CN105629109B - Oil-immersed transformer fault diagnosis method based on ART1 neural network - Google Patents

Oil-immersed transformer fault diagnosis method based on ART1 neural network Download PDF

Info

Publication number
CN105629109B
CN105629109B CN201511024146.0A CN201511024146A CN105629109B CN 105629109 B CN105629109 B CN 105629109B CN 201511024146 A CN201511024146 A CN 201511024146A CN 105629109 B CN105629109 B CN 105629109B
Authority
CN
China
Prior art keywords
art1
neural network
oil
fault
immersed transformer
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.)
Expired - Fee Related
Application number
CN201511024146.0A
Other languages
Chinese (zh)
Other versions
CN105629109A (en
Inventor
宋玉琴
朱紫娟
赵洋
姬引飞
李莹
叶大伟
李超
程诚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Polytechnic University
Original Assignee
Xian Polytechnic University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Polytechnic University filed Critical Xian Polytechnic University
Priority to CN201511024146.0A priority Critical patent/CN105629109B/en
Publication of CN105629109A publication Critical patent/CN105629109A/en
Application granted granted Critical
Publication of CN105629109B publication Critical patent/CN105629109B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01RMEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Housings And Mounting Of Transformers (AREA)

Abstract

Oil-immersed transformer fault diagnosis method disclosed by the invention based on ART1 neural network, is specifically implemented according to the following steps: step 1 utilizes ART1 neural network specific algorithm, oil-immersed transformer fault diagnosis model of the building based on ART1 neural network;Step 2, the input and output amount that oil-immersed transformer fault diagnosis model of the building based on ART1 neural network obtained through step 1 is determined using four ratio methods;Step 3, after step 2, be arranged the oil-immersed transformer fault diagnosis based on ART1 neural network parameter;Step 4 carries out learning training to fault sample using the oil-immersed transformer fault diagnosis model based on ART1 neural network;Step 5 carries out identifying and diagnosing to physical fault data type using the oil-immersed transformer fault diagnosis model based on ART1 neural network.The present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network, correctly can carry out real-time online fault diagnosis for the different types of Superheated steam drier of oil-immersed transformer.

