CN109669087A - A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion - Google Patents
A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion Download PDFInfo
- Publication number
- CN109669087A CN109669087A CN201910101086.XA CN201910101086A CN109669087A CN 109669087 A CN109669087 A CN 109669087A CN 201910101086 A CN201910101086 A CN 201910101086A CN 109669087 A CN109669087 A CN 109669087A
- Authority
- CN
- China
- Prior art keywords
- transformer
- data
- network
- layer
- fault
- 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
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion, multiple limited Boltzmann machines and one layer of backpropagation BP neural network are stacked by the method, constitute deepness belief network model, the model is trained using the transformer fault data of acquisition, transformer station high-voltage side bus data are inputted into trained deepness belief network model, obtain the basic probability assignment of various failures, then it recycles DS evidence theory to merge probability of malfunction, obtains the final result of transformer diagnosis.The present invention combines the deepness belief network in deep learning with DS evidence theory, potential valuable information can be extracted in the complicated data characteristics of transformer, this method compensates for the deficiency of conventional fault diagnosis method well, the accuracy of transformer fault diagnosis result can be effectively improved, provides guarantee for the safety of electric system.
Description
Technical field
The present invention relates to a kind of based on Multi-source Information Fusion, can Accurate Diagnosis Power Transformer Faults method, belong to
Transformer technology field.
Background technique
Power transformer is one of electric equipments of electric system, and operating status directly affects electric system
Production safety and economic benefit, but since its own structure is more complicated, locating working environment is complex, severe, and event occurs
The probability of barrier problem is larger, for example, iron core failure, insulation fault, component faults, discharge fault, short trouble etc..Once electric power
Certain failure occurs for transformer, may will affect the normal operation of entire electric system, it is possible to lead to a large amount of material resources, manpower money
The waste in source, or even cause casualties.Therefore, deep diagnose, in time is carried out to power transformer according to the data monitored
It found the abnormal situation, and effective maintenance measure is taken to power transformer, the safety, reliable to Operation of Electric Systems is ensured
Property has great importance with economy.
Currently, to the research in terms of diagnosing fault of power transformer, there is also some limitations:
(1) in terms of data source, most of researchs are using DGA data, due to the complexity and operation of transformer equipment
The uncertainty of environment cannot completely reflect the operating status of transformer to the information fusion of data mapping.
(2) in terms of research method, the information fusion method that domestic and foreign scholars use at present mainly has artificial neural network, shellfish
Leaf this reasoning, Kalman filtering method, fuzzy theory, DS evidence theory, support vector machines etc., these methods compensate for traditional DGA
The deficiency of method, but one kind of shallow-layer machine learning is belonged to, learning ability is limited, and accuracy is not very high.
(3) in terms of data monitoring, most of at present merged to off-line data, has ignored some on-line monitoring numbers
According to for example, oil chromatography, shelf depreciation, infrared temperature, dielectrical loss etc., to cause real-time diagnosis effect to transformer not
It is good.
In conclusion existing method for diagnosing fault of power transformer problems faced is more, error is larger, with data
The increase of dimension be easy to cause the uncertainty and unstability of data, it is necessary to be improved.
Summary of the invention
It is an object of the invention to for the prior art there are drawback, a kind of electric power based on Multi-source Information Fusion is provided
Diagnosis Method of Transformer Faults, to improve the accuracy of transformer fault diagnosis result.
Problem of the present invention is solved with following technical proposals:
A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion, the method by multiple limited Bohr hereby
Graceful machine and one layer of backpropagation BP neural network are stacked, and constitute deepness belief network model, utilize the transformer of acquisition
Fault data is trained the model, and transformer station high-voltage side bus data are inputted trained deepness belief network model, are obtained
Then the basic probability assignment of various failures recycles DS evidence theory to merge probability of malfunction, obtains transformer diagnosis
Final result.
The above-mentioned method for diagnosing fault of power transformer based on Multi-source Information Fusion, using transformer fault data to depth
The training of belief network model is divided into two steps: the first step is pre-training, i.e., trains each limited glass individually unsupervisedly
The graceful machine of Wurz；Second step is fine tuning, uses BP neural network in the top layer of limited Boltzmann machine network, is limited Boltzmann
The output feature vector of machine network carries out the training for having supervision as the input of BP neural network.
The above-mentioned method for diagnosing fault of power transformer based on Multi-source Information Fusion, using DS evidence theory to probability of malfunction
It is merged method particularly includes: the result of the Fusion Features obtained to DBN network is as probability partition function, according to seemingly
Right function, rule of combination formula obtain final failure decision probability situation.
The above-mentioned method for diagnosing fault of power transformer based on Multi-source Information Fusion, the deepness belief network model are 5
Layer network, input layer are 8 units, respectively the partial discharge quantity of input electric power transformer, casing dielectrical loss factor and
H is dissolved in oil2、C2H2、CH4、C2H6、CO、C2H4Volume fraction；Output layer is 5 Failure probability distribution numbers, respectively corresponds office
Portion's electric discharge, hyperthermia and superheating, middle cryogenic overheating, high-energy discharge and low energy electric discharge.
The present invention combines the deepness belief network in deep learning with DS evidence theory, can be crisscross multiple in transformer
Potential valuable information is extracted in miscellaneous data characteristics, this method compensates for conventional fault diagnosis method not well
Foot, can effectively improve the accuracy of transformer fault diagnosis result, provide guarantee for the safety of electric system.
