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
- value
- firefly
- atom
- fault diagnosis
- 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
- 238000004804 winding Methods 0.000 title claims abstract description 90
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 35
- 238000002405 diagnostic procedure Methods 0.000 title abstract 2
- 238000003745 diagnosis Methods 0.000 claims abstract description 67
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 31
- 238000012360 testing method Methods 0.000 claims abstract description 23
- 239000013598 vector Substances 0.000 claims abstract description 14
- 241000254158 Lampyridae Species 0.000 claims description 48
- GNBHRKFJIUUOQI-UHFFFAOYSA-N fluorescein Chemical compound O1C(=O)C2=CC=CC=C2C21C1=CC=C(O)C=C1OC1=CC(O)=CC=C21 GNBHRKFJIUUOQI-UHFFFAOYSA-N 0.000 claims description 20
- 210000002569 neuron Anatomy 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 17
- 210000004205 output neuron Anatomy 0.000 claims description 10
- 210000002364 input neuron Anatomy 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000010521 absorption reaction Methods 0.000 claims description 2
- 230000002238 attenuated effect Effects 0.000 claims description 2
- 230000007423 decrease Effects 0.000 claims description 2
- 238000012804 iterative process Methods 0.000 claims description 2
- 230000031700 light absorption Effects 0.000 claims description 2
- 238000004088 simulation Methods 0.000 abstract 1
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005520 electrodynamics Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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 invention relates to the technical field of fault diagnosis, in particular to a transformer winding fault diagnosis method based on atomic sparse decomposition.
Background
The failure of the transformer winding is a great hidden trouble for the safe operation of a power system, and the serious failure caused by the mechanical deformation of the winding under the action of electrodynamic force accounts for 70 percent of the total failure of the winding according to statistics. Therefore, the method has very important significance for finding the potential fault hazard of the transformer winding in time, avoiding sudden accidents and developing the research of transformer winding fault diagnosis. For the problem of fault diagnosis of transformer windings, at present, researchers at home and abroad propose various fault diagnosis methods. The main fault diagnosis methods include a short-circuit impedance method, a frequency response method, a rough set theory, a support vector machine, a neural network and the like. The short-circuit impedance method and the frequency response method can be realized only by stopping the transformer, and the detection method is complex and has low precision. The rough set theory has great superiority in processing fuzzy and uncertain information, but the decision rule is unstable and has poor accuracy, and the rough set theory is based on a complete information system, and a data loss phenomenon is often encountered when data is processed. The support vector machine has advantages in solving the problems of small samples, nonlinearity and high-dimensional pattern recognition, but the recognition capability is easily influenced by self parameters. The neural network has a simple structure and strong problem solving capability, and can better process noise data, but the algorithm has a local optimal problem, the convergence is poor, and the reliability is limited.
Therefore, in the prior art, the transformer winding fault diagnosis method has the problems of low precision, poor reliability, large deviation of diagnosis results and the like.
Disclosure of Invention
In view of the above, the main object of the present invention is to provide a transformer winding fault diagnosis method with high precision, good reliability and accurate diagnosis result.
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
a transformer winding fault diagnosis method based on atomic sparse decomposition comprises the following steps:
step 1, simulating transformer winding faults under different working conditions by adopting an S11-M-500/35 type three-phase transformer test, and selecting an LC0154J type voltage type acceleration sensor to acquire vibration signals of different fault types of the transformer winding;
step 2, carrying out atom sparse decomposition on the vibration signals to obtain attenuation modal parameters X ═ X (X) representing different fault types of the transformer winding1，x2，...，x5)T；
Step 3, preprocessing the attenuation modal parameter X to obtain a characteristic vector T (T) of the transformer winding under different fault types1，t2，...，t5)T；
Step 4, randomly extracting a plurality of groups of the feature vectors T as training samples T1＝(t11，t12，...，t15)TThe rest is a test sample T2＝(t21，t22，...，t25)T；
Step 5, training the sample T1As input, the fault type of the transformer winding is used as output, and a GSO-SOM network fault diagnosis model of the transformer winding is established;
step 6, training a transformer winding GSO-SOM network fault diagnosis model;
step 7, judging whether the end condition is met; if yes, executing step 8; if not, executing step 6;
step 8, obtaining the trained transformer winding based on atomic sparse decompositionThe group GSO-SOM network fault diagnosis model is used for analyzing the test sample T2Inputting a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, judging the fault type of the transformer winding, and outputting a diagnosis result.
