CN109975634A - A kind of fault diagnostic method for transformer winding based on atom sparse decomposition - Google Patents
A kind of fault diagnostic method for transformer winding based on atom sparse decomposition Download PDFInfo
- Publication number
- CN109975634A CN109975634A CN201910197065.2A CN201910197065A CN109975634A CN 109975634 A CN109975634 A CN 109975634A CN 201910197065 A CN201910197065 A CN 201910197065A CN 109975634 A CN109975634 A CN 109975634A
- Authority
- CN
- China
- Prior art keywords
- transformer winding
- atom
- firefly
- fault
- value
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N99/00—Subject matter not provided for in other groups of this subclass
- G06N99/007—Molecular computers, i.e. using inorganic molecules
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)
Abstract
The present invention provides a kind of fault diagnostic method for transformer winding based on atom sparse decomposition, this method comprises: using the transformer winding fault under three-phase transformer test simulation difference operating condition, acquire its vibration signal, vibration signal is subjected to atom sparse decomposition, obtain decaying modal parameter, modal parameter of decaying is carried out to the pretreatment of data, obtain the feature vector under transformer winding different faults type, feature vector is divided into training sample and test sample, using training sample as input, the fault type of transformer winding is as output, establish transformer winding GSO-SOM network fault diagnosis model, and fault diagnosis model is trained, obtain the trained transformer winding GSO-SOM network fault diagnosis model based on atom sparse decomposition, test sample is inputted into trained fault diagnosis mould Type judges transformer winding fault, exports diagnostic result.The present invention has the characteristics that high reliablity and accuracy are good, can be widely applied to fault diagnosis field.
Description
Technical field
The present invention relates to fault diagnosis technology fields, more particularly to a kind of transformer winding based on atom sparse decomposition
Method for diagnosing faults.
Background technique
Transformer winding fault be safe operation of power system a major hidden danger according to statistics winding in electrodynamic action
It issues catastrophe failure caused by raw mechanically deform and accounts for 70%. accident rates of winding total failare and have been raised to first place.Therefore, in time
It was found that the potential faults of transformer winding, avoid burst accident, carry out the research of transformer winding fault diagnosis with very heavy
The meaning wanted.For the troubleshooting issue of transformer winding, currently, domestic and foreign scholars propose different fault diagnosis method.
Major failure diagnostic method has short circuit impedance method, frequency response method, rough set theory, support vector machines and neural network etc..It is short
Road impedance method and frequency response method require transformer shutdown and are just able to achieve, and detection method is complicated and precision is lower.Rough set reason
It obscuring by processing and there is biggish superiority on uncertain information, but its decision rule is very unstable, accuracy is poor, and
It and is that loss of data phenomenon when handling data, can often be encountered based on complete information system.Support vector machines is solving sample
Originally, there is advantage, but recognition capability is influenced vulnerable to inherent parameters in non-linear and high dimensional pattern identification problem.Neural network has
Simple structure and very strong problem solving ability, and noise data can be preferably handled, but algorithm has local optimum,
Convergence is poor, limited reliability.
It can be seen that in the prior art, fault diagnostic method for transformer winding there are precision low, poor reliability, diagnosis
As a result there is the problems such as relatively large deviation.
Summary of the invention
In view of this, the main purpose of the present invention is to provide a kind of high-precision, good reliability, diagnostic results accurately to become
Depressor winding failure diagnostic method.
In order to achieve the above object, technical solution proposed by the present invention are as follows:
A kind of fault diagnostic method for transformer winding based on atom sparse decomposition, the transformer winding fault diagnosis side
Method includes the following steps:
Step 1, using the transformer winding fault under S11-M-500/35 type three-phase transformer test simulation difference operating condition,
Choose the vibration signal of LC0154J type voltage-type acceleration transducer acquisition transformer winding different faults type；
The vibration signal is carried out atom sparse decomposition by step 2, obtains characterization transformer winding different faults type
Decay modal parameter X=(x1, x2..., x5)T；
Step 3, the pretreatment that the decaying modal parameter X is carried out to data, obtain transformer winding different faults type
Under feature vector T=(t1, t2..., t5)T；
Described eigenvector T is randomly selected several groups as training sample T by step 41=(t11, t12..., t15)T,
Remaining part is divided into test sample T2=(t21, t22..., t25)T；
Step 5, by the training sample T1As input, the fault type of transformer winding establishes transformation as output
Device winding GSO-SOM network fault diagnosis model；
Step 6, training transformer winding GSO-SOM network fault diagnosis model；
Step 7, whether true for differentiation if meeting termination condition；If set up, 8 are thened follow the steps；If invalid, execute
Step 6；
Step 8 obtains the trained transformer winding GSO-SOM network fault diagnosis mould based on atom sparse decomposition
Type, by the test sample T2The trained transformer winding GSO-SOM network failure based on atom sparse decomposition is inputted to examine
Disconnected model, judges the fault type of transformer winding, and export diagnostic result.