Detailed description of the invention
The invention will be further described with reference to the accompanying drawing.
Fig. 1 is RBM structure chart；
Fig. 2 is RBM learning process figure；
Fig. 3 is DBN structure chart；
Fig. 4 is power transformer Multi-source Information Fusion model；
Fig. 5 is transformer data prediction flow chart；
Fig. 6 is BPNN compared with DBN-DS arithmetic result.
Specific embodiment
Attached drawing of the invention is illustrated first:
Fig. 1 is RBM structure chart；It is made of hidden layer H and visual layers V double-layer structure, using symmetrical connection and
The input of stochastic neural net model without feedback, data source inputs in visual layers V, is used to carry out data in hidden layer H
The extraction of feature.Each layer is that data are indicated with a vector, and the number of each layer of neuron is exactly with the dimension of vector come body
Existing.Entire RBN intuitively appears to be the form of a bipartite graph as can be seen from Fig., and visual layers V and hidden layer H take
0 or 1 binary variable.It is the relationship being connected entirely, hidden layer neuron only between visual layers neuron and hidden layer neuron
Between or visual layers neuron between be there is no associated relation.
Fig. 2 is RBM learning process figure；The training process of RBM is exactly to find out one to generate the maximum probability of training sample
Distribution, the deciding factor of this sample are weight W, and trained target is exactly to seek this weight, and learning process figure is as schemed
Show.G.Hinton proposes the learning algorithm to sdpecific dispersion (Contrastive Divergence), learns to RBM network
To improve calculating speed and precision.
Fig. 3 is DBN structure chart；Specific training is divided into 2 steps: step 1 is pre-training, and training is each individually unsupervisedly
A RBM network, it is ensured that feature vector keeping characteristics information as far as possible during fusion.Step 2 is fine tuning, in DBN network
Top layer use BP neural network, the output feature vector of RBM carries out the training for having supervision, each layer as its input
RBM network can only ensure that the weight in own layer is optimal this layer of maps feature vectors, be not the feature to entire DBN
DUAL PROBLEMS OF VECTOR MAPPING is optimal, so counterpropagation network also propagates to each layer of RBM for error message is top-down, fine tuning is entire
DBN network.
Fig. 4 is power transformer Multi-source Information Fusion model；The information source of transformer is subjected to pretreatment and is sent into DBN network
It is tentatively merged, the Decision fusion of DS evidence theory is finally carried out using obtained result as probability.
Fig. 5 is transformer data prediction flow chart；The data for monitoring voltage device are pre-processed first, will be handled
Result screening carried out to the data of transformer as the input terminal of DBN network excavated with preliminary, then by the processing result of DBN
It is sent into the fusion of D-S evidence decision, diagnostic result is eventually arrived at according to certain rule, transformer is evaluated.
Fig. 6 is BPNN compared with DBN-DS arithmetic result.Carrying out 10 Experimental comparisons by test can be seen that DBN-
The accuracy rate of DS algorithm is higher than the accuracy rate of BPNN, has been well demonstrated that algorithm proposed by the present invention,
The treatment principle of the method for the present invention:
Firstly, deepness belief network is combined with DS evidence theory；Then, by repetition training network and tuning, make
Algorithm accuracy rate reaches best, finally, the validity and feasibility of proposed method is verified by Simulation Example.
(1) deepness belief network (deep belief network, DBN) is one kind of deep learning model, in addition to this
There are also convolutional neural networks, storehouse self-encoding encoders etc..DBN is a generative probabilistic model, is by multiple limited Boltzmann machines
Deep neural network made of being stacked with one layer of counterpropagation network.
1. limited Boltzmann machine (Restricted Boltzmann Machine, RBM) is a part of DBN, it is
The model based on energy of one unsupervised learning, double-layer structure are hidden layer H and visual layers V respectively, symmetrical connection and nothing
The stochastic neural net model of feedback, visual layers V are data input layer, and hidden layer H is characterized extract layer, and each layer can use one
Vector indicates that the dimension of vector is exactly the number of every layer of neuron.Entire RBN is a bipartite graph, visual layers V and hidden layer H
It is all the binary variable for taking 0 or 1.It is not present and is connected between visual layers neuron between hidden layer neuron, only hidden layer
There is side connection between neuron and visual layers neuron, structure is as shown in Figure 1.
RBM is the model based on energy, given state (v, h), visual layers neuron v and hidden layer neuron h's
Energy joint distribution function are as follows:
In formula, viIndicate the display node of input layer, hiIndicate that the layer of output implies node, wijIndicate input layer and output
The weight of layer connecting node, bjIndicate the offset of output node, ciIndicate the offset of input node.
It can be used to minor function and realize normalized, wherein θ is indicated by bj、ci、wijThe weight collection of composition, Z (θ) are usual
Referred to as partition function.
When the state of given visual element v, the activation primitive of j-th of implicit unit are as follows:
When the state of given implicit unit h, the activation primitive of j-th of visual element are as follows:
Wherein, formula (IV), ρ (x) is sigmoid activation primitive in formula (V)
The training process of RBM is exactly to find out one to generate the maximum probability distribution of training sample, the decision of this sample
Sexual factor is weight W, and trained target is exactly to seek this weight, and learning process figure is as shown in Figure 2.The number of N expression iteration.
