CN107884663A - A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine - Google Patents
A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine Download PDFInfo
- Publication number
- CN107884663A CN107884663A CN201711017993.3A CN201711017993A CN107884663A CN 107884663 A CN107884663 A CN 107884663A CN 201711017993 A CN201711017993 A CN 201711017993A CN 107884663 A CN107884663 A CN 107884663A
- Authority
- CN
- China
- Prior art keywords
- mrow
- transformer
- vector machine
- msub
- relevance vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/62—Testing of transformers
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Monitoring And Testing Of Nuclear Reactors (AREA)
Abstract
The present invention relates to a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine, its technical characterstic comprises the following steps:Divide the Status Type of transformer and determine the method for expressing of various Status Types;Choose the Monitoring Data of running state of transformer;Carry out feature extraction and determine the characteristic variable group for reflecting running state of transformer from different perspectives;It is determined that the combination core Method Using Relevance Vector Machine Fusion Model in the different characteristic space being made up of different characteristic set of variables;Choose corresponding kernel function;Gather sample data of the transformer under various running statuses;Enter the study and test of the combination core Method Using Relevance Vector Machine fault diagnosis model of line transformer.The present invention is reasonable in design, and it has merged the various features information for containing running state of transformer, provides more available informations for the operation maintenance of transformer, improves the accuracy of Diagnosis Method of Transformer Faults.
Description
Technical field
It is especially a kind of based on combination core Method Using Relevance Vector Machine the invention belongs to transformer equipment Condition Monitoring Technology field
Diagnosis Method of Transformer Faults.
Background technology
Power transformer is the visual plant of power system, and its running status directly affects the level of security of system.Transformation
Although device has good mechanical performance and enough electrical strengths in design, inevitably have one in manufacturing process
A little local defects, in addition in During Process of Long-term Operation, it can occur absolutely under the factor effect such as heat, the destruction of electricity and outside and influence
Edge aging, Material degradation, and trigger failure and accident.Therefore must take effective measure to the timely latency for finding transformer
Failure, to ensure transformer reliability service.With transformer state detecting/monitoring development and progressively promote, occur at present bag
Oil dissolved gas content detecting/monitoring, measurement of partial discharge (hyperfrequency and ultrasonic method etc.), the broadband of earth current is included to survey
The various detecting/monitoring device and system such as amount, the winding deformation based on optical fiber and temperature survey.Therefore, based on these detecting/monitoring numbers
It is very feasible and necessary according to transformer fault diagnosis system is established.
The content of Gases Dissolved in Transformer Oil and the ratio of gas content can reflect the fortune of transformer different in terms of
Row state, based on the traditional diagnosis methods such as this three-ratio method, the Rogers methods that have formd IEC recommendations, and ANN
The artificial intelligent diagnosing method such as network, SVMs.But these current methods are mainly the content according to dissolved gas, either
According to gas content ratio, but according only to can reflect that the single features information of running state of transformer is difficult that failure situation is done
Go out more correct analysis.
The content of the invention
It is overcome the deficiencies in the prior art the mesh of the present invention, proposes a kind of transformation based on combination core Method Using Relevance Vector Machine
Device method for diagnosing faults, it can make more correct with reference to the various features information of running state of transformer to failure situation
Analysis.
The present invention solves its technical problem and takes following technical scheme to realize:
A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine, comprises the following steps:
Step 1, the Status Type for dividing transformer, determine the method for expressing of various Status Types;
Step 2, the Monitoring Data for choosing running state of transformer;
Step 3, feature extraction is carried out to Monitoring Data, it is determined that reflecting that the feature of running state of transformer becomes from different perspectives
Amount group;
The combination core Method Using Relevance Vector Machine fusion mould in the different characteristic space that step 4, determination are made up of different characteristic set of variables
Type;
Step 5, corresponding each group characteristic variable, choose corresponding kernel function;
The sample data of step 6, collection transformer under various running statuses;
Step 7, the study and test of combining core Method Using Relevance Vector Machine fault diagnosis model for entering line transformer, using quick
Type-II maximal possibility estimations solve Study first, using greatest hope estimation and the method solution core combination parameter of quadratic programming.