In conclusion the fault diagnostic method for transformer winding of the present invention based on atom sparse decomposition uses three phase transformations
Transformer winding fault under depressor test simulation difference operating condition, acquires its vibration signal, and it is sparse that vibration signal is carried out atom
It decomposes, obtains decaying modal parameter, modal parameter of decaying is carried out to the pretreatment of data, obtains transformer winding different faults class
Feature vector is divided into training sample and test sample by the feature vector under type, using training sample as input, transformer winding
Fault type establish transformer winding GSO-SOM network fault diagnosis model as output, and fault diagnosis model is carried out
Training, obtains the trained transformer winding GSO-SOM network fault diagnosis model based on atom sparse decomposition, by test specimens
The trained fault diagnosis model of this input, judges transformer winding fault, diagnostic result is exported, to improve change
Precision, the accuracy and reliability of depressor winding failure diagnosis.
Detailed description of the invention
Fig. 1 is a kind of flow chart of fault diagnostic method for transformer winding based on atom sparse decomposition of the present invention.
Fig. 2 is SOM network topology structure schematic diagram of the present invention.
Fig. 3 is trend chart of the systematic error of the present invention with frequency of training.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, right below in conjunction with the accompanying drawings and the specific embodiments
The present invention is described in further detail.
Fig. 1 is a kind of flow chart of fault diagnostic method for transformer winding based on atom sparse decomposition of the present invention.
As shown in Figure 1, fault diagnostic method for transformer winding of the present invention, includes the following steps:
Step 1, using the transformer winding fault under S11-M-500/35 type three-phase transformer test simulation difference operating condition,
Choose the vibration signal of LC0154J type voltage-type acceleration transducer acquisition transformer winding different faults type；
The vibration signal is carried out atom sparse decomposition by step 2, obtains characterization transformer winding different faults type
Decay modal parameter X=(x1, x2..., x5)T；
Step 3, the pretreatment that the decaying modal parameter X is carried out to data, obtain transformer winding different faults type
Under feature vector T=(t1, t2..., t5)T；
Described eigenvector T is randomly selected several groups as training sample T by step 41=(t11, t12..., t15)T,
Remaining part is divided into test sample T2=(t21, t22..., t25)T；
Step 5, by the training sample T1As input, the fault type of transformer winding establishes transformation as output
Device winding GSO-SOM network fault diagnosis model；
Step 6, training transformer winding GSO-SOM network fault diagnosis model；
Step 7, whether true for differentiation if meeting termination condition；If set up, 8 are thened follow the steps；If invalid, execute
Step 6；
Step 8 obtains the trained transformer winding GSO-SOM network fault diagnosis mould based on atom sparse decomposition
Type, by the test sample T2The trained transformer winding GSO-SOM network failure based on atom sparse decomposition is inputted to examine
Disconnected model, judges the fault type of transformer winding, and export diagnostic result.
Fault diagnostic method for transformer winding of the present invention based on atom sparse decomposition is tested using three-phase transformer
The transformer winding fault under different operating conditions is simulated, its vibration signal is acquired, vibration signal is subjected to atom sparse decomposition, is obtained
Modal parameter of decaying is carried out the pretreatment of data, obtains the spy under transformer winding different faults type by decaying modal parameter
Vector is levied, feature vector is divided into training sample and test sample, using training sample as input, the failure classes of transformer winding
Type is established transformer winding GSO-SOM network fault diagnosis model, and be trained to fault diagnosis model, is obtained as output
To the trained transformer winding GSO-SOM network fault diagnosis model based on atom sparse decomposition, test sample is inputted
Trained fault diagnosis model, judges transformer winding fault, export diagnostic result, thus improve transformer around
Precision, the accuracy and reliability of group fault diagnosis.