G.Hinton proposes the learning algorithm to sdpecific dispersion, is learnt to RBM network to improve calculating speed and precision, herein not
It is described in detail.
2. DBN model is a kind of model being composed of several layers RBM and one layer of BP neural network, as shown in Figure 3.It
Whole network is constructed using the unsupervised learning algorithm of quick greed, and is trained using bottom-up mode, by low layer
Characteristic information be gradually fed to one layer and merged, to obtain stronger robustness.
Specific training is divided into 2 steps: step 1 is pre-training, individually unsupervisedly each RBM network of training, it is ensured that
Feature vector keeping characteristics information as far as possible during fusion.Step 2 is fine tuning, uses BP in the top layer of DBN network
The output feature vector of neural network, RBM carries out the training for having supervision as its input, and each layer of RBM network can only ensure
Weight in own layer is optimal this layer of maps feature vectors, is not to reach most to the maps feature vectors of entire DBN
It is excellent, so counterpropagation network also propagates to each layer of RBM for error message is top-down, finely tune entire DBN network.RBM net
The process of network training pattern can be regarded as making DBN overcome BP network the initialization of one deep layer BP network weight parameter
The deficiency of local optimum and training time length is easily trapped into because of random initializtion weighting parameter.
(2) DS evidence theory initially by Dempster teach in 1967 propose, by its student Shafer to algorithm into
Row improve and perfect ultimately forms a kind of reasoning method under uncertainty.Due to the advantage of DS itself, transformer difference can be come
The information fusion of the information in source, different representation methods becomes significantly more efficient information, and decision-making treatment ability is stronger, wide at present
It is general to apply to uncertain information reasoning, information decision fusion, target identification etc..
DS evidence theory is directed to uncertainty of objective to be processed, introduces probability distribution function, confidence function, likelihood letter
Several methods.By the way that DS evidence theory to be introduced into transformer fault diagnosis, uncertainty caused by various information can solve
Problem improves the robustness of algorithm.
Define a transformer identification framework U={ u1 u2 L un, any one element is all independent from each other, and
Mutual exclusion two-by-two, all elements constitute one and collect the power set for being collectively referred to as U, are denoted as 2U。
U is an identification framework, if m:2U→ [0,1], and meet 2 points following:
First: for the thing that can not occur, probability 0 is expressed as m (Φ)=0.
Second: the probability that all situations occur, and be 1, it is expressed as
M is referred to as the basic probability assignment of proposition A or mass function on U.M (A) indicates the degree of belief to A.
F:2U→[0,1]
Definition: p:2u→ [0,1],
Then p (A) is the likelihood function of A.
WhenWhen, the combining evidences of the upper element of U are as follows:
In formula, K is conflict coefficient, is indicated are as follows:
Conflict spectrum between the size reflection evidence of K value, bigger expression conflict are bigger.It is credible when K value is very big
Degree will be very low, and the effect of fusion results also can be undesirable.
With reference to the accompanying drawing and table data elaborate to the present invention.
Step 1: the present invention takes the typical oil dissolved gas (H of power transformer2、C2H2、CH4、C2H6、CO、C2H4) volume
Score, partial discharge quantity, the pre-processed results of casing dielectrical loss 8 groups of data normalizations of factor are defeated as the data of DBN network
Enter, input is the matrix of n × 8, and n indicates the number of data sample, for determining fault type belonging to transformer.From
The design feature of DBN network, which can be seen that through the study to the network great amount of samples, to be trained with stronger extensive energy
Power so as to preferably describe the Nonlinear Mapping relationship of this complexity, therefore can determine basic probability assignment.Then sharp again
Resulting probability of malfunction is merged with DS evidence theory.This example selects 1200 fault datas of certain city's transformer to do
Experimental verification, 600 are used as training sample, and 600 are used as test sample, wherein the oil chromatography failure of some transformers monitoring
Data are as shown in table 1, and sample data set is as shown in table 2.
1 transformer oil chromatographic fault data of table
2 sample fault data collection of table
Step 2: in this example, the time as used in increase of the DBN algorithm with the number of plies also increases, and selects 5
Layer network is trained.Input layer is 8 units, and the data vector after input transformer data normalization, output layer is 5
Failure probability distribution number, the intermediate number of plies are 30,12 layers respectively.Network settings parameter W, a, b initialization are both configured to 0, learning rate
It is set as 0.1, training error threshold value is set as 0.05, maximum number of iterations 2000.Since the output of DBN network reflects spy
The correlativity in space and fault type is levied, therefore the output of each network can be converted to the probability assignments letter of each failure
Number, it is assumed that OiIt is expressed as the output of i-th of network node, then the apportioning cost for distributing to evidence O calculates by the following method.
In formula, m indicates the number of fault type.
The probability of evidence body distributes, as shown in table 3.
The distribution of 3 probability of table
Step 3: defining a DS identification framework { E1, E2, E3, E4, E5 }, be expressed as low energy electric discharge, middle low temperature mistake
Heat, hyperthermia and superheating, high-energy discharge, shelf depreciation.Transformer fault type is directly judged with maximum membership grade principle by table 3
Inaccuracy, uncertain higher, by DS evidence theory, the results are shown in Table 4 for evidence fusion two-by-two again, it can be seen that no
Certainty has obtained preferable processing, so that failure belongs to hyperthermia and superheating for stronger having determined.
