CN104655948A - Novel multistage fault diagnosis method for power transformer - Google Patents

Novel multistage fault diagnosis method for power transformer Download PDF

Info

Publication number
CN104655948A
CN104655948A CN201310588855.6A CN201310588855A CN104655948A CN 104655948 A CN104655948 A CN 104655948A CN 201310588855 A CN201310588855 A CN 201310588855A CN 104655948 A CN104655948 A CN 104655948A
Authority
CN
China
Prior art keywords
fault
posterior probability
evidence
power transformer
diagnosis
Prior art date
Application number
CN201310588855.6A
Other languages
Chinese (zh)
Inventor
孙志
何菲
梅军
Original Assignee
国家电网公司
江苏省电力公司
江苏省电力公司泰州供电公司
东南大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 国家电网公司, 江苏省电力公司, 江苏省电力公司泰州供电公司, 东南大学 filed Critical 国家电网公司
Priority to CN201310588855.6A priority Critical patent/CN104655948A/en
Publication of CN104655948A publication Critical patent/CN104655948A/en

Links

Abstract

The invention provides a novel multistage fault diagnosis method for a power transformer. The method uses two primary algorithms including a posterior probability support vector machine (SVM) and an evidence theory, and realizes fault diagnosis on the transformer. For a to-be-detected sample, firstly, the to-be-detected sample is preprocessed with a codeless four-ratio method and a characteristic gas method, and two sets of characteristic vectors can be generated; secondly, basic probability assignment (BPA) is provided for the evidence theory by utilizing one-to-one posterior probability estimation of various SVMs; finally, a multistage comprehensive diagnosis system is created. The method can fully utilize redundant or complementary information of chromatographic data and solves a problem of uncertainty between a fault mode and fault characteristics; at the same time, the analyzability of the system can be improved, the identification capability of the fault mode is effectively improved, and the accuracy and reliability of a diagnosis result can be remarkably improved in comparison with those of a conventional diagnosis result.