In the method for the present invention, the step 2 includes the following steps:
Step 21 chooses Gabor atom as time-frequency atom, it is by a modulated Gauss functionIt constitutes, the calculating formula of Gabor atom is as follows:
Wherein, s is contraction-expansion factor, and ε is frequency factor,For phase factor；
Step 22 is found and vibration signal x from excessively complete Gabor atomi(t) atom the most matched, the as the 1st
A most matched atomsSpecific calculating formula is as follows:
Wherein,For initial residual signals,For v-th of time-frequency atom in excessively complete Gabor atom, D is
The excessively complete Gabor atom of vibration signal sparse decomposition；
Step 23, by the 1st most matched atomsFrom vibration signal xi(t) it is extracted in, obtains the 1st residual error letter
NumberSpecific calculating formula is as follows:
Step 24, when its inner product value added is more than current 1%, and current contraction-expansion factor, frequency factor and phase because
When the value added of sub 3 variables is more than itself 10%, repeated iterative operation, specific calculating formula are carried out according to step 22 and step 23
It is as follows:
Wherein,For the atom sought during the m times Decomposition iteration,To be produced during the m times Decomposition iteration
Raw residue signal,The residue signal generated during the m-1 times Decomposition iteration；
New atom and newest residual signals are done into inner product operation when step 25, each iteration, when inner product value added deficiency is worked as
Preceding 1%, or current 3 variable of contraction-expansion factor, frequency factor and phase factor value added it is insufficient itself 10% when, iteration is whole
Only, the atom obtained at this time is best atom gγi(MA), at the same time, resulting contraction-expansion factor, frequency factor and phase factor
3 variables are optimal 3 variable (sγi(MA), εγi(MA),)；
If step 26, current best atom gγi(MA)When waveform is decaying, byIt calculates
Attenuation factor outγi(MA)If current best atom gγi(MA) waveform be diverging when, by It calculates
Attenuation factor outγi(MA)；
Step 27, the sort method of basis from small to large, find best atom gγi(MA)The maximum value of amplitude, as decays
The maximum amplitude A of sinusoidal quantity atomγi(MA)；
Step 28, the optimal 3 variable s that vibration signal can be acquired by above 7 stepγi(MA), εγi(MA),And decaying
Factor-alphaγi(MA), the maximum amplitude A of damped sinusoidal quantity atomγi(MA), as characterize the decaying mode ginseng of different type mechanical breakdown
Number X.
In step 3 of the present invention, the calculating formula of the data prediction is as follows:
T (p) is the sample value of p-th of decaying modal parameter, xact(p) be p-th of decaying modal parameter actual value, xmin
(p) be p-th of decaying modal parameter minimum value, xmax(p) be p-th of decaying modal parameter maximum value.
In the method for the present invention, the step 5 includes the following steps:
Step 51, using training sample T1As the neuron of GSO-SOM network input layer, the output of GSO-SOM network
Layer has the two-dimensional network of 6 × 6 output neurons for one, and the neuron of output layer lines up a neighbour structure, each neuron
Laterally attached with other neurons around it, each input neuron is connected to all output neurons, inputs neuron
With output neuron connection weight wijInitial value be random number in [0,1], the initial threshold b of GSO-SOM network is smaller
Non-zero random number；
Step 52 uses GSO-SOM network inputs neuron and output neuron initial connection weight and initial threshold b
Real number vector form coding, constitutes firefly initial population, initializes the number N of firefly population, attraction factor beta0, light suction
Receive coefficient gamma and randomness factor alpha0, wherein attraction factor beta0=1, absorption coefficient of light γ are the random number of [0,1] distribution, with
Machine property coefficient α0∈ [0,1].