In summary, the transformer winding fault diagnosis method based on atomic sparse decomposition of the invention adopts a three-phase transformer test to simulate transformer winding faults under different working conditions, collects vibration signals of the transformer winding, carries out atomic sparse decomposition on the vibration signals to obtain attenuation modal parameters, carries out data preprocessing on the attenuation modal parameters to obtain characteristic vectors of the transformer winding under different fault types, divides the characteristic vectors into training samples and test samples, takes the training samples as input and the fault types of the transformer winding as output, establishes a transformer winding GSO-SOM network fault diagnosis model, trains the fault diagnosis model to obtain the trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, inputs the test samples into the trained fault diagnosis model to judge the faults of the transformer winding, and the diagnosis result is output, so that the precision, the accuracy and the reliability of the fault diagnosis of the transformer winding are improved.
Drawings
Fig. 1 is a flowchart of a transformer winding fault diagnosis method based on atomic sparse decomposition according to the present invention.
Fig. 2 is a schematic diagram of the SOM network topology according to the present invention.
Fig. 3 is a graph showing the variation trend of the system error according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Fig. 1 is a flowchart of a transformer winding fault diagnosis method based on atomic sparse decomposition according to the present invention. As shown in fig. 1, the transformer winding fault diagnosis method of the present invention includes the following steps:
step 1, simulating transformer winding faults under different working conditions by adopting an S11-M-500/35 type three-phase transformer test, and selecting an LC0154J type voltage type acceleration sensor to acquire vibration signals of different fault types of the transformer winding;
step 2, carrying out atom sparse decomposition on the vibration signals to obtain attenuation modal parameters X ═ X (X) representing different fault types of the transformer winding1，x2，...，x5)T；
Step 3, preprocessing the attenuation modal parameter X to obtain a characteristic vector T (T) of the transformer winding under different fault types1，t2，...，t5)T；
Step 4, randomly extracting a plurality of groups of the feature vectors T as training samples T1＝(t11，t12，...，t15)TThe rest is a test sample T2＝(t21，t22，...，t25)T；
Step 5, training the sample T1As input, the fault type of the transformer winding is used as output, and a GSO-SOM network fault diagnosis model of the transformer winding is established;
step 6, training a transformer winding GSO-SOM network fault diagnosis model;
step 7, judging whether the end condition is met; if yes, executing step 8; if not, executing step 6;
step 8, obtaining a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, and testing the modelSample T2Inputting a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, judging the fault type of the transformer winding, and outputting a diagnosis result.
The transformer winding fault diagnosis method based on atomic sparse decomposition adopts a three-phase transformer test to simulate transformer winding faults under different working conditions, collects vibration signals of the transformer winding faults, carries out atomic sparse decomposition on the vibration signals to obtain attenuation modal parameters, carries out data preprocessing on the attenuation modal parameters to obtain characteristic vectors of the transformer winding under different fault types, divides the characteristic vectors into training samples and test samples, takes the training samples as input and the fault types of the transformer winding as output, establishes a transformer winding GSO-SOM network fault diagnosis model, trains the fault diagnosis model to obtain the trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, inputs the test samples into the trained fault diagnosis model to judge the faults of the transformer winding, and the diagnosis result is output, so that the precision, the accuracy and the reliability of the fault diagnosis of the transformer winding are improved.