Description

The multistage method for diagnosing faults of a kind of novel power transformer
Technical field
The present invention relates to the multistage method for diagnosing faults of a kind of novel power transformer.
Background technology
Power transformer belongs to the visual plant in electric system, once break down, will produce huge economic loss.Therefore, the fault diagnosis of power transformer plays a very important role in the safe operation of electric system.Because the structure of power transformer is comparatively complicated, the uncertain factor that fault relates to is more, traditional characteristic signal (such as, oil chromatography, dielectric loss, iron core grounding current, shelf depreciation etc.) there is larger limitation in the fault that reflects, and therefore need some new intelligent algorithms of research badly and be applied in the fault diagnosis of power transformer.
For many years, dissolved gas analysis technology (DGA) is applicable to the on-line monitoring of power transformer because of it, there is the advantage that analytical effect is good simultaneously, often in conjunction with various artificial intelligence technology (such as, rough set, gray theory, fuzzy clustering etc.), make the widespread use which gived in diagnosing fault of power transformer.Foundation for DGA diagnosis mainly includes improvement three-ratio method, David's triangulation method, characteristic gas method etc.But in the fault diagnosis of power transformer, usual same fault mode can show with different characteristic formps, and same fault signature may be caused by different faults pattern.The interference of various environmental factor in uncertain mapping relations between fault mode and fault signature and diagnostic procedure, single DGA diagnosis basis does not meet diagnostic requirements, needs to carry out comprehensive diagnos by multiple criterion.
In order to improve precision and the reliability of fault diagnosis, correlative study has utilized the information fusion method of evidence theory, space or temporal redundant information and complementary information are merged by rule of combination, describe with the consistance obtaining measurand, thus the type of fault is made judge more accurately.But the rational structure of evidence body basic probability assignment BPA is the difficult problem that evidence theory is applied in transformer fault diagnosis.Have two kinds of concrete solutions at present: one is expertise assignment, two is intelligent algorithm structures.But if indirect assignment BPA, subjectivity is too heavy; If be BPA by neural network Output rusults assignment after treatment, reliability coefficient is wherein still experience value.The above-mentioned method mentioned all has some limitations, and determines that the method for BPA lacks objectivity by expertise.
Fault diagnosis is in fact a kind of recognition and classification problem of failure message, and the core algorithm of fault diagnosis is identification and the classification problem of failure message.Propose more and more higher requirement along with to the safety and reliability of electric system, how inventing a kind of new diagnosing fault of power transformer algorithm with higher reliability becomes the problem needing solution badly.
Summary of the invention
The invention provides the multistage method for diagnosing faults of a kind of novel power transformer, it carries out comprehensive diagnos by multiple criterion to the incipient fault of power transformer, thus obtains the higher result of reliability.
Present invention employs following technical scheme: the multistage method for diagnosing faults of a kind of novel power transformer, comprise the following steps: step one, test sample book is extracted from power transformer, utilize, without coding four ratioing technigues and characteristic gas method, pre-service is carried out to test sample book, obtain two groups of corresponding eigenvectors, wherein, the eigenvector without coding four ratioing technigues is: CH 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C 2h 6/ CH 4; The eigenvector of characteristic gas method is C 2h 2/ (C 1+ C 2), H 2/ (C 1+ C 2+ H 2), C 2h 4/ (C 1+ C 2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2) and the relative size of total hydrocarbon; Step 2, according in step 1 obtain eigenvector, utilize posterior probability support vector machine obtain corresponding SVM posterior probability export r ij ; Then, by all r ij iterative posterior probability vector p,be the Basic probability assignment function BPA of all kinds of fault under this evidence body; Step 3, according to evidence theory definition, calculate the fusion results to all kinds of fault under the synergy of many evidences body, finally carry out judging and reaching a conclusion according to decision rule; Step 4, for transformer fault, complete above-mentioned steps one to step 3; Then, for the result that primary diagnosis draws, again complete above-mentioned steps one to step 3, the result drawn is last diagnosis.Now can form the multilevel diagnostic system of power transformer.
In step 2 of the present invention, posterior probability support vector machine specific algorithm is as follows: step one, the iclass and jclass data sample, after the data prediction of wherein one group of evidence body, obtains corresponding eigenvector; Step 2, for iclass and jthe eigenvector of class data sample, obtains corresponding SVM posterior probability and exports r ij ; Step 3, by all r ij iterative posterior probability vector p, that is the basic probability assignment BPA of all kinds of fault under this evidence body.
Step 3 kind evidence theory specific algorithm of the present invention is as follows: step one, the BPA of all kinds of fault under two the evidence bodies utilizing posterior probability support vector machine to calculate, and according to evidence theory definition, calculates degree of belief and the likelihood score of each fault type; Step 2, for the belief function of different evidence body, calculates the fusion results to all kinds of fault under the synergy of many evidences body; Step 3, according to decision rule: the degree of belief of target faults is the maximal value in all degree of beliefs, judges and reach a conclusion to the incipient fault of testing data.