The Status Type of step 1 transformer be divided into normal condition, low energy discharge condition, high-energy discharge state, in
Cryogenic overheating state, hyperthermia and superheating state and local discharge condition.
When the step 3 carries out feature extraction, become from two groups of features of characteristic gas content and characteristic gas content ratio
Amount.
The step 4 selects H2、CH4、C2H6、C2H4、C2H2Gas content is as one group of characteristic variable, and as the following formula to it
It is normalized:
In formula:x、xnewRepresent to normalize forward and backward sample value respectively;xmin、xmaxThe minimum value and most of sample is represented respectively
Big value.
The step 5 is using RBF as kernel function, and the combination core Method Using Relevance Vector Machine of the kernel function is using layering
Bayesian model structure, having for non-isomorphic multiple information data or multiple feature spaces is realized by introducing multinomial probability likelihood function
Machine merges and more classification.
The kernel functional parameter optimizes processing using K-CV and GA kernel functional parameter optimization method, its specific processing
Process is:
(1) by S sample data XsIt is each random to be divided into approximately equalised K separate subsets of element number
(2) useIt is trained, uses as training the set pair analysis modelAs checking set pair mould
Type is verified, is obtained K combination nuclear phase and is closed the right judging rate of vector machine model and the K model on corresponding checking collection;
(3) the kernel function for treating selection is assessed using following formula as GA fitness function using the average right judging rate of K model
Parameter;
Wherein,
In formula:For by data setThe grader for learning to obtain is to input vector xiClassification
As a result;For data setThe sample number contained;
(4) kernel functional parameter is chosen using GA:GA is met given using real coding mode, elitist selection method with accuracy
Value or constant generations optimum individual fitness are all mutually end condition.
The advantages and positive effects of the present invention are:
The present invention is reasonable in design, and it has merged the various features information for containing running state of transformer;Exporting transformer is
Various shape probability of states, more available informations are provided for the operation maintenance of transformer, improve the standard of transformer fault detection
True property.The present invention gives the concrete methods of realizing based on combination core Method Using Relevance Vector Machine diagnostic method, and to merge dissolved gas in oil
Exemplified by the different characteristic space that volume data proposes, using the effective of diagnosis example validation group synkaryon Method Using Relevance Vector Machine diagnostic method
Property.
Brief description of the drawings
Fig. 1 is the process chart of the present invention;
Fig. 2 is the combination core Method Using Relevance Vector Machine Fusion Model schematic diagram of the present invention;
Fig. 3 is the combination core Method Using Relevance Vector Machine principle schematic of the present invention;
Fig. 4 is the combination Method Using Relevance Vector Machine of the present invention using model structure schematic diagram;
Fig. 5 is the flow chart of the kernel functional parameter optimization method based on K-CV and GA of the present invention.
Embodiment
The embodiment of the present invention is further described below in conjunction with accompanying drawing.
A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine, as shown in figure 1, comprising the following steps:
Step 1:According to the characteristics of studying a question, the Status Type of transformer is divided, and determine the table of various Status Types
Show method.
Status Type according to needs and feature the division transformer to study a question;Such as the overall operation of research transformer
State, the running status of transformer can be divided into normal (N), low energy electric discharge (D1), high-energy discharge (D2), middle cryogenic overheating
(T12), hyperthermia and superheating (T3), 6 kinds of states of shelf depreciation (PD).To the electric discharge type for the shelf depreciation for studying transformer, can incite somebody to action
The shelf depreciation Type division of transformer is the electric discharge types such as needle point electric discharge, bubble electric discharge, suspended discharge, creeping discharge.Transformation
Device fault diagnosis is classification problem more than one, and transformer state is divided into N, D by the present invention1、D2、T12、T3, 6 kinds of states of PD, and
It is respectively adopted vectorial [0,0,0,0,0,1]T、[0,0,0,0,1,0]T、[0,0,0,1,0,0]T、[0,0,1,0,0,0]T、[0,1,
0,0,0,0]T、[1,0,0,0,0,0]TRepresent.