Table 4 merges posterior probability distribution
This experiment is shown by carrying out 10 DBN-DS experiments and BPNN Experimental comparison by DBN and DS evidence theory phase
It is greatly improved in conjunction with the accuracy rate than traditional BP neural network algorithm, as shown in fig. 6, demonstrating side of the present invention
The validity and feasibility of method.
The present invention use by deep learning deepness belief network and DS evidence theory combine, devise a kind of multi-source
Information fusion model makes up the deficiency of conventional fault diagnosis method, can extract in the complicated data characteristics of transformer
Potential valuable information out, and fusion treatment is carried out simultaneously to off-line data and online data, demonstrate this method has
Effect property.
(1) this method has preferable superiority than traditional algorithm, and the uncertainty that can tentatively solve transformer complexity is asked
Topic.
(2) this method has preferable convergence and stability than traditional algorithm.
(3) this method has development potentiality, following in combination with association analysis method, and advanced optimizes deep learning net
Network, so that the result of fault diagnosis is more accurate.
Correlation technique data introduction in the present invention:
(1) Multi-source Information Fusion model
The data of power transformer are complicated and type is various, so as to cause the reason of failure there are certain uncertainty,
Therefore experimental datas various in transformer, such as the gas content in oil chromatogram analysis, oil medium in oiling experimental analysis is hit
Wear voltage, micro-water content, repair history data, running environment data (such as temperature, humidity, pollution level), appearance detection data
It is the pass solved the problems, such as that much informations such as (such as whether oil leak, sound are abnormal, electric discharge flashover situation), which are effectively combined,
Key.Multi-source Information Fusion be exactly by by a variety of monitoring information sources of transformer spatially and temporally on complementation, realize it is superfluous
Remaining information combines according to certain Optimality Criteria, generates and the consistency of transformer fault phenomenon is explained and described.
Current Multi-source Information Fusion model mostly uses 3 layer architectures, is respectively: the fusion of data Layer information, characteristic layer information
Fusion, the fusion of decision-making level's information.The characteristics of present invention is according to Power Transformer Faults are established in conjunction with the method for Multi-source Information Fusion
The model of transformer Multi-source Information Fusion, as shown in Figure 4.
Data prediction is handled transformer data using data normalization method, and Fusion Features are using DBN
Information converged network finally carries out decision level fusion processing with DS evidence theory, to obtain the most termination of transformer diagnosis
Fruit.
(2) data prediction
The monitoring data of power transformer are not only measured greatly, but also have structuring and semi-structured data, are caused data and are deposited
A series of difficulties, the validity of big data such as storage, excavation, analysis need to be broken through.Transformer data have following feature.
1. data mode is various.The data source form of transformer is different, there is factory information data, operation data, online
Monitoring data, fault data etc., data format is also in different poses and with different expressions, includes the diversified forms such as one-dimensional data 2-D data and picture.
2. data have relevance.Although the factor of failure is caused to be present in multiple independent components, will appear
The exception of one component causes other failures, how to solve to be instantly urgently to be resolved ask extremely caused by the relevance of data
Topic.
3. data value density is low.Such as DGA data, it is the data that transformer station high-voltage side bus always exists, most of data
All it is normally, only only a few is abnormal data, and abnormal data is only our main research objects.
Therefore, by being pre-processed to transformer data, it can tentatively solve the above problems, facilitate the efficient of data
Using with efficiency shared, and that data analysis can be improved and excavated, it is more convenient to the research of transformer.It is pair shown in Fig. 5
The flow chart of power transformer data prediction.
The standardized method of data normalization processing is to carry out a series of linear transformation to original transformer data.Assuming that
Min and max is respectively the minimum value and maximum value of attribute M, by an original value x of M by standardization be mapped to section [0,
1] value in, mathematical formulae are as follows: new data=(initial data-minimum)/(maximum-minimum), also referred to as deviation mark
Standardization is the linear transformation to initial data, is mapped to its end value between [0,1].Transfer function is as follows:
In formula, max is the maximum value of sample data, and min is the minimum value of sample data.
The explanation of term used in the present invention
(1) artificial neural network (Artificial Neural Network, ANN), it is artificial since being the 1980s
The research hotspot that smart field rises.It is abstracted human brain neuroid from information processing angle, and it is simple to establish certain
Model is formed different networks by different connection types.
(2) the statistical induction reasoning of Bayesian inference (Bayesian Theory) classics --- estimation and hypothesis testing
On the basis of the new inference method of one kind that grows up.Compared with classical statistical induction inference method, Bayesian inference is being obtained
Out not only will be according to current observed sample information when conclusion, but also to pass by related experience by inference and know
Know.
(3) Kalman filtering method (Kalman Filter) is a kind of utilizes linear system state equation, is inputted by system defeated
Data are observed out, and the algorithm of optimal estimation is carried out to system mode.
(4) fuzzy theory (Fuzzy Theory) refers to the basic conception for having used fuzzy set or continuous subordinating degree function
Theory.