The present invention has following beneficial effect: 1, posterior probability support vector machine not only remains SVM small sample, feature that generalization ability is strong, ensure the model also setting up stronger Generalization Ability in limited fault sample situation, overcome the shortcoming of SVM hard decision simultaneously, probability output is provided, ensure that the objectivity of follow-up BPA assignment; 2, posterior probability support vector machine provides many class probabilities of approximate true value, meets the feature of unascertained information input in evidence theory, forms mutual supplement with each other's advantages between the two; The multiple diagnostic method based on DGA foundation is merged by evidence theory, efficiently solves criterion result contradiction problem; 3, the belief function value of physical fault that after merging, comprehensive diagnos method judges is compared with the belief function value under single-mode and is increased to some extent, relatively reduces the belief function value of other type, thus the uncertainty of system is significantly declined.When two class posterior probability support vector machine identify separately, there will be situation about cannot pass judgment on.After evidence fusion, integrated diagnostic system can accurately identify, and that is, only identify by single failure pattern, uncertainty is higher, sometimes accurately cannot identify fault type.Evidence fusion based on posterior probability support vector machine adds the analyticity of system, effectively improves the recognition capability of fault mode.
Accompanying drawing explanation
Fig. 1 is the multistage integrated diagnostic system process flow diagram of the present invention.
Fig. 2 is single-stage diagnostic procedure schematic diagram of the present invention.
Embodiment
The invention provides the multistage method for diagnosing faults of a kind of novel power transformer, comprise the following steps: step one, test sample book is extracted from power transformer, utilize, without coding four ratioing technigues and characteristic gas method, pre-service is carried out to test sample book, obtain two groups of corresponding eigenvectors, wherein, the eigenvector without coding four ratioing technigues is: CH 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C 2h 6/ CH 4; The eigenvector of characteristic gas method is C 2h 2/ (C 1+ C 2), H 2/ (C 1+ C 2+ H 2), C 2h 4/ (C 1+ C 2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2) and the relative size of total hydrocarbon; Step 2, according in step 1 obtain eigenvector, utilize posterior probability support vector machine obtain corresponding SVM posterior probability export r ij ; Then, by all r ij iterative posterior probability vector p,be the Basic probability assignment function BPA of all kinds of fault under this evidence body, in step 2, posterior probability support vector machine specific algorithm is as follows: step one, and iclass and jclass data sample, after the data prediction of wherein one group of evidence body, obtains corresponding eigenvector; Step 2, for iclass and jthe eigenvector of class data sample, obtains corresponding SVM posterior probability and exports r ij ; Step 3, by all r ij iterative posterior probability vector p, that is the basic probability assignment BPA of all kinds of fault under this evidence body; Step 3, to define according to evidence theory, calculate the fusion results to all kinds of fault under the synergy of many evidences body, finally carry out judging and reaching a conclusion according to decision rule, step 3 kind evidence theory specific algorithm of the present invention is as follows: step one, the BPA of all kinds of fault under two the evidence bodies utilizing posterior probability support vector machine to calculate, according to evidence theory definition, calculate degree of belief and the likelihood score of each fault type; Step 2, for the belief function of different evidence body, calculates the fusion results to all kinds of fault under the synergy of many evidences body; Step 3, according to decision rule: the degree of belief of target faults is the maximal value in all degree of beliefs, judges and reach a conclusion to the incipient fault of testing data; Step 4, for transformer fault, complete above-mentioned steps one to step 3; Then, for the result that primary diagnosis draws, again complete above-mentioned steps one to step 3, the result drawn is last diagnosis.Now can form the multilevel diagnostic system of power transformer.
The present invention illustrates further: the present invention extracts test sample book from power transformer, for data sample to be measured, first utilizes and carries out pre-service without coding four ratioing technigues and characteristic gas method to it, produce two groups of corresponding eigenvectors; Then " one to one " multiclass SVM posterior probability estimation is utilized to provide basic probability assignment (BPA) for evidence theory; Finally construct a multistage integrated diagnostic system, the incipient fault of power transformer is diagnosed.Based on the multistage method for diagnosing faults of power transformer of posterior probability support vector machine and evidence theory, comprise following step:
Step 1, utilization carry out pre-service without coding four ratioing technigues and characteristic gas method to test sample book, obtain two groups of corresponding eigenvectors.Wherein, the eigenvector without coding four ratioing technigues is: CH 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C 2h 6/ CH 4; The eigenvector of characteristic gas method is C 2h 2/ (C 1+ C 2), H 2/ (C 1+ C 2+ H 2), C 2h 4/ (C 1+ C 2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2) and the relative size of total hydrocarbon.Nothing coding four ratioing technigues described in step 1 are: oil dissolved gas component generally comprises H 2, CH 4, C 2h 6, C 2h 4, C 2h 2, CO, CO 2.CH is had without the Gas Ratio classification in coding rate method 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C 2h 2/ (C 1+ C 2), H 2/ (C 1+ C 2+ H 2), C 2h 4/ (C 1+ C 2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2), (CH 4+ C 2h 4)/(C 1+ C 2).In conjunction with DGA diagnosis criterion, the eigenvector without the employing of coding four ratioing technigue that the present invention relates to is: CH 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C 2h 6/ CH 4.