Description

Oil-immersed transformer fault diagnosis method based on ART1 neural network
Technical field
The invention belongs to power transformer resultant fault monitoring method technical fields, and in particular to one kind is based on ART1 nerve The oil-immersed transformer fault diagnosis method of network.
Background technique
Power transformer is most important electric device equipment in electric system, and causes power system accident most One of electrical equipment.The security and stability that its operating status directly affects power supply system is horizontal.Power transformer is found in time Potentiality failure ensures that transformer safety is effectively run, is power department to improve the reliability of transformer stable power-supplying One problem of crucial importance of concern.Therefore, diagnosing fault of power transformer technology is studied, the operation and maintenance of transformer are improved Level has important practical significance.
Current diagnosing fault of power transformer technology, although a variety of effective method for diagnosing faults have been developed, Major part is superposition accumulation to a certain extent, and most of method for diagnosing faults itself still lack perfect theoretical basis and are The concept system of systemization.In addition, most of method for diagnosing faults are all the characteristic signals that are shown using object to diagnose spy The failure for levying type, generally requires a large amount of failure training sample, and large-scale effectively fault sample data are in practice It is difficult to obtain.Power transformer all makes solution gas in transformer oil because of voltage class, insulation system and fault degree difference Body content has larger randomness, and there are certain redundant datas, and this requires method for diagnosing faults to have very strong fault-tolerance, The research of this respect is also not nearly enough deep at present.
On the basis for fully understanding various fault conditions and fault diagnosis the relevant technologies existing for oil-immersed power transformer On, in conjunction with home and abroad related scientific research mechanism power transformer on-line monitoring, intelligent diagnostics and state repair the research in field at It is particularly important to develop a kind of method mainly for research diagnosing fault of power transformer for fruit and practical experience.
Summary of the invention
The purpose of the present invention is to provide a kind of oil-immersed transformer fault diagnosis method based on ART1 neural network, energy It is enough correctly to carry out real-time online fault diagnosis for the different types of Superheated steam drier of oil-immersed transformer.
Institute of the invention is using technical solution, based on the oil-immersed transformer fault diagnosis method of ART1 neural network, tool Body follows the steps below to implement:
Step 1 utilizes ART1 neural network specific algorithm, oil-immersed transformer failure of the building based on ART1 neural network Diagnostic model;
Step 2 determines oil-immersed transformer of the building based on ART1 neural network obtained through step 1 using four ratio methods The input and output amount of fault diagnosis model;
Step 3, after step 2, be arranged the oil-immersed transformer fault diagnosis based on ART1 neural network parameter;
Step 4, using the oil-immersed transformer fault diagnosis model based on ART1 neural network to fault sample Practise training;
Step 5, using the oil-immersed transformer fault diagnosis model based on ART1 neural network to physical fault data class Type carries out identifying and diagnosing.
The features of the present invention also characterized in that:
Step 1 is specifically implemented according to the following steps:
Step 1.1 carries out initialization process to the connection weight between ART1 neural network model input layer and output layer, The specific method is as follows:
The input layer of setting ART1 neural network model has N number of neuron, and output layer has M neuron;
Two-value failure input vector Mode AkWith output vector BkIt is as follows respectively:
Wherein, k=1,2 ..., p, p are the number for inputting mode of learning;
Enable tij(0)=1,I=1,2 ..., N, j=1,2 ..., M;
Wherein, Vigilance parameter 0 < ρ≤1, tijFor network model feedback link weight, WijFor network model feedforward connection weight Value;
Step 1.2, after step 1.1, by two-value failure input vector modeIt is supplied to net The input layer of network;
Step 1.3, after step 1.2, calculate each neuron of ART1 neural network model output layer weighted input and, tool Body is implemented according to following algorithm:
Wherein, WijFor network model feedforward connection weight;
For specific composition two-value failure input vector Mode AkValue, specially 0 or 1;
SjFor each neuron of ART1 neural network model output layer weighted input and;
Step 1.4, after step 1.3, select the network model optimal classification of input pattern as a result, the result is according to following Algorithm is computed acquisition:
In formula, enabling the output of neuron g is 1, SjFor the weighted input of each neuron of ART1 neural network model output layer With;
Step 1.5 calculates following three formulas, and judges calculated result, judges final calculation result The Vigilance parameter ρ (0 < ρ≤1) for whether meeting the setting of ART1 neural network model, that is, judge that can ART1 neural network model connect By this recognition result:
If last formula is set up, 1.7 are entered step;
If last formula is invalid, 1.6 are entered step;
Step 1.6, after step 1.5, cancel recognition result, method particularly includes:
The output valve of output layer neuron g is first reset to 0, and by this neuron exclude identification next time range it Outside, then return step 1.5;
If all sharp used neurons are all unable to satisfy last formula in step 1.5, select one it is new Then neuron is transferred to step 1.7 as classification results;
Step 1.7 receives recognition result, and adjusts connection weight according to following algorithm:
Wherein, i=1,2 ..., N;
T is current time;
For the value of specific composition two-value failure input vector Mode A k, specifically: 0 or 1;
tgiFor network model feedback link weight;
wigFor network model feedforward connection weight;
Step 1.8 rejoins all neurons resetted in step 1.6 in identification range, and return step 1.2 is under One mode carries out identification storage;
The oil-immersed transformer fault diagnosis based on ART1 neural network has just been constructed by 1.1~step of step 1.8 Model.
Step 2 is specific to be implemented in accordance with the following methods:
First, the extraction of fault characteristic gases component is carried out to failure transformer;
Secondly, measurement oil dissolved gas CH4、C2H2、C2H6、C2H4、H2Concentration of component content;
Finally, fault feature vector T is obtained using four ratio methods:
Two concentration of component ratios are then indicated with 1 if more than 1;
If two concentration of component ratios are indicated less than 1 with 0;
1 or so, then it represents that the exposure of the intermediate change process of nature of trouble, i.e. nature of trouble is less obvious;
Ratio is bigger, then the display of nature of trouble is more obvious.
The method for obtaining fault feature vector T using four ratio methods involved in step 2 is specific as follows:
Utilize five kinds of main feature gas CH4、C2H2、C2H6、C2H4、H2Form four reduced values, it may be assumed that takeBinary coding is carried out to its ratio, obtains fault feature vector T.
Step 3 is specifically implemented according to the following steps:
Step 3.1, environment and condition according to oil-immersed power transformer actual motion, not for transformer fault diagnosis The specific Vigilance parameter ρ (0 < ρ≤1) of ART1 neural network is arranged to realize to the fast of transformer fault type in same detection accuracy Speed accurately identification judgement;
Step 3.2 carries out data initialization after step 3.1, then to ART1 neural network, waits fault feature vector T Input, then realize that ART1 neural network model identifies the diagnosis of transformer fault.
Step 4 is specific to be implemented in accordance with the following methods:
Using fault sample data, the study of fault sample is carried out to ART1 neural network, is realized to transformer fault class The classification of type;Input of the binary coding fault feature vector T obtained in conjunction with four ratio methods as ART1 neural network, selection The sample data of 20 groups of known fault types comprising nine kinds of transformer fault types carries out study instruction to ART1 neural network Practice.
Step 5 is specific to be implemented in accordance with the following methods:
The fault feature vector T that the fault data of the oil-immersed transformer under actual working environment is obtained through step 2, will be real Oil-immersed transformer fault diagnosis model of the fault feature vector T input in border based on ART1 neural network carries out accident analysis, defeated Be out of order type.
The beneficial effects of the present invention are:
(1) the present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network, can be realized transformer event The real-time online of barrier data learns and accurately identifies transformer fault type.
(2) the present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network, are not necessarily to a large amount of transformer Fault data sample is trained, and carries out reduction without the fault data sample to redundancy.
(3) the present invention is based in the oil-immersed transformer fault diagnosis method of ART1 neural network, network is to transformer Fault data being capable of self-organizing, adaptive learning, moreover it is possible to can carry out classifying from normalizing to unknown transformer fault data type.
(4) compared with ART2 and ART3 neural network, the present invention is based on the oil-immersed transformer failures of ART1 neural network to examine The ART1 neural network algorithm used in disconnected method is simply easily realized, causes failure to be examined it is not necessary that excessive parameter value is manually set Disconnected objectivity reduces.
(5) the present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network, by adjusting Vigilance parameter ρ The value of (0 < ρ≤1) can carry out the diagnosis of technique of different accuracy to the oil-immersed transformer failure under the conditions of particular job, real The maximization of existing trouble diagnosibility.