In the method of the present invention, the step 2 comprises the following steps:
step 21, selecting Gabor atoms as time-frequency atoms, which are formed by a modulated Gaussian window functionThe calculated formula for the Gabor atom is as follows:
wherein s is a scaling factor, epsilon is a frequency factor,is a phase factor;
step 22, from overcompleteFinding and vibrating signal x in Gabor atom libraryi(t) the most closely matching atom, i.e. the 1 st most closely matching atomThe specific calculation formula is as follows:
wherein,in order to be the original residual signal,the number is the v time-frequency atom in the overcomplete Gabor atom library, and D is the overcomplete Gabor atom library of vibration signal sparse decomposition;
step 23, the 1 st best matching atomFrom vibration signals xi(t) extracting to obtain the 1 st residual signalThe specific calculation formula is as follows:
step 24, when the increase value of the inner product exceeds the current 1% and the increase values of the current scaling factor, frequency factor and phase factor 3 variable exceed 10% of the current value, repeating iterative operation according to the step 22 and the step 23, wherein the specific calculation formula is as follows:
wherein,for the atom found during the mth decomposition iteration,for the residual signal generated during the mth decomposition iteration,decomposing a residual signal generated in the iterative process of the (m-1) th time;
step 25, performing inner product operation on the new atom and the latest residual error signal during each iteration, and terminating the iteration when the inner product added value is less than the current 1%, or the added values of the current scaling factor, frequency factor and phase factor 3 variables are less than 10% of the self-value, wherein the atom obtained at this time is the optimal atom gγi(MA)Meanwhile, the obtained scaling factor, frequency factor and phase factor 3 variables are the optimal 3 variables(s)γi(MA)，εγi(MA)，)；
Step 26, if the current best atom gγi(MA)When the waveform is attenuated, the wave form is composed ofCalculate an attenuation factor αγi(MA)If the current best atom gγi(MA) When the waveform is divergent, the wave form is composed of Calculate an attenuation factor αγi(MA)；
Step 27, according to the sorting method from small to large, finding the best atom gγi(MA)Of amplitudeMaximum value, i.e. maximum amplitude A of decaying sinusoidal atomsγi(MA)；
Step 28, obtaining the optimal 3 variables s of the vibration signal according to the above 7 stepsγi(MA)，εγi(MA)，And an attenuation factor αγi(MA)Attenuating the maximum amplitude A of the sine-wave atomsγi(MA)I.e. attenuation mode parameters X characterizing different types of mechanical faults.
In step 3 of the present invention, the calculation formula of the data preprocessing is as follows:
t (p) is the sample value of the p-th attenuation mode parameter, xact(p) is the actual value of the p-th attenuation mode parameter, xmin(p) is the minimum value of the p-th attenuation mode parameter, xmax(p) is the maximum value of the p-th attenuation mode parameter.
In the method of the present invention, the step 5 comprises the following steps:
step 51, using training sample T1The neurons are used as input layers of the GSO-SOM network, an output layer of the GSO-SOM network is a two-dimensional network with 6 multiplied by 6 output neurons, the neurons of the output layer are arranged in a neighborhood structure, each neuron is laterally connected with other neurons around the neuron, each input neuron is connected with all the output neurons, and the input neurons and the output neurons are connected with a weight wijIs an initial value of [0, 1 ]]The initial threshold b of the GSO-SOM network is a smaller non-zero random number;
step 52, encoding the initial connection weight value and the initial threshold value b of the GSO-SOM network input neuron and output neuron by adopting a real number vector form to form a firefly initial population, initializing the number N of the firefly population and an attraction coefficient β0Light absorption coefficient gamma and randomness coefficient α0Wherein the attraction coefficient β01, the light absorption coefficient γ is [0, 1 ]]Distributed random number, randomness coefficient α0∈[0，1]。
Step 53, calculating the individual fitness function value of the firefly, wherein the specific calculation formula is as follows:
wherein f is the individual fitness value of the firefly, and Z is the training sample T1Number of (a), ykIs the actual output value, okThe expected output value is obtained, and N is the number of the firefly populations;
step 54, calculating the fluorescein value lu(g) Fluorescein value lu(g) And current position xu(g) Represents the u, fluorescein value l of each firefly individualu(g) The specific calculation formula is as follows:
lu(g+1)＝(1-δ1)×lu(g)+ξ1×J(xu(g+1))
wherein u is a firefly individual, xu(g) Is the current location of the firefly individual u,/u(g) The magnitude of the fluorescein value, l, of the firefly individual u at the g-th iterationu(g +1) is the magnitude of the fluorescein value of firefly individual u at g +1 iteration, J (x)u(g +1)) is the value of the objective function, delta1Volatility coefficient of fluorescein value, ξ1Is the enhancement factor;
step 55, calculating the number of the fireflies larger than the fireflies, wherein the calculation formula is as follows:
Mu(g)＝{Q：duQ(g)＜ru；lu(g)＜lQ(g)}
wherein M isu(g) The number of all the fireflies with the fluorescein value larger than the firefly number per se in the firefly u sensing range, duQ(g) Is composed ofDistance, r, between individual firefly u and individual firefly QuTo sense the radius,/Q(g) The magnitude of the fluorescein value of the firefly individual Q at the g-th iteration is obtained;
step 56, obtaining the individuals with the strongest fluorescence, and updating the positions of the fireflies, wherein the specific calculation formula is as follows:
wherein, PijThe most intense fluorescent individual, P is all firefly individuals with fluorescein value greater than that of the firefly individuals in the sensing range, lp(g) The magnitude of the fluorescein value at the g-th iteration of P, s is the moving step length, xu(g +1) is the location of firefly u after updating, j is the individual firefly, xj(g) Is the current location of firefly individual j.