Characteristic gas method described in step 1 is: the characteristic gas dissolved in transformer oil can reflect surrounding's oil, paper insulated thermal decomposition essence that trouble spot causes.Gas composition feature causes change along with fault type, fault energy and the difference of insulating material that relates to thereof, and namely trouble spot produces between the degree of unsaturation of hydrocarbon gas and the energy density of the source of trouble and has substantial connection, and physical relationship is as shown in table 1.
Table 1 judges the characteristic gas method of transformer fault character
The eigenvector of the characteristic gas method that the present invention adopts is C 2h 2/ (C 1+ C 2), H 2/ (C 1+ C 2+ H 2), C 2h 4/ (C 1+ C 2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2) and the relative size of total hydrocarbon.Wherein the relative size of total hydrocarbon adopts fuzzy membership to process.Subordinate function selects common and important ridge shape distribution to describe, and with reference to total hydrocarbon data, Construction of subordinate function is as follows:
Wherein represent the relative size of total hydrocarbon, represent the actual measured value of total hydrocarbon;
Step 2, in step 1 obtain eigenvector, utilize posterior probability support vector machine obtain corresponding SVM posterior probability export r ij ; Then, by all r ij iterative posterior probability vector p,be the Basic probability assignment function (BPA) of all kinds of fault under this evidence body, the posterior probability algorithm of support vector machine described in step 2 is as follows: for input test sample x, the standard output of SVM is: y=sgn ( f( x)), wherein f( x)= wx+ b, w, bbe constant.The present invention uses the direct matching posterior probability of sigmoid function parameter model, by standard SVM output valve fbe mapped as probable value , that is:
Wherein, parameter a, bobtain by solving the maximum likelihood problem shown in following formula according to training airplane
Wherein, lfor the sample number in training set; for sample x i probabilistic estimated value.
In view of the diagnosing fault of power transformer that the present invention relates to is many classification problems, therefore, the present invention adopts Pairwise Coupling method to complete the conversion of two class probabilities to many class probabilities.Pairwise Coupling method utilizes minimized with relative entropy set up objective function to solve posterior probability vector p, its objective function is
Wherein, n ij ? iclass and jthe number of class training sample; r ij be iclass and jthe posterior probability of the single SVM that class training sample obtains exports.Due to , then following form can be rewritten as:
Finally by following formula to all r ij iterative
Step 3, according to evidence theory definition, for the belief function of different evidence body, utilize following formula can obtain merge after belief assignment be:
(wherein, n ij ? iclass and jthe number of class training sample; r ij be iclass and jthe posterior probability of the single SVM that class training sample obtains exports) calculate the fusion results to all kinds of fault under the synergy of many evidences body, finally carry out judging and reaching a conclusion according to decision rule.The decision rule that the present invention adopts is: the degree of belief of target faults is the maximal value in all degree of beliefs, and the evidence theory described in step 3 is: the power set 2 of given an identification framework Q, Q qon a mapping m: 2 q→ [0,1], meets: and , claim m( a) be the BPA function on framework Q, it illustrates right adirect degree of support, wherein m( a) >0 abe called burnt unit; Degree of belief is ; Likelihood score is; Title interval [ bel (A), pl (A)] be aconfidence interval, represent right aconfirmation degree, determine current evidence pair ahold the bound of trusting degree.
The general type of combining evidences rule is: establish bel1, bel2 ..., beltfor on domain U tthe belief function of individual corroboration, its corresponding BPA function is respectively m 1 , m 2 ..., m t , corresponding burnt unit is respectively , then the belief assignment after merging is:
Step 4, for transformer fault, complete above-mentioned steps 1 to step 3, the primary diagnosis shown in accompanying drawing 1 can be completed; Then, for the result that primary diagnosis draws, complete above-mentioned steps 1 to step 3, can complete the secondary diagnosis shown in accompanying drawing 1, the result drawn is last diagnosis.The multilevel diagnostic system being power transformer now formed.
The present invention is mainly used in the fault diagnosis of power transformer, and concrete embodiment is as follows:
In fig. 2, the determination of step 1, Power Transformer Faults type.If for the set of transformer fault type, practice shows: power transformer interior fault can be divided into overheating fault and discharge fault by character.The statistics of pertinent literature to oil-filled electric equipment fault type is as shown in table 2:
The statistics of table 2 oil-filled electric equipment fault type
Fault type Platform Ratio/(%)
Superheated steam drier 226 53.0
High-energy discharge fault 65 18.1
Overheated double high-energy discharge fault 36 10.0
Spark discharge fault 25 7.0
Make moist or partial discharges fault 7 1.9
Therefore thus, transformer level fault may be defined as:
In formula: f 0for Superheated steam drier; f 1for discharging fault; f 2for compositeness fault; f 3for other fault.
Transformer secondary failure may be defined as:
In formula: f 0for cryogenic overheating (<300oC); f 1for middle temperature overheated (300oC-700oC); f 2for hyperthermia and superheating (>700oC); f 3for shelf depreciation; f 4for low energy electric discharge; f 5for low energy electric discharge is double overheated; f 6for arc discharge; f 7for arc discharge with over heat.
Step 2, utilization carry out pre-service without coding four ratioing technigues and characteristic gas method to test sample book, obtain two groups of corresponding eigenvectors.Wherein, the eigenvector without coding four ratioing technigues is: CH 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C2H6/CH4; The eigenvector of characteristic gas method is C2H2/(C1+C2), H2/(C1+C2+H2), C2H4/(C1+C2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2) and the relative size of total hydrocarbon; Wherein the relative size of total hydrocarbon adopts fuzzy membership to process.Subordinate function selects common and important ridge shape distribution to describe, and with reference to total hydrocarbon data, Construction of subordinate function is as follows:
Wherein represent the relative size of total hydrocarbon, represent the actual measured value of total hydrocarbon.
Step 3, in step 2 obtain eigenvector, utilize posterior probability support vector machine obtain corresponding SVM posterior probability export r ij ;
Wherein, parameter a, bobtain by solving the maximum likelihood problem shown in following formula according to training airplane
Wherein, lfor the sample number in training set; for sample x i probabilistic estimated value.
Then, when hwhen value reaches minimum, now corresponding pbe posterior probability vector p,also be the Basic probability assignment function (BPA) of all kinds of fault under this evidence body simultaneously;
Step 4, according to evidence theory definition, for the belief function of different evidence body, utilize following formula can obtain merge after belief assignment be:
(wherein, n ij ? iclass and jthe number of class training sample; r ij be iclass and jthe posterior probability of the single SVM that class training sample obtains exports) calculate the fusion results to all kinds of fault under the synergy of many evidences body, finally carry out judging and reaching a conclusion according to decision rule.The decision rule that the present invention adopts is: the degree of belief of target faults is the maximal value in all degree of beliefs;
Step 5, for the level fault of transformer described in step 1, completing steps 2 to step 4, completes the primary diagnosis shown in accompanying drawing 1; Then, for the result that primary diagnosis draws, completing steps 2 to step 4, complete the secondary diagnosis shown in accompanying drawing 1, the result drawn is last diagnosis.The multilevel diagnostic system being power transformer now formed.
Do below by three embodiments and further illustrate:
1, example 1
Certain main-transformer model is SFZ9-31500-110, the heavy 11.2t of oil, in April, 1997 goes into operation, find that main transformer oil reaches 65oC on March 25th, 2003, do not meet with the environment temperature of be with 20000kVA load and 18oC at that time, judge that main transformer inside may exist Superheated steam drier, the gas composition content in oil is:
Code of direct ratio is 0,1,1, does not exist within the scope of code of direct ratio, therefore cannot judge.But adopt the inventive method to diagnose, have result shown in table 2:
Table 2 example 1 diagnostic result
As can be seen from Table 2, in primary diagnosis, the mass value of Superheated steam drier is maximum, and in secondary diagnosis, the mass value of cryogenic overheating is maximum, therefore thinks that this type of fault is cryogenic overheating (<300oC).From on-the-spot actual diagnosis, by on-the-spot pendant-core examination, discovery is transformer core multipoint earth faults, belongs to cryogenic overheating type.
2, example 2
On June 2nd, 2006, the 35KV south of a city becomes 1# main transformer and puts into operation, and model is SFZ9-31500-110, and when putting into operation, oil chromatogram analysis is reported as:
Code of direct ratio is 1,0,0, and tentative diagnosis is arc discharge.But adopt context of methods to diagnose, have result shown in table 3:
Table 3 example 2 diagnostic result
As can be seen from Table 3, the mass value of other fault is maximum, therefore thinks that this type of fault is other fault.In actual maintenance, no abnormal through the test of transformer conventional electrical, check and find have one to loosen in the non-end shield fixed screw of sleeve pipe three after draining the oil, but without obvious spark tracking, fasteningly carry out degassed process to oil afterwards, main transformer test run stratographic analysis is normal.
3, example 3
Taihu Lake becomes 110kV#2 main transformer rated capacity 31500 kVA, and rated voltage is 110 kV/38 kV/11 kV, and model is SFSLZB-31500/110, and in Dec, 1997 puts into operation, on February 26th, 2009, becomes 110kV#2 main transformer carry out periodicity oil sample collecting work to Taihu Lake.By oil chromatogram analysis, find main transformer body oil analysis data exception, acetylene content increases by a fairly big margin, and other gas content also has the increase of different amplitude, and the gas composition content at that time in oil is:
Code of direct ratio is 0,2,2, and tentative diagnosis is hyperthermia and superheating (>700oC).But adopt context of methods to diagnose, have result shown in table 4:
Table 4 example 3 diagnostic result
As can be seen from Table 4, in primary diagnosis, the mass value of compositeness fault is maximum, and in secondary diagnosis, to hold concurrently mass value overheated maximum for high-energy discharge, determines that this type of fault type is that high-energy discharge is held concurrently overheated.By the historical data analysis of this transformer and the power failure inspection at scene, draw this transformer because multipoint earthing of iron core, be in operation, the iron core be in highfield can form closed-loop path at earth point, produce circulation, this electric current will cause the local pyrexia of iron core, and long-term heating make transformer core certain a bit burn out, produce electric arc, thus cause the continuous increase of multiple gases content in insulating oil analysis.
For precision and the reliability of assessment models, diagnose the transformer fault data that other 45 have clear and definite conclusion by context of methods, its diagnostic result compares with the diagnostic result of improvement three-ratio method and the many classification of SVM, and it the results are shown in Table 5.
Table 53 kinds of diagnostic result summary sheets
As can be seen from Table 7, total discrimination is diagnosed to be 75.6% based on improvement three ratio; Be 82.2% based on the total discrimination of the many classification diagnosis of SVM, and reach 91.1% based on the total discrimination of multistage comprehensive diagnos herein, discrimination significantly improves, and describes the validity of this model.