Step 2:According to the needs that study a question, the Monitoring Data of the selection all types of running statuses of transformer, and data volume phase
Closely.
Step 3:Feature extraction is carried out, it is determined that the characteristic variable group of running state of transformer can be reflected from different perspectives,
It is required that every group of characteristic variable can be independent sign transformer running status;Combination core Method Using Relevance Vector Machine can both merge together
The different characteristic space of one Monitoring Data extraction, can also merge the different characteristic space extracted by different Monitoring Datas, because
This can reflect that the Monitoring Data of running state of transformer can be selected.The former is for example, what fusion was extracted by dissolved gas in oil volume data
The different characteristic informations of running state of transformer can be reflected from different aspect, if dissolved gas content is one group of characteristic variable group
Into feature space, gas content ratio be the feature space of one group of characteristic variable composition, factor of created gase be one group of characteristic variable group
Into feature space etc.;The latter is for example, fusion monitors number by the pulse current Monitoring Data of transformer, ultrahigh frequency partial discharge respectively
According to the different characteristic set of variables group that can reflect running state of transformer from different aspect of the extractions such as, ultrasonic partial discharge Monitoring Data
Into different characteristic space.Monitoring Data of the present invention uses dissolved gas in oil volume data, from characteristic gas content and characteristic gas
Two groups of characteristic variables of content ratio;
Step 4:It is determined that the combination core Method Using Relevance Vector Machine fusion for the different feature spaces being made up of different characteristic set of variables
Model.Fusion Features model is as shown in Fig. 2 β in figure1、β2..., βnFor core combination parameter;During transformer normal operation, H2、CH4、
C2H6、C2H4、C2H2Content is seldom, the heat or aerogenesis speed of these gases is accelerated during electric fault;When trouble point temperature is relatively low, oil
The composition of middle dissolved gas is mainly CH4, as temperature raises, the maximum gas of factor of created gase is CH successively4、C2H6、C2H4With
C2H2;And C2H6It is unstable, it is extremely easy in decomposition as C2H4And H2, C2H4And H2Always it is accompanied, usual C2H6Content be less than CH4;
During general cryogenic overheating, H2Account for hydrogen hydrocarbon (H2、CH4、C2H6、C2H4、C2H2Content sum) more than 27%.And medium temperature
During overheat, H2Account for less than the 27% of hydrogen hydrocarbon;During hyperthermia and superheating, characteristic gas is mainly C2H4, next to that CH4, the two can account for total hydrocarbon
(CH4、C2H6、C2H4、C2H2Content sum) more than 80%;During serious overheat, micro C can be also produced2H2, it is usually no more than
The 6% of total hydrocarbon;During general shelf depreciation, total hydrocarbon content is not high, and characteristic gas is mainly H2, next to that CH4;Usual H2Hydrogen can be accounted for
More than 90%, CH of hydrocarbon4Account for more than the 90% of total hydrocarbon;It is possible that a small amount of C when discharge energy density is big2H2, it is generally less than total
The 2% of hydrocarbon.During general low energy electric discharge, total hydrocarbon content is not high, and characteristic gas is mainly C2H2And H2。C2H2Can account for total hydrocarbon 25%~
90%, C2H4Less than 20%, H of total hydrocarbon2Account for more than the 30% of hydrogen hydrocarbon.During usual high-energy discharge, characteristic gas is mainly C2H2With
H2And a considerable amount of CH4And C2H4, usual C2H2Content be higher than CH4, C2H2Account for 20%~70%, H of total hydrocarbon2Account for hydrogen hydrocarbon
30%~90%.