(6) the present invention is based on the oil-immersed transformer fault diagnosis method of ART1 neural network, principles are as follows: in conjunction with tradition Four ratio methods that are used in the analysis of DGA technology, by oil-immersed transformer work in different Superheated steam drier types carry out two Scale coding constitutes fault feature vector, and the fault feature vector recycled is trained and learns to ART1 neural network It practises;Shown by theory analysis and emulation experiment data result: the present invention is based on the events of the oil-immersed transformer of ART1 neural network Hindering diagnostic method correctly can carry out real-time online fault diagnosis for the different types of Superheated steam drier of oil-immersed transformer, It has broad application prospects.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the oil-immersed transformer fault diagnosis method of ART1 neural network.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
The present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network, as shown in Figure 1, specifically according to Lower step is implemented:
Step 1 utilizes ART1 neural network specific algorithm, oil-immersed transformer failure of the building based on ART1 neural network Diagnostic model is specifically implemented according to the following steps:
Step 1.1 carries out initialization process to the connection weight between ART1 neural network model input layer and output layer, The specific method is as follows:
The input layer of setting ART1 neural network model has N number of neuron, and output layer has M neuron;
Two-value failure input vector Mode AkWith output vector BkIt is as follows respectively:
Wherein, k=1,2 ..., p, p are the number for inputting mode of learning;
Enable tij(0)=1,I=1,2 ..., N, j=1,2 ..., M;
Wherein, Vigilance parameter 0 < ρ≤1, tijFor network model feedback link weight, WijFor network model feedforward connection weight Value;
Step 1.2, after step 1.1, by two-value failure input vector modeIt is supplied to net The input layer of network;
Step 1.3, after step 1.2, calculate each neuron of ART1 neural network model output layer weighted input and, tool Body is implemented according to following algorithm:
Wherein, WijFor network model feedforward connection weight;
For specific composition two-value failure input vector Mode AkValue (0 or 1);
SjFor each neuron of ART1 neural network model output layer weighted input and.
Step 1.4, after step 1.3, select the network model optimal classification of input pattern as a result, the result is according to following Algorithm is computed acquisition:
In formula, enabling the output of neuron g is 1, SjFor the weighted input of each neuron of ART1 neural network model output layer With;
Step 1.5 calculates following three formulas, and judges calculated result, judges final calculation result The Vigilance parameter ρ (0 < ρ≤1) for whether meeting the setting of ART1 neural network model, that is, judge that can ART1 neural network model connect By this recognition result:
If last formula is set up, 1.7 are entered step;
If last formula is invalid, 1.6 are entered step;
Step 1.6, after step 1.5, cancel recognition result, method particularly includes:
The output valve of output layer neuron g is first reset to 0, and by this neuron exclude identification next time range it Outside, then return step 1.5;
If all sharp used neurons are all unable to satisfy last formula in step 1.5, select one it is new Then neuron is transferred to step 1.7 as classification results;
Step 1.7 receives recognition result, and adjusts connection weight according to following algorithm:
Wherein, i=1,2 ..., N;T is current time;
For specific composition two-value failure input vector Mode AkValue (0 or 1);
tgiFor network model feedback link weight;
wigFor network model feedforward connection weight;
Step 1.8 rejoins all neurons resetted in step 1.6 in identification range, and return step 1.2 is under One mode carries out identification storage;
The oil-immersed transformer fault diagnosis based on ART1 neural network has just been constructed by 1.1~step of step 1.8 Model.
Step 2 determines oil-immersed transformer of the building based on ART1 neural network obtained through step 1 using four ratio methods The input and output amount of fault diagnosis model is specifically implemented in accordance with the following methods:
The extraction for carrying out fault characteristic gases component to failure transformer first, measures oil dissolved gas CH later4、 C2H2、C2H6、C2H4、CO、CO2And H2Concentration of component content;
In the inventive solutions, not seven kinds of main feature gases in the above transformer oil directly as network Input, but use four ratio methods and obtain fault feature vector, five kinds of main feature gas CH therein is only utilized4、 C2H2、C2H6、C2H4、H2
Four ratio methods are to make full use of five kinds of main feature gas CH4、C2H2、C2H6、C2H4、H2Form four reduced values, it may be assumed that It takesBinary coding is carried out to its ratio, obtains fault feature vector T:
Two concentration of component ratios are then indicated with 1 if more than 1;
If two concentration of component ratios are indicated less than 1 with 0;
1 or so, then it represents that the exposure of the intermediate change process of nature of trouble, i.e. nature of trouble is less obvious;
Ratio is bigger, then the display of nature of trouble is more obvious;
If there are two types of the failures of property to exist simultaneously, such as: 1011, then it may be interpreted as continuous electric spark and overheat.