Claims (3)

1. the multistage method for diagnosing faults of novel power transformer, comprises the following steps:
Step one, extracts test sample book from power transformer, utilizes and carries out pre-service without coding four ratioing technigues and characteristic gas method to test sample book, obtain two groups of corresponding eigenvectors, and wherein, the eigenvector without coding four ratioing technigues is: CH 4/ H 2, C 2h 2/ C 2h 4, C 2h 4/ C 2h 6, C 2h 6/ CH 4; The eigenvector of characteristic gas method is C 2h 2/ (C 1+ C 2), H 2/ (C 1+ C 2+ H 2), C 2h 4/ (C 1+ C 2), CH 4/ (C 1+ C 2), C 2h 6/ (C 1+ C 2) and the relative size of total hydrocarbon;
Step 2, according in step 1 obtain eigenvector, utilize posterior probability support vector machine obtain corresponding SVM posterior probability export r ij ; Then, by all r ij iterative posterior probability vector p,be the Basic probability assignment function BPA of all kinds of fault under this evidence body;
Step 3, according to evidence theory definition, calculate the fusion results to all kinds of fault under the synergy of many evidences body, finally carry out judging and reaching a conclusion according to decision rule;
Step 4, for transformer fault, complete above-mentioned steps one to step 3; Then, for the result that primary diagnosis draws, again complete above-mentioned steps one to step 3, the result drawn is last diagnosis, now can form the multilevel diagnostic system of power transformer.
2. the multistage method for diagnosing faults of power transformer according to claim 1, is characterized in that in step 2, posterior probability support vector machine specific algorithm is as follows:
Step one, the iclass and jclass data sample, after the data prediction of wherein one group of evidence body, obtains corresponding eigenvector;
Step 2, for iclass and jthe eigenvector of class data sample, obtains corresponding SVM posterior probability and exports r ij ;
Step 3, by all r ij iterative posterior probability vector p, that is the basic probability assignment BPA of all kinds of fault under this evidence body.
3. the multistage method for diagnosing faults of power transformer according to claim 1, is characterized in that step 3 kind evidence theory specific algorithm is as follows:
Step one, the BPA of all kinds of fault under two the evidence bodies utilizing posterior probability support vector machine to calculate, according to evidence theory definition, calculates degree of belief and the likelihood score of each fault type;
Step 2, for the belief function of different evidence body, calculates the fusion results to all kinds of fault under the synergy of many evidences body;
Step 3, according to decision rule: the degree of belief of target faults is the maximal value in all degree of beliefs, judges and reach a conclusion to the incipient fault of testing data.
CN201310588855.6A 2013-11-21 2013-11-21 Novel multistage fault diagnosis method for power transformer CN104655948A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310588855.6A CN104655948A (en) 2013-11-21 2013-11-21 Novel multistage fault diagnosis method for power transformer