Therefore, gas content of the present invention selects H2、CH4、C2H6、C2H4、C2H25 kinds of gas contents as one group of characteristic variable,
Because above-mentioned 5 kinds of characteristic gas contents contain the information of total hydrocarbon content, so characteristic variable does not select total hydrocarbon content;This
Outside, above-mentioned five kinds of characteristic gas content differences are larger, in order to reduce calculation error, it are normalized as the following formula first place
Reason;Gas content ratio selects H2Account for the ratio of hydrogen hydrocarbon, CH4、C2H6、C2H4、C2H2The ratio for each accounting for total hydrocarbon becomes as feature
Amount;
In formula:x、xnewRepresent to normalize forward and backward sample value respectively;xmin、xmaxThe minimum value and most of sample is represented respectively
Big value;
Step 5:Corresponding each group characteristic variable, chooses corresponding kernel function;Corresponding core is chosen for every group of characteristic variable group
Function.The present invention need to choose two kernel functions due to selecting two groups of gas content, gas content ratio characteristic variables, this
Invention uses the RBF of generally use in transformer fault diagnosis, and Radial basis kernel function is as follows, wherein being core
The high wide parameter of function;
It is as shown in Figure 3 to combine core Method Using Relevance Vector Machine principle.Combine core Method Using Relevance Vector Machine and use layering Bayesian model structure,
By introducing multinomial probability likelihood function (multinomial probitlikelihood), non-isomorphic multi information number is realized
According to or multiple feature spaces organically blend and classify more, wherein S is the species of information data, β1,β2,…,βsFor combination core ginseng
Number;
Provided with S kind information datas, S feature space is obtained through feature extraction.Sample data set note from feature space s
ForWherein N is sample number, and D is characterized the dimension of vector
Number, C is classification number;When given kernel function, base nuclear matrix K can be obtainedS, KS∈RN×N, combinations of definitions nuclear matrixThe element of combination nuclear matrix can be obtained by following formula, wherein KsFor kernel function, KβFor compound kernel function;
Introduce auxiliary regression target variable Y ∈ RN×CWith weight parameter W ∈ RN×C, it is as follows that standard noise regression model can be obtained
Formula, wherein yncRepresent the element of Y line n c row, wcW c row are represented,Represent kβLine n, Nx(m, v) represents x
Obedience average is m, the normal distribution that variance is v;
Introduce multinomial probability Copula to be shown below, regressive object is converted into class label;
It is hereby achieved that multinomial probability likelihood function such as following formula, wherein ε is standardized normal distribution p (u)=N's (0,1)
It is expected, Φ is Gauss Cumulative Distribution Function;
It is openness in order to ensure model, it is that weight vectors introduce zero-mean, variance isStandard normal priori point
Cloth,For Study first matrix A ∈ RN×CIn element,The gamma that hyper parameter is a, b is obeyed to be distributed;
More combination Method Using Relevance Vector Machines are as shown in Figure 4 using layering Bayesian model structure, model structure as can be seen here.
Although combination core Method Using Relevance Vector Machine determines the problem of difficult in the absence of regularization coefficient, core combination parameter also can be in mould
The process Automatic Optimal of type study, but kernel functional parameter does not use automatic relational learning algorithm, it is necessary to be manually set in advance;
K-CV may insure that all sample datas are involved in the training and checking of model, and GA then has the preferably overall situation
Optimizing ability;In consideration of it, the present invention proposes the kernel functional parameter optimization method based on K-CV and GA, closed with improving combination nuclear phase
The performance of vector machine;This method treats the kernel functional parameter of selection using K-CV to assess, and kernel functional parameter is chosen using GA;
The handling process of kernel functional parameter optimization method based on K-CV and GA is as shown in figure 5, main process is as follows:
1) by S sample data XsIt is each random to be divided into approximately equalised K separate subsets of element number
2) use("-" is set difference operation) is trained as training the set pair analysis model, usesVerified as checking the set pair analysis model, can so obtain K combination nuclear phase and close vector machine model
And right judging rate of the K model on corresponding checking collection;
3) using the average right judging rate of K model as GA fitness function, it is shown below, selection is treated to assess
Kernel functional parameter;
Wherein,
In formula:For by data setThe grader for learning to obtain is to input vector xiClassification
As a result;For data setThe sample number contained;
4) kernel functional parameter is chosen using GA.GA is met given using real coding mode, elitist selection method with accuracy
Value or constant generations optimum individual fitness are all mutually end condition;In optimal of 90% or continuous 10 generation, is reached with accuracy herein
Body fitness is identical to be used as end condition.