Step 53 calculates firefly individual adaptation degree functional value, and specific calculating formula is as follows:
Wherein, f is firefly ideal adaptation angle value, and Z is training sample T1Number, ykFor actual output valve, okBy a definite date
The output valve of prestige, N are firefly population number；
Step 54 calculates fluorescein value lu(g), fluorescein value lu(g) and current location xu(g) represent each firefly
Individual u, fluorescein value lu(g) specific calculating formula is as follows:
lu(g+1)=(1- δ1)×lu(g)+ξ1×J(xu(g+1))
Wherein, u is firefly individual, xu(g) current location for being firefly individual u, luIt (g) is firefly individual u the
The size of fluorescein value, l when g iterationuIt (g+1) is the size of firefly individual u fluorescein value in the g+1 times iteration, J (xu
It (g+1)) is target function value, δ1For fluorescein value volatility coefficient, ξ1To enhance coefficient；
Step 55 calculates the firefly quantity for being greater than itself, and calculating formula is as follows:
Mu(g)={ Q:duQ(g) < ru；lu(g) < lQ(g)}
Wherein, Mu(g) it is greater than the firefly number of itself, d for all fluorescein values in firefly u sensing rangeuQ(g) it is
The distance between firefly individual u and firefly individual Q, ruFor the perception radius, lQIt (g) is firefly individual Q in the g times iteration
When fluorescein value size；
Step 56 obtains the most strong individual of fluorescence, updates firefly position, specific formula for calculation is as follows:
Wherein, PijFor most hyperfluorescence individual, P is the firefly individual that all fluorescein values are greater than itself in sensing range,
lpIt (g) is the size of P fluorescein value in the g times iteration, s is moving step length, xuIt (g+1) is the position of firefly u after update, j
For firefly individual, xj(g) current location for being firefly individual j.
In the method for the present invention, the step 6 includes the following steps:
Step 61 calculates the training sample T2In moment t to the distance of all output nodes, using Eucliden away from
From calculating formula is as follows:
Wherein, TiIt (t) is training sample T2In the value of t moment, wijIt is exported for i-th of input neuron node and j-th
Connection weight between neuron node；
Step 62, selection generate minimum range djNode be used as most matched neuron,Neuron i
It (x) is triumph neuron；
Step 63, to triumph neuron, update the weight of SOM network, the calculating formula of weight is as follows:
wij(t+1)=wij(t)+η(t)hJ, i (x)[Ti(t)-wij(t)]
Wherein, η (t) is learning efficiency, 0 < η (t) < 1, and t dullness reduction at any time, hJ, i (t)It (t) is nerve of winning
Neighborhood function around first, calculating formula are as follows:
Wherein, rj, ri(x)It is the position of SOM network output node j, i (x) respectively.
In step 7 of the present invention, termination condition is specially that training error less than 0.0001 or the number of iterations reaches 2500.
Step 8 of the present invention specifically: obtain the trained transformer winding GSO-SOM network based on atom sparse decomposition
Fault diagnosis model, by the test sample T2As the input of fault diagnosis model, the fault type conduct of transformer winding
Output, judges the fault type of transformer winding, exports diagnostic result.
Embodiment
By test sample T2As input, the fault type of transformer winding is as output, based on atom sparse decomposition
Transformer winding fault diagnostic model test sample T2Partial data is as shown in table 1.The fault diagnosis result of transformer winding is such as
Shown in table 2.
1 training sample T of table2Partial data
2 diagnostic result of table
From the data in table 2, it can be seen that training data 1,2 is divided into one kind, 3,4,5,6,7,8 quilts when train epochs are 1000
It is divided into another kind of, GSO-SOM network has carried out preliminary classification to data, when train epochs are 2500,1 and 2,3 and 4,5 and 6,
7 and 8 are divided into same class, and at this moment GSO-SOM network can carry out data further division to the fault type of transformer winding
Correct classification.From the point of view of the test result of GSO-SOM network, using atom sparse decomposition GSO-SOM network to transformer around
The failure of group is diagnosed, and diagnostic result is consistent with physical fault type.From the point of view of the diagnostic result of GSO-SOM network, use
The GSO-SOM network diagnosis model of atom sparse decomposition can accurately judge the fault type of transformer winding, and accuracy is high.
In conclusion the above is merely preferred embodiments of the present invention, being not intended to limit the scope of the present invention.
All within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in of the invention
Within protection scope.