(5) support vector machines (Support Vector Machine, SVM) is a kind of two disaggregated models, basic model
Definition is the maximum linear classifier in interval (when using linear kernel) on feature space, the i.e. learning strategy of support vector machines
It is margin maximization, can be finally converted into the solution of a convex quadratic programming problem.
(6) convolutional neural networks (Convolutional Neural Networks, CNN) are a kind of feedforward neural networks,
Artificial neuron can respond surrounding cells, can carry out large-scale image procossing.
Claims (4)
1. a kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion, characterized in that the method by it is multiple by
Limit Boltzmann machine and one layer of backpropagation BP neural network are stacked, and constitute deepness belief network model, utilize acquisition
Transformer fault data the model is trained, transformer station high-voltage side bus data are inputted into trained deepness belief network mould
Type obtains the basic probability assignment of various failures, then DS evidence theory is recycled to merge probability of malfunction, obtains transformation
The final result of device diagnosis.
2. a kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion according to claim 1, feature
It is that be divided into two steps to the training of deepness belief network model using transformer fault data: the first step is pre-training, i.e., it is single respectively
Each limited Boltzmann machine is solely trained unsupervisedly；Second step is fine tuning, in the top layer of limited Boltzmann machine network
Using BP neural network, the output feature vector of limited Boltzmann machine network has carried out supervision as the input of BP neural network
Training.
3. a kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion according to claim 2, feature
It is to be merged using DS evidence theory to probability of malfunction method particularly includes: to the result for the Fusion Features that DBN network obtains
As probability partition function, final failure decision probability situation is obtained according to likelihood function, rule of combination formula.
4. a kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion according to claim 3, feature
It is that the deepness belief network model is 5 layer networks, and input layer is 8 units, and the part of input electric power transformer is put respectively
H is dissolved in electricity, casing dielectrical loss factor and oil2、C2H2、CH4、C2H6、CO、C2H4Volume fraction；Output layer is 5
Failure probability distribution number respectively corresponds shelf depreciation, hyperthermia and superheating, middle cryogenic overheating, high-energy discharge and low energy electric discharge.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910101086.XA CN109669087A (en) | 2019-01-31 | 2019-01-31 | A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910101086.XA CN109669087A (en) | 2019-01-31 | 2019-01-31 | A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109669087A true CN109669087A (en) | 2019-04-23 |
Family
ID=66151112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910101086.XA Pending CN109669087A (en) | 2019-01-31 | 2019-01-31 | A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109669087A (en) |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110784A (en) * | 2019-04-30 | 2019-08-09 | 贵州电网有限责任公司 | A kind of transformer fault discrimination method based on transformer correlation operation data |
| CN110334865A (en) * | 2019-07-05 | 2019-10-15 | 上海交通大学 | A kind of electrical equipment fault rate prediction technique and system based on convolutional neural networks |
| CN110334740A (en) * | 2019-06-05 | 2019-10-15 | 武汉大学 | The electrical equipment fault of artificial intelligence reasoning fusion detects localization method |
| CN110414552A (en) * | 2019-06-14 | 2019-11-05 | 中国人民解放军海军工程大学 | A kind of spare part reliability Bayesian Assessment Method and system based on multi-source fusion |
| CN110929763A (en) * | 2019-10-30 | 2020-03-27 | 西安交通大学 | Multi-source data fusion-based mechanical fault diagnosis method for medium-voltage vacuum circuit breaker |
| CN111062245A (en) * | 2019-10-31 | 2020-04-24 | 北京交通大学 | Locomotive driver fatigue state monitoring method based on upper body posture |
| CN111090024A (en) * | 2019-11-14 | 2020-05-01 | 国网上海市电力公司 | GIS state evaluation method and device based on external thermal and acoustic characteristic information |
| CN111325233A (en) * | 2019-10-29 | 2020-06-23 | 国网辽宁省电力有限公司电力科学研究院 | Transformer fault detection method and device |
| CN111340118A (en) * | 2020-02-27 | 2020-06-26 | 河南大学 | Conflict evidence fusion method based on reliability entropy and BJS divergence |
| CN111368885A (en) * | 2020-02-24 | 2020-07-03 | 大连理工大学 | Aero-engine gas circuit fault diagnosis method based on deep learning and information fusion |
| CN111426904A (en) * | 2019-10-23 | 2020-07-17 | 合肥申芯电子技术有限责任公司 | Transformer substation grounding grid fault diagnosis method based on depth self-encoder |