Step 6:Gather sample data of the transformer under various running statuses.
Step 7:Enter the study and test of the combination core Method Using Relevance Vector Machine fault diagnosis model of line transformer, using quick
Type-II maximal possibility estimations solve Study first, using greatest hope estimation and the method solution core combination parameter of quadratic programming.
The output for combining core Method Using Relevance Vector Machine is that the probabilistic diagnosis result that transformer is various states is the various shapes of transformer
The output vector of the combinations of states core Method Using Relevance Vector Machine of maximum probability is [p in state1,p2,p3,p4,p5,p6], wherein p1,p2,p3,
p4,p5,p6The state of transformer is represented respectively as PD, T3、T12、D2、D1, N probability.
In order to verify the validity of institute's extracting method of the present invention, provide combination nuclear phase in conjunction with an example and close vector machine method
Examined with the transformer fault of BP neural network method (back propagation neural network, BPNN), SVM methods
The comparative analysis of disconnected example.
The BPNN input layers number of this example is 5, and hidden nodes 15, output layer neuron is 6, and hidden layer swashs
Encourage function and select tansig, output layer excitation function selects purelin, and training function selects trainlm, the choosing of weight learning function
With learngdm, validity function selects mse;SVM kernel functions select RBF, using the sorting technique of " one-to-one ", using net
The method selection rule coefficient and kernel functional parameter that lattice search is combined with 10 folding cross validations;Combine the choosing of core Method Using Relevance Vector Machine
With two groups of characteristic variables of the gas content after normalization and gas content ratio, two RBF kernel functions need to be selected;
Following table lists BPNN, SVM each respectively from the gas content after normalization, content ratio as characteristic variable
6 groups of typical transformer fault diagnosis examples of vector machine method are closed with combining nuclear phase;
The transformer fault diagnosis example of table 1.
To upper table analysis understand, same diagnostic method, characteristic variable from gas content diagnostic result for it is wrong when, and
It is probably then correct from gas content ratio diagnostic result, vice versa;As can be seen here transformer dissolved gas content and
Gas content ratio can reflect the running status of transformer in different aspect;
Corresponding to 6 transformer fault diagnosis examples in upper table, the output of combination core Method Using Relevance Vector Machine is shown below;
Matrix P row vector represents the state of transformer as PD, T3、T12、D2、D1, N probability;Such as matrix P the first row
p1=[0.1467,0.0137,0.0686,0.0000,0.0313,0.7397] represents that the state of First transformer in upper table is
PD、T3、T12、D2、D1, N probability be respectively 0.1467,0.0137,0.0686,0.0000,0.0313,0.7397, wherein transformation
Device is maximum for the probable value of normal condition, therefore diagnostic result is N, the like can obtain other 5 transformations that upper table provides
Device combination nuclear phase closes the diagnostic result of vector machine method.
From above formula, it is various shape probability of states that combination core Method Using Relevance Vector Machine, which can directly export transformer, is easy to point
The uncertainty of analysis problem.