Fault feature vector T is obtained using four ratio methods, every kind of fault feature vector T corresponds to a kind of fault signature type; Then fault feature vector T is the input vector of ART1 neural network, and the output of network is corresponding to fault feature vector T Fault signature type.
Step 3, after step 2, be arranged the oil-immersed transformer fault diagnosis based on ART1 neural network parameter, specifically It follows the steps below to implement:
Step 3.1, environment and condition according to oil-immersed power transformer actual motion, not for transformer fault diagnosis The specific Vigilance parameter ρ (0 < ρ≤1) of ART1 neural network is arranged to realize to the fast of transformer fault type in same detection accuracy Speed accurately identification judgement;
Step 3.2 carries out data initialization after step 3.1, then to ART1 neural network, waits fault feature vector T Input, then realize that ART1 neural network model identifies the diagnosis of transformer fault.
Step 4, using the oil-immersed transformer fault diagnosis model based on ART1 neural network to fault sample Training is practised, is specifically implemented in accordance with the following methods:
Using fault sample data, the study of fault sample is carried out to ART1 neural network, to realize to transformer event Hinder the classification of type;Input of the binary coding fault feature vector T obtained in conjunction with four ratio methods as ART1 neural network, Select the sample data of 20 groups of known fault types comprising nine kinds of transformer fault types to ART1 neural network Practise training;
In order to precisely objectively embody the trouble diagnosibility of ART1 neural network, input data sample learning 2 is taken turns to come Training network, then examines ART1 neural network to the accuracy rate of transformer fault diagnosis by test data sample.
Step 5, using the oil-immersed transformer fault diagnosis model based on ART1 neural network to physical fault data class Type carries out identifying and diagnosing, specifically implements in accordance with the following methods:
The fault feature vector T that the fault data of the oil-immersed transformer under actual working environment is obtained through step 2, will be real Oil-immersed transformer fault diagnosis model of the fault feature vector T input in border based on ART1 neural network carries out accident analysis, defeated Be out of order type.
The transformer of certain substation 180MVA 10 groups of fault datas since running have been put into production into as ART1 nerve net The input of network, and diagnose and also compare ART1 neural network and routine three-ratio method, ART1 nerve net is compared with this Diagnosis capability of the network to the transformer fault type.Show the oil-immersed transformer event based on ART1 neural network by comparing Barrier diagnosis to transformer fault type have more accurate diagnostic analysis ability, accuracy rate 90%, and routine three-ratio method Accuracy rate only has 70%.
The present invention is based on be applied to ART1 nerve net in the oil-immersed transformer fault diagnosis method of ART1 neural network Network, ART1 neural network can while study while work, to transformer fault data can self-organizing, self study, to unknown change Depressor fault type can carry out classifying and remembering from normalizing.When testing it, ART1 neural network can be to having remembered The sample automatic identification and quick response of study go out transformer fault type.
The present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network to have the advantage that
(1) the present invention is based on refreshing based on ART1 involved in the oil-immersed transformer fault diagnosis method of ART1 neural network Fault Diagnosis Model for Power Transformer through network is not necessarily to a large amount of fault data sample, does not require to carry out reduction to fault data, keep away Redundant data and processing big data problem bring inconvenience are exempted from.
(2) it the present invention is based in the oil-immersed transformer fault diagnosis method of ART1 neural network, is obtained using four ratio methods To transformer fault feature binary input vectors, dimension-reduction treatment is carried out to fault data feature input vector, reduces ART1 The complexity of neural network, so that ART1 Neural Network Diagnosis model calculation speed is faster, diagnostic accuracy is also higher.
(3) the present invention is based on constructed in the oil-immersed transformer fault diagnosis method of ART1 neural network based on ART1 mind Oil-immersed transformer fault diagnosis diagnostic model through network can be to the event according to the transformer fault instance data being collected into Barrier diagnostic model is trained and emulates.
The present invention is based on the oil-immersed transformer fault diagnosis methods of ART1 neural network, are tied by theory analysis and emulation Fruit shows the transformer fault diagnosis energy with higher of the oil-immersed transformer fault diagnosis method based on ART1 neural network Power;More importantly the strategy can self-organizing, self study, oil-immersed transformer unknown failure type can be carried out from normalizing Reason memory, is not necessarily to a large amount of learning training fault data sample, the diagnosis of real-time online transformer fault may be implemented;It is in oil Also superior to conventional method and other traditional neural network methods in immersion transformer fault diagnosis performance.