| CN111488946A (en) * | 2020-04-28 | 2020-08-04 | 东南大学 | Radar servo system fault diagnosis method based on information fusion |
| CN111539486A (en) * | 2020-05-12 | 2020-08-14 | 国网四川省电力公司电力科学研究院 | Transformer fault diagnosis method based on Dropout deep confidence network |
| CN111650451A (en) * | 2020-04-28 | 2020-09-11 | 中国南方电网有限责任公司超高压输电公司广州局 | Converter transformer fault reason identification method and system |
| CN111693488A (en) * | 2020-06-08 | 2020-09-22 | 济南大学 | Fruit grade classification method and system based on DS evidence theory fusion |
| CN111985820A (en) * | 2020-08-24 | 2020-11-24 | 深圳市加码能源科技有限公司 | FNN and DS fusion-based fault identification method for charging operation management system |
| CN112085084A (en) * | 2020-08-24 | 2020-12-15 | 宁波大学 | Transformer fault diagnosis method based on multi-feature fusion common vector |
| CN112163619A (en) * | 2020-09-27 | 2021-01-01 | 北华大学 | Transformer fault diagnosis method based on two-dimensional tensor |
| CN112307918A (en) * | 2020-10-21 | 2021-02-02 | 华北电力大学 | Diagnosis method for transformer direct-current magnetic biasing based on fuzzy neural network |
| CN112327219A (en) * | 2020-10-29 | 2021-02-05 | 国网福建省电力有限公司南平供电公司 | Distribution transformer fault diagnosis method with automatic feature mining and automatic parameter optimization |
| CN112419243A (en) * | 2020-11-11 | 2021-02-26 | 国网山东省电力公司青岛供电公司 | Power distribution room equipment fault identification method based on infrared image analysis |
| CN112530584A (en) * | 2020-12-15 | 2021-03-19 | 贵州小宝健康科技有限公司 | Medical diagnosis assisting method and system |
| CN112633493A (en) * | 2020-12-01 | 2021-04-09 | 北京理工大学 | Fault diagnosis method and system for industrial equipment data |
| CN112668754A (en) * | 2020-12-03 | 2021-04-16 | 国网山西省电力公司大同供电公司 | Power equipment defect diagnosis method based on multi-source characteristic information fusion |
| CN112684012A (en) * | 2020-12-02 | 2021-04-20 | 青岛科技大学 | Equipment key force-bearing structural part fault diagnosis method based on multi-parameter information fusion |
| CN112733588A (en) * | 2020-08-13 | 2021-04-30 | 精英数智科技股份有限公司 | Machine running state detection method and device and electronic equipment |
| CN112907114A (en) * | 2021-03-18 | 2021-06-04 | 三一重工股份有限公司 | Oil leakage fault detection method and device, electronic equipment and storage medium |
| CN113092899A (en) * | 2021-03-25 | 2021-07-09 | 国网湖南省电力有限公司 | Transformer electrical fault identification method, system, terminal and readable storage medium based on multi-source information fusion |
| CN113326881A (en) * | 2021-05-31 | 2021-08-31 | 西安思安云创科技有限公司 | Power transformer fault diagnosis method |
| CN113640596A (en) * | 2021-07-13 | 2021-11-12 | 中国南方电网有限责任公司超高压输电公司广州局 | Converter transformer abnormity detection method and device, computer equipment and storage medium |
| CN113822421A (en) * | 2021-10-14 | 2021-12-21 | 平安科技（深圳）有限公司 | Neural network based anomaly positioning method, system, equipment and storage medium |
| CN114065814A (en) * | 2021-11-16 | 2022-02-18 | 中国南方电网有限责任公司超高压输电公司广州局 | Method and device for identifying defect types of GIL partial discharge |
| CN114636882A (en) * | 2022-03-24 | 2022-06-17 | 国网江西省电力有限公司电力科学研究院 | Digital twin-based transformer magnetic bias detection system and method |
| CN115713027A (en) * | 2022-10-31 | 2023-02-24 | 国网江苏省电力有限公司泰州供电分公司 | Transformer state evaluation method, device and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108196143A (en) * | 2017-12-11 | 2018-06-22 | 囯网河北省电力有限公司电力科学研究院 | Power Transformer Faults depth diagnostic method and terminal device |
| CN109088407A (en) * | 2018-08-06 | 2018-12-25 | 河海大学 | The State Estimation for Distribution Network of modeling is measured based on deepness belief network puppet |
| CN109086817A (en) * | 2018-07-25 | 2018-12-25 | 西安工程大学 | A kind of Fault Diagnosis for HV Circuit Breakers method based on deepness belief network |
| CN109214416A (en) * | 2018-07-23 | 2019-01-15 | 华南理工大学 | A kind of multidimensional information fusion Diagnosis Method of Transformer Faults based on deep learning |
-
2019
- 2019-01-31 CN CN201910101086.XA patent/CN109669087A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108196143A (en) * | 2017-12-11 | 2018-06-22 | 囯网河北省电力有限公司电力科学研究院 | Power Transformer Faults depth diagnostic method and terminal device |
| CN109214416A (en) * | 2018-07-23 | 2019-01-15 | 华南理工大学 | A kind of multidimensional information fusion Diagnosis Method of Transformer Faults based on deep learning |
| CN109086817A (en) * | 2018-07-25 | 2018-12-25 | 西安工程大学 | A kind of Fault Diagnosis for HV Circuit Breakers method based on deepness belief network |
| CN109088407A (en) * | 2018-08-06 | 2018-12-25 | 河海大学 | The State Estimation for Distribution Network of modeling is measured based on deepness belief network puppet |
Non-Patent Citations (1)
| Title |
|---|
| 石鑫: "基于深度信念网络的电力变压器故障分类建模", 《电力系统保护与控制》 * |