Following table gives BPNN, SVM method and is characterized change using the gas content after normalization and gas content ratio
Fault diagnosis right judging rate during amount, and combination nuclear phase close the contrast situation of the fault diagnosis right judging rate of vector machine method.Pass through
The analysis of table 2 is understood, compared with the Diagnosis Method of Transformer Faults based on BPNN, SVM, institute's extracting method of the present invention has higher
Fault diagnosis accuracy.
The transformer fault diagnosis right judging rate contrast of table 2.BPNN, SVM and MKL-RVM method
It is emphasized that embodiment of the present invention is illustrative, rather than it is limited, therefore the present invention is simultaneously
The embodiment described in embodiment is not limited to, it is every to be drawn by those skilled in the art's technique according to the invention scheme
Other embodiment, also belong to the scope of protection of the invention.
Claims (6)
1. a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine, it is characterised in that comprise the following steps:
Step 1, the Status Type for dividing transformer, determine the method for expressing of various Status Types;
Step 2, the Monitoring Data for choosing running state of transformer;
Step 3, feature extraction is carried out to Monitoring Data, it is determined that reflecting the characteristic variable of running state of transformer from different perspectives
Group;
The combination core Method Using Relevance Vector Machine Fusion Model in the different characteristic space that step 4, determination are made up of different characteristic set of variables;
Step 5, corresponding each group characteristic variable, choose corresponding kernel function;
The sample data of step 6, collection transformer under various running statuses;
Step 7, the study and test of combining core Method Using Relevance Vector Machine fault diagnosis model for entering line transformer, using quick type-
II maximal possibility estimations solve Study first, using greatest hope estimation and the method solution core combination parameter of quadratic programming.
2. a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine according to claim 1, its feature
It is:The Status Type of step 1 transformer is divided into normal condition, low energy discharge condition, high-energy discharge state, middle low temperature
Superheat state, hyperthermia and superheating state and local discharge condition.
3. a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine according to claim 1, its feature
It is:When the step 3 carries out feature extraction, from two groups of characteristic variables of characteristic gas content and characteristic gas content ratio.
4. a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine according to claim 1, its feature
It is:The step 4 selects H2、CH4、C2H6、C2H4、C2H2Gas content is carried out to it as the following formula as one group of characteristic variable
Normalized:
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>n</mi>
<mi>e</mi>
<mi>w</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>x</mi>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>min</mi>
</msub>
</mrow>
</mfrac>
</mrow>
In formula:x、xnewRepresent to normalize forward and backward sample value respectively;xmin、xmaxThe minimum value and maximum of sample are represented respectively
Value.
5. a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine according to claim 1, its feature
It is:The step 5 is using RBF as kernel function, and the combination core Method Using Relevance Vector Machine of the kernel function is using layering pattra leaves
This model structure, realize that non-isomorphic multiple information data or the organic of multiple feature spaces are melted by introducing multinomial probability likelihood function
Close and classify more.