Claims (7)
1. a kind of fault diagnostic method for transformer winding based on atom sparse decomposition, which is characterized in that the transformer winding
Method for diagnosing faults includes the following steps:
Step 1, using the transformer winding fault under S11-M-500/35 type three-phase transformer test simulation difference operating condition, choose
The vibration signal of LC0154J type voltage-type acceleration transducer acquisition transformer winding different faults type；
The vibration signal is carried out atom sparse decomposition by step 2, obtains the decaying of characterization transformer winding different faults type
Modal parameter X=(x1, x2..., x5)T；
Step 3, the pretreatment that the decaying modal parameter X is carried out to data, obtain under transformer winding different faults type
Feature vector T=(t1, t2..., t5)T；
Described eigenvector T is randomly selected several groups as training sample T by step 41=(t11, t12..., t15)T, remaining part
It is divided into test sample T2=(t21, t22..., t25)T；
Step 5, by the training sample T1As input, the fault type of transformer winding establishes transformer winding as output
GSO-SOM network fault diagnosis model；
Step 6, training transformer winding GSO-SOM network fault diagnosis model；
Step 7, whether true for differentiation if meeting termination condition；If set up, 8 are thened follow the steps；If invalid, then follow the steps
6；
Step 8 obtains the trained transformer winding GSO-SOM network fault diagnosis model based on atom sparse decomposition, will
The test sample T2Input the trained transformer winding GSO-SOM network fault diagnosis mould based on atom sparse decomposition
Type judges the fault type of transformer winding, and exports diagnostic result.
2. fault diagnostic method for transformer winding according to claim 1, which is characterized in that the step 2 includes following
Specific steps:
Step 21 chooses Gabor atom as time-frequency atom, it is by a modulated Gauss functionStructure
At the calculating formula of Gabor atom is as follows:
Wherein, s is contraction-expansion factor, and ε is frequency factor,For phase factor；
Step 22 is found and vibration signal x from excessively complete Gabor atomi(t) atom the most matched, as the 1st most
Matched atomsSpecific calculating formula is as follows:
Wherein,For initial residual signals,For v-th of time-frequency atom in excessively complete Gabor atom, D is vibration
The excessively complete Gabor atom of signal sparse decomposition；
Step 23, by the 1st most matched atomsFrom vibration signal xi(t) it is extracted in, obtains the 1st residual signalsSpecific calculating formula is as follows:
Step 24, when its inner product value added is more than current 1%, and current contraction-expansion factor, frequency factor and phase factor 3 become
When the value added of amount is more than itself 10%, repeated iterative operation is carried out according to step 22 and step 23, specific calculating formula is as follows:
Wherein,For the atom sought during the m times Decomposition iteration,For what is generated during the m times Decomposition iteration
Residue signal,The residue signal generated during the m-1 times Decomposition iteration；
New atom and newest residual signals are done into inner product operation when step 25, each iteration, when inner product value added is insufficient current
1%, or current 3 variable of contraction-expansion factor, frequency factor and phase factor value added it is insufficient itself 10% when, iteration ends,
The atom obtained at this time is best atom gγi(MA), at the same time, resulting contraction-expansion factor, frequency factor and phase factor 3 become
Amount is optimal 3 variable
If step 26, current best atom gγi(MA)When waveform is decaying, byIt calculates and declines
Subtracting coefficient αγi(MA)If current best atom gγi(MA)When waveform is diverging, by Calculate decaying
Factor-alphaγi(MA)；
Step 27, the sort method of basis from small to large, find best atom gγi(MA)The maximum value of amplitude, as attenuated sinusoidal
Measure the maximum amplitude A of atomγi(MA)；
Step 28, the optimal 3 variable s that vibration signal can be acquired by above 7 stepγi(MA), εγi(MA),And decay factor
αγi(MA), the maximum amplitude A of damped sinusoidal quantity atomγi(MA), as characterize the decaying mode of transformer winding different faults type
Parameter X.
3. fault diagnostic method for transformer winding according to claim 1, which is characterized in that in step 3, the data are pre-
The calculating formula of processing is as follows:
T (p) is the sample value of p-th of decaying modal parameter, xact(p) be p-th of decaying modal parameter actual value, xmin(p)
It is the minimum value of p-th of decaying modal parameter, xmax(p) be p-th of decaying modal parameter maximum value.