Cited By (50)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110784A (en) * | 2019-04-30 | 2019-08-09 | 贵州电网有限责任公司 | A kind of transformer fault discrimination method based on transformer correlation operation data |
| 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 |
| CN110334740A (en) * | 2019-06-05 | 2019-10-15 | 武汉大学 | The electrical equipment fault of artificial intelligence reasoning fusion detects localization method |
| CN110414552B (en) * | 2019-06-14 | 2021-07-16 | 中国人民解放军海军工程大学 | Bayesian evaluation method and system for spare part reliability based on multi-source fusion |
| CN110414552A (en) * | 2019-06-14 | 2019-11-05 | 中国人民解放军海军工程大学 | A kind of spare part reliability Bayesian Assessment Method and system based on multi-source fusion |
| CN110334865A (en) * | 2019-07-05 | 2019-10-15 | 上海交通大学 | A kind of electrical equipment fault rate prediction technique and system based on convolutional neural networks |
| CN110334865B (en) * | 2019-07-05 | 2023-04-18 | 上海交通大学 | Power equipment fault rate prediction method and system based on convolutional neural network |
| CN111426904A (en) * | 2019-10-23 | 2020-07-17 | 合肥申芯电子技术有限责任公司 | Transformer substation grounding grid fault diagnosis method based on depth self-encoder |
| CN111325233B (en) * | 2019-10-29 | 2024-02-09 | 国网辽宁省电力有限公司电力科学研究院 | Transformer fault detection method and device |
| CN111325233A (en) * | 2019-10-29 | 2020-06-23 | 国网辽宁省电力有限公司电力科学研究院 | Transformer fault detection method and device |
| CN110929763A (en) * | 2019-10-30 | 2020-03-27 | 西安交通大学 | Multi-source data fusion-based mechanical fault diagnosis method for medium-voltage vacuum circuit breaker |
| CN110929763B (en) * | 2019-10-30 | 2022-06-21 | 西安交通大学 | Multi-source data fusion-based mechanical fault diagnosis method for medium-voltage vacuum circuit breaker |
| CN111062245A (en) * | 2019-10-31 | 2020-04-24 | 北京交通大学 | Locomotive driver fatigue state monitoring method based on upper body posture |
| CN111090024A (en) * | 2019-11-14 | 2020-05-01 | 国网上海市电力公司 | GIS state evaluation method and device based on external thermal and acoustic characteristic information |
| CN111090024B (en) * | 2019-11-14 | 2021-11-12 | 国网上海市电力公司 | GIS state evaluation method and device based on external thermal and acoustic characteristic information |
| CN111368885A (en) * | 2020-02-24 | 2020-07-03 | 大连理工大学 | Aero-engine gas circuit fault diagnosis method based on deep learning and information fusion |
| CN111368885B (en) * | 2020-02-24 | 2021-12-03 | 大连理工大学 | Gas circuit fault diagnosis method for aircraft engine |
| CN111340118B (en) * | 2020-02-27 | 2021-07-23 | 河南大学 | Conflict evidence fusion method based on reliability entropy and BJS divergence |
| CN111340118A (en) * | 2020-02-27 | 2020-06-26 | 河南大学 | Conflict evidence fusion method based on reliability entropy and BJS divergence |
| CN111488946A (en) * | 2020-04-28 | 2020-08-04 | 东南大学 | Radar servo system fault diagnosis method based on information fusion |
| CN111650451A (en) * | 2020-04-28 | 2020-09-11 | 中国南方电网有限责任公司超高压输电公司广州局 | Converter transformer fault reason identification method and system |
| CN111650451B (en) * | 2020-04-28 | 2021-12-31 | 中国南方电网有限责任公司超高压输电公司广州局 | Converter transformer fault reason identification method and system |
| CN111539486A (en) * | 2020-05-12 | 2020-08-14 | 国网四川省电力公司电力科学研究院 | Transformer fault diagnosis method based on Dropout deep confidence network |
| CN111693488B (en) * | 2020-06-08 | 2022-12-06 | 济南大学 | Fruit grade classification method and system based on DS evidence theory fusion |
| CN111693488A (en) * | 2020-06-08 | 2020-09-22 | 济南大学 | Fruit grade classification method and system based on DS evidence theory fusion |
| CN112733588A (en) * | 2020-08-13 | 2021-04-30 | 精英数智科技股份有限公司 | Machine running state detection method and device and electronic equipment |
| CN112085084B (en) * | 2020-08-24 | 2023-12-15 | 宁波大学 | Transformer fault diagnosis method based on multi-feature fusion common vector |
| CN112085084A (en) * | 2020-08-24 | 2020-12-15 | 宁波大学 | Transformer fault diagnosis method based on multi-feature fusion common vector |
| CN111985820A (en) * | 2020-08-24 | 2020-11-24 | 深圳市加码能源科技有限公司 | FNN and DS fusion-based fault identification method for charging operation management system |
| CN112163619A (en) * | 2020-09-27 | 2021-01-01 | 北华大学 | Transformer fault diagnosis method based on two-dimensional tensor |
| CN112307918A (en) * | 2020-10-21 | 2021-02-02 | 华北电力大学 | Diagnosis method for transformer direct-current magnetic biasing based on fuzzy neural network |
| CN112327219B (en) * | 2020-10-29 | 2024-03-12 | 国网福建省电力有限公司南平供电公司 | Distribution transformer fault diagnosis method with automatic feature mining and parameter automatic optimizing functions |
| CN112327219A (en) * | 2020-10-29 | 2021-02-05 | 国网福建省电力有限公司南平供电公司 | Distribution transformer fault diagnosis method with automatic feature mining and automatic parameter optimization |
| CN112419243B (en) * | 2020-11-11 | 2023-08-11 | 国网山东省电力公司青岛供电公司 | Power distribution room equipment fault identification method based on infrared image analysis |