6. a kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine according to claim 5, its feature
It is:The kernel functional parameter optimizes processing using K-CV and GA kernel functional parameter optimization method, and it is specific treated
Cheng Wei:
(1) by S sample data XsIt is each random to be divided into approximately equalised K separate subsets of element number
(2) useIt is trained, uses as training the set pair analysis modelEnter as checking the set pair analysis model
Row checking, obtain K combination nuclear phase and close the right judging rate of vector machine model and the K model on corresponding checking collection;
(3) the kernel functional parameter for treating selection is assessed using following formula as GA fitness function using the average right judging rate of K model;
<mrow>
<mi>f</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>K</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>K</mi>
</munderover>
<mfrac>
<mn>100</mn>
<mrow>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>k</mi>
<mi>s</mi>
</msubsup>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mo><</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>t</mi>
<mi>i</mi>
</msub>
<mo>></mo>
<mo>&Element;</mo>
<msubsup>
<mi>X</mi>
<mi>k</mi>
<mi>s</mi>
</msubsup>
</mrow>
</munder>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mo>(</mo>
<mrow>
<mrow>
<mo>(</mo>
<mrow>
<msup>
<mi>X</mi>
<mi>s</mi>
</msup>
<mo>-</mo>
<msubsup>
<mi>X</mi>
<mi>k</mi>
<mi>s</mi>
</msubsup>
</mrow>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>,</mo>
<msub>
<mi>t</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
Wherein,
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mi>j</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mo>&NotEqual;</mo>
<mi>j</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
In formula:For by data setThe grader for learning to obtain is to input vector xiClassification results;For data setThe sample number contained;
(4) kernel functional parameter is chosen using GA:GA using real coding mode, elitist selection method, with accuracy meet set-point or
Constant generations optimum individual fitness is all mutually end condition.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711017993.3A CN107884663A (en) | 2017-10-27 | 2017-10-27 | A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711017993.3A CN107884663A (en) | 2017-10-27 | 2017-10-27 | A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107884663A true CN107884663A (en) | 2018-04-06 |
Family
ID=61782496
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711017993.3A Pending CN107884663A (en) | 2017-10-27 | 2017-10-27 | A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107884663A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717149A (en) * | 2018-05-25 | 2018-10-30 | 西安工程大学 | Diagnosis Method of Transformer Faults based on M-RVM fusion dynamic weightings AdaBoost |
CN109032107A (en) * | 2018-06-05 | 2018-12-18 | 国家电网公司 | Equipment fault signal based on Bayes's classification takes place frequently prediction technique |
CN109540808A (en) * | 2018-11-02 | 2019-03-29 | 湖南文理学院 | A kind of transformer detection system and method for diagnosing faults |
CN109782109A (en) * | 2019-02-15 | 2019-05-21 | 南京力通达电气技术有限公司 | It is a kind of for promoting the restrainable algorithms of transformer state diagnostic accuracy |
CN112085084A (en) * | 2020-08-24 | 2020-12-15 | 宁波大学 | Transformer fault diagnosis method based on multi-feature fusion common vector |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102735761A (en) * | 2012-06-26 | 2012-10-17 | 河海大学 | Method for predicting transformer oil chromatographic data based on relevance vector machine |
CN102944769A (en) * | 2012-10-18 | 2013-02-27 | 上海市电力公司 | Fault diagnosis method of power transformer based on extreme learning machine |
CN106093612A (en) * | 2016-05-26 | 2016-11-09 | 国网江苏省电力公司电力科学研究院 | A kind of method for diagnosing fault of power transformer |
CN106443259A (en) * | 2016-09-29 | 2017-02-22 | 国网山东省电力公司电力科学研究院 | Transformer fault diagnosis new method based on Euclidean clustering and SPO-SVM |
US20170061305A1 (en) * | 2015-08-28 | 2017-03-02 | Jiangnan University | Fuzzy curve analysis based soft sensor modeling method using time difference Gaussian process regression |
-
2017
- 2017-10-27 CN CN201711017993.3A patent/CN107884663A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102735761A (en) * | 2012-06-26 | 2012-10-17 | 河海大学 | Method for predicting transformer oil chromatographic data based on relevance vector machine |
CN102944769A (en) * | 2012-10-18 | 2013-02-27 | 上海市电力公司 | Fault diagnosis method of power transformer based on extreme learning machine |