Claims (3)

1. the oil-immersed transformer fault diagnosis method based on ART1 neural network, which is characterized in that specifically according to the following steps Implement:
Step 1 utilizes ART1 neural network specific algorithm, oil-immersed transformer fault diagnosis of the building based on ART1 neural network Model;
Step 2 determines oil-immersed transformer failure of the building based on ART1 neural network obtained through step 1 using four ratio methods The input and output amount of diagnostic model;
The step 2 is specific to be implemented in accordance with the following methods:
First, the extraction of fault characteristic gases component is carried out to failure transformer;
Secondly, measurement oil dissolved gas CH4、C2H2、C2H6、C2H4、H2Concentration of component content;
Finally, fault feature vector T is obtained using four ratio methods:
Two concentration of component ratios are then indicated with 1 if more than 1;
If two concentration of component ratios are indicated less than 1 with 0;
1 or so, then it represents that the exposure of the intermediate change process of nature of trouble, i.e. nature of trouble is less obvious;
Ratio is bigger, then the display of nature of trouble is more obvious;
The method for obtaining fault feature vector T using four ratio methods involved in the step 2 is specific as follows:
Utilize five kinds of main feature gas CH4、C2H2、C2H6、C2H4、H2Form four reduced values, it may be assumed that takeBinary coding is carried out to its ratio, obtains fault feature vector T;
Step 3, after step 2, according to the environment and condition of oil-immersed transformer actual motion, not for transformer fault diagnosis The specific Vigilance parameter ρ of ART1 neural network, 0 ρ≤1 <, to realize to the fast of transformer fault type is arranged in same detection accuracy Speed accurately identification judgement;Data initialization is carried out to ART1 neural network again, waits the input of fault feature vector T;
Step 4, using fault sample data, the study of fault sample is carried out to ART1 neural network, is realized to transformer fault The classification of type;Input of the binary coding fault feature vector T obtained in conjunction with four ratio methods as ART1 neural network, choosing The sample data for having selected 20 comprising nine kinds of transformer fault types groups of known fault types learns ART1 neural network Training;
Step 5, using the oil-immersed transformer fault diagnosis model based on ART1 neural network to physical fault data type into Row identifying and diagnosing.
2. the oil-immersed transformer fault diagnosis method according to claim 1 based on ART1 neural network, feature exist In the step 1 is specifically implemented according to the following steps:
Step 1.1 carries out initialization process, specific method to the connection weight between ART1 neural network input layer and output layer It is as follows:
The input layer of setting ART1 neural network has N number of neuron, and output layer has M neuron;
Two-value failure input vector Mode AkWith output vector BkIt is as follows respectively:
Wherein, k=1,2 ..., p, p are the number for inputting mode of learning;
Enable tij(0)=1,
Wherein, tijFor network model feedback link weight, WijFor network model feedforward connection weight;
Step 1.2, after step 1.1, by two-value failure input vector modeIt is supplied to network Input layer;
Step 1.3, after step 1.2, calculate each neuron of ART1 neural network output layer weighted input and, specifically according to Lower algorithm is implemented:
Wherein, WijFor network model feedforward connection weight;
For specific composition two-value failure input vector Mode AkValue, specifically: 0 or 1;
SjFor each neuron of ART1 neural network output layer weighted input and;
Step 1.4, after step 1.3, select the network model optimal classification of input pattern as a result, the result is according to following algorithm It is computed acquisition:
In formula, enabling the output of neuron g is 1, SjFor each neuron of ART1 neural network output layer weighted input and;
Step 1.5 calculates following three formulas, and judges calculated result whether judge final calculation result Meet the Vigilance parameter ρ of ART1 neural network setting, 0 ρ≤1 < judges that can ART1 neural network receive this identification knot Fruit:
For specific composition two-value failure input vector Mode AkValue, specifically: 0 or 1;
tgiFor network model feedback link weight;
If last formula is set up, 1.7 are entered step;
If last formula is invalid, 1.6 are entered step;
Step 1.6, after step 1.5, cancel recognition result, method particularly includes:
The output valve of the neuron g of output layer is first reset to 0, and by this neuron exclude identification next time range it Outside, then return step 1.5;
If all sharp used neurons are all unable to satisfy last formula in step 1.5, a new nerve is selected Member is used as classification results, is then transferred to step 1.7;
Step 1.7 receives recognition result, and adjusts connection weight according to following algorithm:
Wherein, i=1,2 ..., N;
T is current time;
For specific composition two-value failure input vector Mode AkValue, specifically: 0 or 1;
tgiFor network model feedback link weight;
wigFor network model feedforward connection weight;
Step 1.8 rejoins all neurons resetted in step 1.6 in identification range, and return step 1.2 is to next Mode carries out identification storage;
The oil-immersed transformer fault diagnosis model based on ART1 neural network has just been constructed by 1.1~step of step 1.8.
3. the oil-immersed transformer fault diagnosis method according to claim 1 based on ART1 neural network, feature exist In the step 5 is specific to be implemented in accordance with the following methods:
The fault feature vector T of fault data is obtained through step 2, fault feature vector T is inputted based on ART1 neural network Oil-immersed transformer fault diagnosis model carries out accident analysis, exports fault type.
CN201511024146.0A 2015-12-30 2015-12-30 Oil-immersed transformer fault diagnosis method based on ART1 neural network Expired - Fee Related CN105629109B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201511024146.0A CN105629109B (en) 2015-12-30 2015-12-30 Oil-immersed transformer fault diagnosis method based on ART1 neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201511024146.0A CN105629109B (en) 2015-12-30 2015-12-30 Oil-immersed transformer fault diagnosis method based on ART1 neural network