In the method of the present invention, the step 6 includes the following steps:
step 61, calculating the training sample T2The Eucliden distance is used for the distances from the time t to all output nodes, and the calculation formula is as follows:
wherein, Ti(T) is a training sample T2Value at time t, wijThe connection weight value between the ith input neuron node and the jth output neuron node is obtained;
step 62, selecting the minimum distance djThe node of (a) is taken as the best-matching neuron,neuron i (x) is a winning neuron;
and 63, updating the weight of the SOM network for the winning neuron, wherein the weight is calculated 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 monotonically decreases with time t, hj，i(t)(t) is the neighborhood function around the winning neuron, calculated as follows:
wherein r isj，ri(x)The positions of the SOM network output nodes j, i (x), respectively.
In step 7 of the present invention, the termination condition is specifically that the training error is less than 0.0001 or the iteration number reaches 2500.
The step 8 of the invention is specifically as follows: obtaining a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, and testing the sample T2And as the input of the fault diagnosis model, the fault type of the transformer winding is used as the output, the fault type of the transformer winding is judged, and the diagnosis result is output.
Examples
Test specimen T2As input and fault type of transformer winding as output, and testing sample T based on transformer winding fault diagnosis model of atomic sparse decomposition2Some of the data are shown in table 1. The results of the fault diagnosis of the transformer windings are shown in table 2.
TABLE 1 training sample T2Partial data
TABLE 2 diagnostic results
As can be seen from the data in table 2, when the number of training steps is 1000, the training data 1 and 2 are classified into one type, the training data 3, 4, 5, 6, 7 and 8 are classified into another type, the GSO-SOM network preliminarily classifies the data, when the number of training steps is 2500, the training data 1 and 2, 3 and 4, 5 and 6, and 7 and 8 are classified into the same type, and then the GSO-SOM network further classifies the data, so that the fault types of the transformer windings can be correctly classified. From the test result of the GSO-SOM network, the GSO-SOM network with atomic sparse decomposition is adopted to diagnose the fault of the transformer winding, and the diagnosis result is consistent with the actual fault type. From the diagnosis result of the GSO-SOM network, the fault type of the transformer winding can be accurately judged by adopting the GSO-SOM network diagnosis model of atomic sparse decomposition, and the accuracy is high.
In summary, the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A transformer winding fault diagnosis method based on atomic sparse decomposition is characterized by comprising the following steps:
step 1, simulating transformer winding faults under different working conditions by adopting an S11-M-500/35 type three-phase transformer test, and selecting an LC0154J type voltage type acceleration sensor to acquire vibration signals of different fault types of the transformer winding;
step 2, carrying out atom sparse decomposition on the vibration signals to obtain attenuation modal parameters X which represent different fault types of the transformer winding(x1，x2，...，x5)T；
Step 3, preprocessing the attenuation modal parameter X to obtain a characteristic vector T (T) of the transformer winding under different fault types1，t2，...，t5)T；
Step 4, randomly extracting a plurality of groups of the feature vectors T as training samples T1＝(t11，t12，...，t15)TThe rest is a test sample T2＝(t21，t22，...，t25)T；
Step 5, training the sample T1As input, the fault type of the transformer winding is used as output, and a GSO-SOM network fault diagnosis model of the transformer winding is established;
step 6, training a transformer winding GSO-SOM network fault diagnosis model;
step 7, judging whether the end condition is met; if yes, executing step 8; if not, executing step 6;
step 8, obtaining a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, and testing the sample T2Inputting a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, judging the fault type of the transformer winding, and outputting a diagnosis result.