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310588855.6A CN104655948A (en) 2013-11-21 2013-11-21 Novel multistage fault diagnosis method for power transformer

Publications (1)

Publication Number Publication Date
CN104655948A true CN104655948A (en) 2015-05-27

Family

ID=53247313

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310588855.6A CN104655948A (en) 2013-11-21 2013-11-21 Novel multistage fault diagnosis method for power transformer

Country Status (1)

Country Link
CN (1) CN104655948A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105510729A (en) * 2014-10-11 2016-04-20 国家电网公司 Overheating fault diagnosis method of transformer
CN105629109A (en) * 2015-12-30 2016-06-01 西安工程大学 ARTI-neural network-based oil-immersed transformer fault diagnosis method
CN107271867A (en) * 2017-06-27 2017-10-20 国网河南省电力公司检修公司 GIS partial discharge fault type recognition method based on D S evidence theories
CN107273924A (en) * 2017-06-06 2017-10-20 上海电力学院 The Fault Analysis of Power Plants method of multi-data fusion based on fuzzy cluster analysis
CN107895194A (en) * 2017-10-20 2018-04-10 上海电力学院 A kind of nuclear power plant's main coolant system method for diagnosing faults
CN108021942A (en) * 2017-12-01 2018-05-11 朱震 A kind of power transformer incipient fault diagnostic method
CN109738595A (en) * 2019-03-07 2019-05-10 福建工程学院 A kind of system and method based on Internet of Things detection transformer fault
CN110208658A (en) * 2019-05-23 2019-09-06 国网天津市电力公司电力科学研究院 The method that a kind of pair of shelf depreciation diagnostic result carries out multivariate complement cross validation

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587155A (en) * 2009-06-08 2009-11-25 浙江大学 Oil soaked transformer fault diagnosis method
CN101692113A (en) * 2009-10-12 2010-04-07 天津大学 Method for diagnosing fault of power transformer on the basis of interval mathematical theory
CN101701940A (en) * 2009-10-26 2010-05-05 南京航空航天大学 On-line transformer fault diagnosis method based on SVM and DGA
CN102411106A (en) * 2011-11-18 2012-04-11 广东电网公司广州供电局 Fault monitoring method and device for power transformer
CN103245861A (en) * 2013-05-03 2013-08-14 云南电力试验研究院(集团)有限公司电力研究院 Transformer fault diagnosis method based on Bayesian network