4. fault diagnostic method for transformer winding according to claim 1, which is characterized in that the step 5 includes following
Specific steps:
Step 51, using training sample T1As the neuron of GSO-SOM network input layer, the output layer of GSO-SOM network is one
A two-dimensional network for having 6 × 6 output neurons, the neuron of output layer line up a neighbour structure, each neuron with it week
Other neurons enclosed are laterally attached, and each input neuron is connected to all output neurons, input neuron and output
Neuron connection weight wijInitial value be random number in [0,1], the initial threshold b of GSO-SOM network is lesser non-zero
Random number；
Step 52, by GSO-SOM network inputs neuron and output neuron connection weight wijIt is sweared with initial threshold b using real number
Form coding is measured, firefly initial population is constituted, initializes the number N of firefly population, attraction factor beta0, the absorption coefficient of light
γ and randomness factor alpha0, wherein attraction factor beta0=1, absorption coefficient of light γ are the random number of [0,1] distribution, randomness system
Number α0∈ [0,1].
Step 53 calculates firefly individual adaptation degree functional value, and specific calculating formula is as follows:
Wherein, f is firefly ideal adaptation angle value, and Z is the number of training sample, ykFor actual output valve, okIt is desired defeated
It is worth out, N is firefly population number；
Step 54 calculates fluorescein value lu(g), fluorescein value lu(g) and current location xu(g) represent each firefly individual
U, fluorescein value lu(g) specific calculating formula is as follows:
lu(g+1)=(1- δ1)×lu(g)+ξ1×J(xu(g+1))
Wherein, u is firefly individual, xu(g) current location for being firefly individual u, luIt (g) is firefly individual u at the g times
The size of fluorescein value, l when iterationuIt (g+1) is the size of firefly individual u fluorescein value in the g+1 times iteration, J (xu(g+
It 1)) is target function value, δ1For fluorescein value volatility coefficient, ξ1To enhance coefficient；
Step 55 calculates the firefly quantity for being greater than itself, and calculating formula is as follows:
Mu(g)={ Q:duQ(g) < ru；lu(g) < lQ(g)}
Wherein, Mu(g) it is greater than the firefly number of itself, d for all fluorescein values in firefly u sensing rangeuQIt (g) is the light of firefly
The distance between worm individual u and firefly individual Q, ruFor the perception radius, lQ(g) glimmering in the g times iteration for firefly individual Q
The size of light element value；
Step 56 obtains the most strong individual of fluorescence, updates firefly position, specific formula for calculation is as follows:
Wherein, PijFor most hyperfluorescence individual, P is the firefly individual that all fluorescein values are greater than itself in sensing range, lp(g)
For the size of P fluorescein value in the g times iteration, s is moving step length, xuIt (g+1) is the position of firefly u after update, j is firefly
Fireworm individual, xj(g) current location for being firefly individual j.
5. fault diagnostic method for transformer winding according to claim 1, which is characterized in that the step 6 includes following
Specific steps:
Step 61 calculates the training sample T2It is calculated in moment t to the distance of all output nodes using Eucliden distance
Formula is as follows:
Wherein, TiIt (t) is training sample T2In the value of t moment, wijFor i-th of input neuron node and j-th of output neuron
Connection weight between node；
Step 62, selection generate minimum range djNode be used as most matched neuron,Neuron i (x) is
Triumph neuron；
Step 63, to triumph neuron, update the weight of SOM network, the calculating formula of weight is as follows:
wij(t+1)=wij(t)+η(t)hJ, i (x)[Ti(t)-wij(t)]
Wherein, η (t) is learning efficiency, 0 < η (t) < 1, and t dullness reduction at any time, hJ, i (t)It (t) is triumph neuron week
The neighborhood function enclosed, calculating formula are as follows:
Wherein, rj, ri(x)It is the position of SOM network output node j, i (x) respectively.
6. fault diagnostic method for transformer winding according to claim 1, which is characterized in that in the step 7, the end
Only condition reaches 2500 less than 0.0001 or the number of iterations for training error.