2. The transformer winding fault diagnosis method according to claim 1, characterized in that the step 2 comprises the following specific steps:
step 21, selecting Gabor atoms as time-frequency atoms, which are formed by a modulated Gaussian window functionThe calculated formula for the Gabor atom is as follows:
wherein s is a scaling factor, epsilon is a frequency factor,is a phase factor;
step 22, finding and vibrating signal x from over-complete Gabor atom libraryi(t) the most closely matching atom, i.e. the 1 st most closely matching atomThe specific calculation formula is as follows:
wherein,in order to be the original residual signal,the number is the v time-frequency atom in the overcomplete Gabor atom library, and D is the overcomplete Gabor atom library of vibration signal sparse decomposition;
step 23, the 1 st best matching atomFrom vibration signals xi(t) extracting to obtain the 1 st residual signalThe specific calculation formula is as follows:
step 24, when the increase value of the inner product exceeds the current 1% and the increase values of the current scaling factor, frequency factor and phase factor 3 variable exceed 10% of the current value, repeating iterative operation according to the step 22 and the step 23, wherein the specific calculation formula is as follows:
wherein,for the atom found during the mth decomposition iteration,for the residual signal generated during the mth decomposition iteration,decomposing a residual signal generated in the iterative process of the (m-1) th time;
step 25, performing inner product operation on the new atom and the latest residual error signal during each iteration, and terminating the iteration when the inner product added value is less than the current 1%, or the added values of the current scaling factor, frequency factor and phase factor 3 variables are less than 10% of the self-value, wherein the atom obtained at this time is the optimal atom gγi(MA)Meanwhile, the obtained scaling factor, frequency factor and phase factor 3 variables are optimal 3 variables
Step 26, if the current best atom gγi(MA)When the waveform is attenuated, the wave form is composed ofCalculate an attenuation factor αγi(MA)If the current best atom gγi(MA)When the waveform is divergent, the wave form is composed of Calculate an attenuation factor αγi(MA)；
Step 27, according to the sorting method from small to large, finding the best atom gγi(MA)The maximum value of the amplitude is the maximum amplitude A of the decaying sine-wave atomsγi(MA)；
Step 28, obtaining the optimal 3 variables s of the vibration signal according to the above 7 stepsγi(MA)，εγi(MA)，And an attenuation factor αγi(MA)Attenuating the maximum amplitude A of the sine-wave atomsγi(MA)Namely, the attenuation mode parameter X is the attenuation mode parameter X for representing different fault types of the transformer winding.
3. The transformer winding fault diagnosis method according to claim 1, characterized in that in step 3, the data preprocessing is calculated as follows:
t (p) is the sample value of the p-th attenuation mode parameter, xact(p) is the actual value of the p-th attenuation mode parameter, xmin(p) is the minimum value of the p-th attenuation mode parameter, xmax(p) is the maximum value of the p-th attenuation mode parameter.
4. The transformer winding fault diagnosis method according to claim 1, characterized in that said step 5 comprises the following specific steps:
step 51, using training sample T1The neurons are used as input layers of the GSO-SOM network, the output layer of the GSO-SOM network is a two-dimensional network with 6 multiplied by 6 output neurons, the neurons of the output layer are arranged in a neighborhood structure, each neuron is laterally connected with other neurons around the neuron, each input neuron is connected with all the output neurons, and the input neurons and the output neurons are connectedConnection weight wijIs an initial value of [0, 1 ]]The initial threshold b of the GSO-SOM network is a smaller non-zero random number;
step 52, connecting the GSO-SOM network input neuron and the output neuron with a weight wijAnd the initial threshold b adopts real vector form coding to form a firefly initial population, the number N of the firefly population is initialized, and the attraction coefficient β0Light absorption coefficient gamma and randomness coefficient α0Wherein the attraction coefficient β01, the light absorption coefficient γ is [0, 1 ]]Distributed random number, randomness coefficient α0∈[0，1]。
Step 53, calculating the individual fitness function value of the firefly, wherein the specific calculation formula is as follows:
wherein f is the individual fitness value of the firefly, Z is the number of training samples, ykIs the actual output value, okThe expected output value is obtained, and N is the number of the firefly populations;
step 54, calculating the fluorescein value lu(g) Fluorescein value lu(g) And current position xu(g) Represents the u, fluorescein value l of each firefly individualu(g) The specific calculation formula is as follows:
lu(g+1)＝(1-δ1)×lu(g)+ξ1×J(xu(g+1))
wherein u is a firefly individual, xu(g) Is the current location of the firefly individual u,/u(g) The magnitude of the fluorescein value, l, of the firefly individual u at the g-th iterationu(g +1) is the magnitude of the fluorescein value of firefly individual u at g +1 iteration, J (x)u(g +1)) is the value of the objective function, delta1Volatility coefficient of fluorescein value, ξ1Is the enhancement factor;
step 55, calculating the number of the fireflies larger than the fireflies, wherein the calculation formula is as follows:
Mu(g)＝{Q：duQ(g)＜ru；lu(g)＜lQ(g)}
wherein M isu(g) The number of all the fireflies with the fluorescein value larger than the firefly number per se in the firefly u sensing range, duQ(g) Is the distance, r, between the individual firefly u and the individual firefly QuTo sense the radius,/Q(g) The magnitude of the fluorescein value of the firefly individual Q at the g-th iteration is obtained;
step 56, obtaining the individuals with the strongest fluorescence, and updating the positions of the fireflies, wherein the specific calculation formula is as follows:
wherein, PijThe most intense fluorescent individual, P is all firefly individuals with fluorescein value greater than that of the firefly individuals in the sensing range, lp(g) The magnitude of the fluorescein value at the g-th iteration of P, s is the moving step length, xu(g +1) is the location of firefly u after updating, j is the individual firefly, xj(g) Is the current location of firefly individual j.