Publications (2)

Publication Number Publication Date
CN105629109A CN105629109A (en) 2016-06-01
CN105629109B true CN105629109B (en) 2019-03-08

Family

ID=56044231

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201511024146.0A Expired - Fee Related CN105629109B (en) 2015-12-30 2015-12-30 Oil-immersed transformer fault diagnosis method based on ART1 neural network

Country Status (1)

Country Link
CN (1) CN105629109B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106292631A (en) * 2016-08-25 2017-01-04 哈尔滨理工大学 A kind of PWM rectifier fault diagnosis system based on neutral net
CN106338673A (en) * 2016-09-29 2017-01-18 国网山东省电力公司电力科学研究院 Transformer winding fault characteristic parameter extraction method
CN106840406B (en) * 2016-11-29 2019-08-16 浙江中新电力工程建设有限公司自动化分公司 Isolation switch method for diagnosing faults based on matrix neural network
CN107133378B (en) * 2017-03-28 2019-07-23 浙江大学 Oil-immersed transformer fault early warning method based on the estimation of Higher Dimensional Linear Models height
CN110110784B (en) * 2019-04-30 2020-03-24 贵州电网有限责任公司 Transformer fault identification method based on transformer related operation data

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100222661B1 (en) * 1997-07-08 1999-10-01 서정욱 Method for preestimating electric intensity using neural network
CN101587155B (en) * 2009-06-08 2011-05-18 浙江大学 Oil soaked transformer fault diagnosis method
CN103268516A (en) * 2013-04-16 2013-08-28 郑州航空工业管理学院 Transformer fault diagnosing method based on neural network
CN103942240B (en) * 2013-10-29 2018-03-27 北京科东电力控制系统有限责任公司 A kind of building method of intelligent substation integrated data Information application platform
CN104655948A (en) * 2013-11-21 2015-05-27 国家电网公司 Novel multistage fault diagnosis method for power transformer
CN104102214A (en) * 2014-02-19 2014-10-15 盐城华盛变压器制造有限公司 Transformer and transformer peripheral circuit fault control method

Also Published As

Publication number Publication date
CN105629109A (en) 2016-06-01

Similar Documents

Publication Publication Date Title
CN105629109B (en) Oil-immersed transformer fault diagnosis method based on ART1 neural network
CN108492000B (en) Fault diagnosis method for nonstationary characteristic of million-kilowatt ultra-supercritical unit
CN109917205A (en) A kind of solenoid valve failure diagnostic device and method based on feature extraction and multi-layer perception (MLP)
CN108921230A (en) Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network
CN108536128A (en) A kind of machine learning fault diagnosis system of parameter optimization
CN105320987B (en) A kind of satellite telemetering data intelligent interpretation method based on BP neural network
CN105182141A (en) Transformer fault detection method and system
CN109597396B (en) A kind of distribution transforming on-line fault diagnosis method based on high amount of traffic and transfer learning
CN109871000A (en) A kind of closed loop industrial process sensor method for diagnosing faults of data-driven
WO2021114320A1 (en) Wastewater treatment process fault monitoring method using oica-rnn fusion model
CN109472097A (en) A kind of transmission line of electricity on-line monitoring equipment method for diagnosing faults
CN108805195A (en) A kind of motor group method for diagnosing faults based on two-value deep-neural-network
CN111881627A (en) Nuclear power device fault diagnosis method and system
CN106292631A (en) A kind of PWM rectifier fault diagnosis system based on neutral net
CN108959498A (en) A kind of big data processing platform and its design method for health monitoring
CN112365009A (en) Secondary equipment abnormity diagnosis method based on deep learning network
CN112200263A (en) Self-organizing federal clustering method applied to power distribution internet of things
CN110209145A (en) One kind being based on the approximate carbon dioxide absorption tower method for diagnosing faults of nuclear matrix
CN112465010A (en) DKPCA and KFDA-based transformer fault detection and classification method
CN107689015A (en) A kind of improved power system bad data recognition method
CN104101795A (en) Transformer fault control method
CN104102214A (en) Transformer and transformer peripheral circuit fault control method
CN110119579B (en) OICA-based complex industrial process fault monitoring method
CN105279553B (en) A kind of height adds to water system fault degree recognition methods
CN114358116A (en) Oil-immersed transformer fault diagnosis method and system and readable storage medium

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20190308

Termination date: 20191230