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587155A (en) * 2009-06-08 2009-11-25 浙江大学 Oil soaked transformer fault diagnosis method
CN101692113A (en) * 2009-10-12 2010-04-07 天津大学 Method for diagnosing fault of power transformer on the basis of interval mathematical theory
CN101701940A (en) * 2009-10-26 2010-05-05 南京航空航天大学 On-line transformer fault diagnosis method based on SVM and DGA
CN102411106A (en) * 2011-11-18 2012-04-11 广东电网公司广州供电局 Fault monitoring method and device for power transformer
CN103245861A (en) * 2013-05-03 2013-08-14 云南电力试验研究院(集团)有限公司电力研究院 Transformer fault diagnosis method based on Bayesian network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
司马莉萍等: "基于SVM和D-S证据理论的电力变压器内部故障部位识别", 《电力自动化设备》 *
陈新岗等: "一种结合DS证据信息融合的变压器故障诊断方法", 《变压器》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105510729A (en) * 2014-10-11 2016-04-20 国家电网公司 Overheating fault diagnosis method of transformer
CN105629109A (en) * 2015-12-30 2016-06-01 西安工程大学 ARTI-neural network-based oil-immersed transformer fault diagnosis method
CN107273924B (en) * 2017-06-06 2020-05-15 上海电力学院 Multi-data fusion power plant fault diagnosis method based on fuzzy clustering analysis
CN107273924A (en) * 2017-06-06 2017-10-20 上海电力学院 The Fault Analysis of Power Plants method of multi-data fusion based on fuzzy cluster analysis
CN107271867A (en) * 2017-06-27 2017-10-20 国网河南省电力公司检修公司 GIS partial discharge fault type recognition method based on D S evidence theories
CN107895194A (en) * 2017-10-20 2018-04-10 上海电力学院 A kind of nuclear power plant's main coolant system method for diagnosing faults
CN108021942A (en) * 2017-12-01 2018-05-11 朱震 A kind of power transformer incipient fault diagnostic method
CN109738595A (en) * 2019-03-07 2019-05-10 福建工程学院 A kind of system and method based on Internet of Things detection transformer fault
CN110208658A (en) * 2019-05-23 2019-09-06 国网天津市电力公司电力科学研究院 The method that a kind of pair of shelf depreciation diagnostic result carries out multivariate complement cross validation

Similar Documents

Publication Publication Date Title
Liu et al. A systematic approach for dynamic security assessment and the corresponding preventive control scheme based on decision trees
Ray et al. Optimal feature and decision tree-based classification of power quality disturbances in distributed generation systems
CN105846780B (en) A kind of photovoltaic module method for diagnosing faults based on decision-tree model
Zhang et al. Evaluating clean energy alternatives for Jiangsu, China: An improved multi-criteria decision making method
CN102621421B (en) Transformer state evaluation method based on correlation analysis and variable weight coefficients
CN102221651B (en) Fault on-line diagnosis and early warning method of flameproof dry-type transformer for mine
CN102662113B (en) Comprehensive diagnosis method of oil-immersed transformer based on fault tree
Li et al. Interpretation of DGA for transformer fault diagnosis with complementary SaE-ELM and arctangent transform
CN104764984B (en) The improved method of transformer oil paper insulation dielectric response Equivalent Circuit Parameter identification
Rudd et al. A generic knowledge-based approach to the analysis of partial discharge data
CN106777984B (en) A method of photovoltaic array Working state analysis and fault diagnosis are realized based on density clustering algorithm
Peng et al. Application of K-Means method to pattern recognition in on-line cable partial discharge monitoring
CN102663412B (en) Power equipment current-carrying fault trend prediction method based on least squares support vector machine
CN104914327A (en) Transformer fault maintenance prediction method based on real-time monitoring information
CN103760464B (en) Solve and the low current neutral grounding system fault route selecting method of SVM based on analysis diagram
CN105930901A (en) RBPNN-based transformer fault diagnosis method
CN106156485A (en) Method for diagnosing fault of power transformer and device
CN103713237B (en) A kind of power system transmission line short trouble diagnostic method
CN103136587A (en) Power distribution network operating state classification recognition method based on support vector machine
CN105512448B (en) A kind of appraisal procedure of power distribution network health index
CN103323755A (en) Method and system for recognition of GIS ultrahigh frequency partial discharge signal
CN104502795A (en) Intelligent fault diagnosis method suitable for microgrid
CN103245861B (en) A kind of transformer fault diagnosis method based on Bayesian network
CN103576050B (en) A kind of running status appraisal procedure of capacitance type potential transformer
CN108107324B (en) Power distribution network fault positioning method based on deep convolutional neural network

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20150527

RJ01 Rejection of invention patent application after publication