5. The transformer winding fault diagnosis method according to claim 1, characterized in that said step 6 comprises the following specific steps:
step 61, calculating the training sample T2The Eucliden distance is used for the distances from the time t to all output nodes, and the calculation formula is as follows:
wherein, Ti(T) is a training sample T2Value at time t, wijThe connection weight value between the ith input neuron node and the jth output neuron node is obtained;
step 62, selecting the minimum distance djThe node of (a) is taken as the best-matching neuron,neuron i (x) is a winning neuron;
and 63, updating the weight of the SOM network for the winning neuron, wherein the weight is calculated 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 monotonically decreases with time t, hj，i(t)(t) is the neighborhood function around the winning neuron, calculated as follows:
wherein r isj，ri(x)The positions of the SOM network output nodes j, i (x), respectively.
6. The transformer winding fault diagnosis method according to claim 1, wherein in the step 7, the termination condition is that a training error is less than 0.0001 or that the number of iterations reaches 2500.
7. The transformer winding fault diagnosis method according to claim 1, wherein the step 8 specifically comprises: obtaining a trained transformer winding GSO-SOM network fault diagnosis model based on atomic sparse decomposition, and testing the sample T2And as the input of the fault diagnosis model, the fault type of the transformer winding is used as the output, the fault type of the transformer winding is judged, and the diagnosis result is output.
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 |
|---|---|---|
| Yang et al. | SuperGraph: Spatial-temporal graph-based feature extraction for rotating machinery diagnosis | |
| CN113935460B (en) | Intelligent diagnosis method for mechanical faults under unbalanced-like data set | |
| CN109116834B (en) | Intermittent process fault detection method based on deep learning | |
| CN111832353B (en) | Steam turbine rotor fault diagnosis method based on EMD and BA optimization SVM | |
| CN108875772B (en) | Fault classification model and method based on stacked sparse Gaussian Bernoulli limited Boltzmann machine and reinforcement learning | |
| CN109975697A (en) | A kind of Mechanical Failure of HV Circuit Breaker diagnostic method based on atom sparse decomposition | |
| CN111275004B (en) | Bearing fault diagnosis method based on LMD and impulse neural network | |
| CN101404071B (en) | Electronic circuit fault diagnosis neural network method based on grouping particle swarm algorithm | |
| CN109389171B (en) | Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology | |
| CN109975634A (en) | A kind of fault diagnostic method for transformer winding based on atom sparse decomposition | |
| CN111046916A (en) | Motor fault diagnosis method and system based on void convolution capsule network | |
| Liu et al. | Imbalanced fault diagnosis of rotating machinery using autoencoder-based SuperGraph feature learning | |
| CN111275108A (en) | Method for performing sample expansion on partial discharge data based on generation countermeasure network | |
| CN111553495A (en) | Small circuit breaker fault analysis method based on probabilistic neural network | |
| CN105606914A (en) | IWO-ELM-based Aviation power converter fault diagnosis method | |
| CN110702966B (en) | Fault arc detection method, device and system based on probabilistic neural network | |
| CN116451123A (en) | Transformer early fault diagnosis method based on ISSA-MKELM | |
| CN112926728B (en) | Small sample turn-to-turn short circuit fault diagnosis method for permanent magnet synchronous motor | |
| CN115859077A (en) | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions | |
| Tao et al. | Fault diagnosis of rolling bearing using deep belief networks | |
| CN111753751A (en) | Fan fault intelligent diagnosis method for improving firework algorithm | |
| CN112287592A (en) | Industrial equipment fault diagnosis method and system based on deep confidence network | |
| CN113884844A (en) | Transformer partial discharge type identification method and system | |
| CN110458071B (en) | DWT-DFPA-GBDT-based optical fiber vibration signal feature extraction and classification method | |
| CN114331214A (en) | Domain-adaptive bearing voiceprint fault diagnosis method and system based on reinforcement 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 |