学 | Transformer fault diagnosis method based on characteristic information quantization and weighted KNN |
CN109164343A (en) * | 2018-08-30 | 2019-01-08 | 西华大学 | Quantify the Diagnosis Method of Transformer Faults with weighting KNN based on characteristic information |
CN109298099A (en) * | 2018-11-26 | 2019-02-01 | 国网河北省电力有限公司电力科学研究院 | Transformer oil chromatographic determination method, system and terminal device |
CN110717552A (en) * | 2019-10-23 | 2020-01-21 | 智洋创新科技股份有限公司 | Method for determining visible mechanical continuous alarm of power transmission line channel |
CN112014692A (en) * | 2020-07-20 | 2020-12-01 | 国网安徽省电力有限公司电力科学研究院 | Partial discharge ultrahigh frequency signal blind source separation denoising method based on principal component analysis |
CN112147444A (en) * | 2020-09-25 | 2020-12-29 | 广东电网有限责任公司佛山供电局 | Power transformer working state monitoring method and system |
CN112801135A (en) * | 2020-12-31 | 2021-05-14 | 浙江浙能镇海发电有限责任公司 | Fault line selection method and device for power plant service power system based on characteristic quantity correlation |
CN112801135B (en) * | 2020-12-31 | 2023-04-18 | 浙江浙能镇海发电有限责任公司 | Fault line selection method and device for power plant service power system based on characteristic quantity correlation |
CN114022114A (en) * | 2021-11-03 | 2022-02-08 | 广州智算信息技术有限公司 | Data governance platform based on telecommunication industry |
CN114022114B (en) * | 2021-11-03 | 2022-07-15 | 广州智算信息技术有限公司 | Data management system and method based on telecommunication industry |
CN114019282A (en) * | 2021-11-04 | 2022-02-08 | 云南电网有限责任公司电力科学研究院 | Transformer fault diagnosis method based on principal component analysis and random forest phase fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106324405A (en) | Transformer fault diagnosis method based on improved principal component analysis | |
Hooper et al. | Evaluating model fit: a synthesis of the structural equation modelling literature | |
CN106021826B (en) | One kind is based on aero-engine complete machine method for predicting residual useful life under operating mode's switch and the matched variable working condition of similitude | |
CN109711663A (en) | Substation's oil-immersed transformer status assessment and modification method and system based on big data analysis | |
CN104750979B (en) | The synthesization risk priority number calculating method of Architecture-oriented | |
CN106169001B (en) | A kind of aero-engine complete machine method for predicting residual useful life based on gas circuit performance parameter Graphic Pattern Matching | |
CN106156485A (en) | Method for diagnosing fault of power transformer and device | |
CN106447210A (en) | Distribution network equipment health degree dynamic diagnosis method involving credibility evaluation | |
CN103207950A (en) | Intelligent transformer fault diagnostic method based on RBF (radial basis function) neural network | |
CN101701940A (en) | On-line transformer fault diagnosis method based on SVM and DGA | |
CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
CN108268905A (en) | A kind of Diagnosis Method of Transformer Faults and system based on support vector machines | |
CN109063734B (en) | Oil-immersed transformer fault state evaluation method combining multi-level local density clustering | |
CN107632590B (en) | A kind of bottom event sort method priority-based | |
CN110161343A (en) | A kind of non-intrusion type real-time dynamic monitoring method of intelligence train exterior power receiving device | |
CN115828140B (en) | Method, system and application for detecting fault by fusing neighborhood mutual information and random forest | |
CN107037306A (en) | Transformer fault dynamic early-warning method based on HMM | |
CN107340766A (en) | Power scheduling alarm signal text based on similarity sorts out and method for diagnosing faults | |
CN107276072A (en) | A kind of method of utilization steady state information qualitative assessment power system transient stability margin | |
Sun et al. | A novel life prediction method for railway safety relays using degradation parameters | |
CN113342784A (en) | Database design method for risk assessment of main transformer equipment of power grid | |
Bakar et al. | Improvement of DGA interpretation using scoring index method | |
CN116992362A (en) | Transformer fault characterization feature quantity screening method and device based on Xia Puli value | |
CN107122829B (en) | Method for diagnosing transformer fault by training neural network through virtual sample | |
Li et al. | MSTI: a new clustering validity index for hierarchical clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into 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: 20170111 |