7. fault diagnostic method for transformer winding according to claim 1, which is characterized in that the step 8 specifically:
To the trained transformer winding GSO-SOM network fault diagnosis model based on atom sparse decomposition, by the test sample
T2As the input of fault diagnosis model, the fault type of transformer winding is as output, to the fault type of transformer winding
Judged, exports diagnostic result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910197065.2A CN109975634A (en) | 2019-03-04 | 2019-03-04 | A kind of fault diagnostic method for transformer winding based on atom sparse decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910197065.2A CN109975634A (en) | 2019-03-04 | 2019-03-04 | A kind of fault diagnostic method for transformer winding based on atom sparse decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109975634A true CN109975634A (en) | 2019-07-05 |
Family
ID=67079014
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910197065.2A Pending CN109975634A (en) | 2019-03-04 | 2019-03-04 | A kind of fault diagnostic method for transformer winding based on atom sparse decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109975634A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110726957A (en) * | 2019-11-05 | 2020-01-24 | 国网江苏省电力有限公司宜兴市供电分公司 | Fault identification method of dry-type reactor |
| CN111767675A (en) * | 2020-06-24 | 2020-10-13 | 国家电网有限公司大数据中心 | Transformer vibration fault monitoring method and device, electronic equipment and storage medium |
| CN112557966A (en) * | 2020-12-02 | 2021-03-26 | 国网江苏省电力有限公司南京供电分公司 | Transformer winding looseness identification method based on local mean decomposition and support vector machine |
-
2019
- 2019-03-04 CN CN201910197065.2A patent/CN109975634A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110726957A (en) * | 2019-11-05 | 2020-01-24 | 国网江苏省电力有限公司宜兴市供电分公司 | Fault identification method of dry-type reactor |
| CN111767675A (en) * | 2020-06-24 | 2020-10-13 | 国家电网有限公司大数据中心 | Transformer vibration fault monitoring method and device, electronic equipment and storage medium |
| CN112557966A (en) * | 2020-12-02 | 2021-03-26 | 国网江苏省电力有限公司南京供电分公司 | Transformer winding looseness identification method based on local mean decomposition and support vector machine |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11474163B2 (en) | Power transformer winding fault positioning method based on deep convolutional neural network integrated with visual identification | |
| CN106443310B (en) | A kind of transformer fault detection method based on SOM neural network | |
| CN110297141A (en) | Fault Locating Method and system based on multilayer assessment models | |
| CN109975634A (en) | A kind of fault diagnostic method for transformer winding based on atom sparse decomposition | |
| CN108535572B (en) | Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics | |
| CN112200694B (en) | Dominant instability mode identification model construction and application method based on graph neural network | |
| CN110045317B (en) | Mutual inductor metering error online detection method and system | |
| CN109975697A (en) | A kind of Mechanical Failure of HV Circuit Breaker diagnostic method based on atom sparse decomposition | |
| CN107942210A (en) | The classification of transformer pulse electric current Partial Discharge and denoising method and system | |
| CN111275108A (en) | Method for performing sample expansion on partial discharge data based on generation countermeasure network | |
| CN112836719B (en) | Indicator diagram similarity detection method integrating two classifications and triplets | |
| CN112766301B (en) | Oil extraction machine indicator diagram similarity judging method | |
| CN114386452B (en) | Nuclear power circulating water pump sun gear fault detection method | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN115600136A (en) | High-voltage bushing fault diagnosis method, system and medium based on multiple sensors | |
| CN111985528A (en) | PDGAN-based cable partial discharge data enhancement method | |
| CN115130546A (en) | Abnormal parameter detection method and device, computer readable medium and electronic equipment | |
| CN113514743A (en) | Construction method of GIS partial discharge pattern recognition system based on multi-dimensional features | |
| Hagh et al. | Improving bad data detection in state estimation of power systems | |
| CN115291039A (en) | Single-phase earth fault line selection method of resonance earthing system | |
| CN115099506A (en) | Wind speed prediction method, system, electronic device and medium | |
| CN109506936B (en) | Bearing fault degree identification method based on flow chart and non-naive Bayes inference | |
| Prabu et al. | Classification of single PD sources of HV transformer insulation faults using PRPD pattern features and ANN approach | |
| CN115563529A (en) | Fault diagnosis method for superconducting magnet of ultra-high-speed low-vacuum pipeline magnetic levitation traffic system | |
| CN112966345A (en) | Rotary machine residual life prediction hybrid shrinkage method based on countertraining and transfer learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190705 |