US20170061305A1 (en) * | 2015-08-28 | 2017-03-02 | Jiangnan University | Fuzzy curve analysis based soft sensor modeling method using time difference Gaussian process regression |
CN106093612A (en) * | 2016-05-26 | 2016-11-09 | 国网江苏省电力公司电力科学研究院 | A kind of method for diagnosing fault of power transformer |
CN106443259A (en) * | 2016-09-29 | 2017-02-22 | 国网山东省电力公司电力科学研究院 | Transformer fault diagnosis new method based on Euclidean clustering and SPO-SVM |
Non-Patent Citations (1)
Title |
---|
朱永利等: ""组合核相关向量机在电力变压器故障诊断中的应用研究"", 《中国电机工程学报》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717149A (en) * | 2018-05-25 | 2018-10-30 | 西安工程大学 | Diagnosis Method of Transformer Faults based on M-RVM fusion dynamic weightings AdaBoost |
CN109032107A (en) * | 2018-06-05 | 2018-12-18 | 国家电网公司 | Equipment fault signal based on Bayes's classification takes place frequently prediction technique |
CN109032107B (en) * | 2018-06-05 | 2021-07-20 | 国家电网公司 | Equipment fault signal frequency prediction method based on Bayesian classification |
CN109540808A (en) * | 2018-11-02 | 2019-03-29 | 湖南文理学院 | A kind of transformer detection system and method for diagnosing faults |
CN109782109A (en) * | 2019-02-15 | 2019-05-21 | 南京力通达电气技术有限公司 | It is a kind of for promoting the restrainable algorithms of transformer state diagnostic accuracy |
CN112085084A (en) * | 2020-08-24 | 2020-12-15 | 宁波大学 | Transformer fault diagnosis method based on multi-feature fusion common vector |
CN112085084B (en) * | 2020-08-24 | 2023-12-15 | 宁波大学 | Transformer fault diagnosis method based on multi-feature fusion common vector |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107884663A (en) | A kind of Diagnosis Method of Transformer Faults based on combination core Method Using Relevance Vector Machine | |
Dai et al. | Dissolved gas analysis of insulating oil for power transformer fault diagnosis with deep belief network | |
CN105930901B (en) | A kind of Diagnosis Method of Transformer Faults based on RBPNN | |
CN102289682B (en) | Transformer fault diagnosis method based on integrated learning Bagging algorithm | |
CN110133146A (en) | A kind of Diagnosis Method of Transformer Faults and system considering unbalanced data sample | |
CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
CN108732528A (en) | A kind of digitalized electrical energy meter method for diagnosing faults based on depth confidence network | |
CN106548230A (en) | Diagnosis Method of Transformer Faults based on Modified particle swarm optimization neutral net | |
CN110879373B (en) | Oil-immersed transformer fault diagnosis method with neural network and decision fusion | |
Yan et al. | Transformer fault diagnosis based on BP‐Adaboost and PNN series connection | |
CN107958292A (en) | Transformer based on cost sensitive learning obscures careful reasoning method for diagnosing faults | |
CN107992880A (en) | A kind of optimal lump classification method for diagnosing faults of power transformer | |
CN109254219B (en) | A kind of distribution transforming transfer learning method for diagnosing faults considering multiple factors Situation Evolution | |
CN108020781A (en) | A kind of circuit breaker failure diagnostic method | |
CN116562114A (en) | Power transformer fault diagnosis method based on graph convolution neural network | |
Zhang et al. | A fault diagnosis method of power transformer based on cost sensitive one-dimensional convolution neural network | |
CN111539486A (en) | Transformer fault diagnosis method based on Dropout deep confidence network | |
CN111091141B (en) | Photovoltaic backboard fault diagnosis method based on layered Softmax | |
Guo et al. | Fault diagnosis of power transformers based on comprehensive machine learning of dissolved gas analysis | |
Liu et al. | An Intelligent Fault Diagnosis Method for Transformer Based on IPSO‐gcForest | |
Jian et al. | Transformer fault diagnosis model with unbalanced samples based on SMOTE algorithm and focal loss | |
Gastoni et al. | Robust state-estimation procedure based on the maximum agreement between measurements | |
CN103400213A (en) | Backbone net rack survivability assessment method based on LDA (Linear Discriminant Analysis) and PCA (Principal Component Analysis) | |
CN112036496A (en) | Nuclear power device fault diagnosis method and system | |
Jamshed et al. | Random forest classifier based dissolved gas analysis for identification of power transformer faults using gas ratio data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180406 |
|
RJ01 | Rejection of invention patent application after publication |