| CN112419243A (en) * | 2020-11-11 | 2021-02-26 | 国网山东省电力公司青岛供电公司 | Power distribution room equipment fault identification method based on infrared image analysis |
| CN112633493A (en) * | 2020-12-01 | 2021-04-09 | 北京理工大学 | Fault diagnosis method and system for industrial equipment data |
| CN112684012A (en) * | 2020-12-02 | 2021-04-20 | 青岛科技大学 | Equipment key force-bearing structural part fault diagnosis method based on multi-parameter information fusion |
| CN112668754A (en) * | 2020-12-03 | 2021-04-16 | 国网山西省电力公司大同供电公司 | Power equipment defect diagnosis method based on multi-source characteristic information fusion |
| CN112530584A (en) * | 2020-12-15 | 2021-03-19 | 贵州小宝健康科技有限公司 | Medical diagnosis assisting method and system |
| CN112907114A (en) * | 2021-03-18 | 2021-06-04 | 三一重工股份有限公司 | Oil leakage fault detection method and device, electronic equipment and storage medium |
| CN113092899A (en) * | 2021-03-25 | 2021-07-09 | 国网湖南省电力有限公司 | Transformer electrical fault identification method, system, terminal and readable storage medium based on multi-source information fusion |
| CN113092899B (en) * | 2021-03-25 | 2022-06-10 | 国网湖南省电力有限公司 | Transformer electrical fault identification method, system, terminal and readable storage medium |
| CN113326881B (en) * | 2021-05-31 | 2023-02-14 | 西安思安云创科技有限公司 | Power transformer fault diagnosis method |
| CN113326881A (en) * | 2021-05-31 | 2021-08-31 | 西安思安云创科技有限公司 | Power transformer fault diagnosis method |
| CN113640596A (en) * | 2021-07-13 | 2021-11-12 | 中国南方电网有限责任公司超高压输电公司广州局 | Converter transformer abnormity detection method and device, computer equipment and storage medium |
| CN113822421A (en) * | 2021-10-14 | 2021-12-21 | 平安科技（深圳）有限公司 | Neural network based anomaly positioning method, system, equipment and storage medium |
| CN113822421B (en) * | 2021-10-14 | 2024-05-14 | 平安科技（深圳）有限公司 | Neural network-based anomaly locating method, system, equipment and storage medium |
| CN114065814A (en) * | 2021-11-16 | 2022-02-18 | 中国南方电网有限责任公司超高压输电公司广州局 | Method and device for identifying defect types of GIL partial discharge |
| CN114636882A (en) * | 2022-03-24 | 2022-06-17 | 国网江西省电力有限公司电力科学研究院 | Digital twin-based transformer magnetic bias detection system and method |
| CN115713027A (en) * | 2022-10-31 | 2023-02-24 | 国网江苏省电力有限公司泰州供电分公司 | Transformer state evaluation method, device and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109669087A (en) | A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion | |
| CN106251059B (en) | Cable state evaluation method based on probabilistic neural network algorithm | |
| CN110007652A (en) | A kind of Hydropower Unit degradation trend interval prediction method and system | |
| CN109635928A (en) | A kind of voltage sag reason recognition methods based on deep learning Model Fusion | |
| CN109543901A (en) | Short-Term Load Forecasting Method based on information fusion convolutional neural networks model | |
| Wan et al. | Day-ahead prediction of wind speed with deep feature learning | |
| CN104155574A (en) | Power distribution network fault classification method based on adaptive neuro-fuzzy inference system | |
| Ma et al. | Multisensor decision approach for HVCB fault detection based on the vibration information | |
| Ali et al. | COMSATS University Islamabad | |
| Ren et al. | An interpretable deep learning method for power system transient stability assessment via tree regularization | |
| Li et al. | Adaptive assessment of power system transient stability based on active transfer learning with deep belief network | |
| Zhang et al. | Fault diagnosis based on non-negative sparse constrained deep neural networks and Dempster–Shafer theory | |
| CN114661905A (en) | Power grid fault diagnosis method based on BERT | |
| CN116205265A (en) | Power grid fault diagnosis method and device based on deep neural network | |
| CN116245033A (en) | Artificial intelligent driven power system analysis method and intelligent software platform | |
| Dai et al. | Fault Diagnosis of Data‐Driven Photovoltaic Power Generation System Based on Deep Reinforcement Learning | |
| Xu et al. | An improved ELM-WOA–based fault diagnosis for electric power | |
| Ichisugi | A cerebral cortex model that self-organizes conditional probability tables and executes belief propagation | |
| Zhang et al. | Partition fault diagnosis of power grids based on improved PNN and GRA | |
| Du et al. | A hierarchical power system transient stability assessment method considering sample imbalance | |
| CN112163474A (en) | Intelligent gearbox diagnosis method based on model fusion | |
| Yangyong et al. | Modeling of false information on microblog with block matching and fuzzy neural network | |
| CN114707613B (en) | Layered depth strategy gradient network-based power grid regulation and control method | |
| Wang et al. | Cluster division in wind farm through ensemble modelling | |
| CN110109005A (en) | A kind of analog circuit fault test method based on sequential test |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190423 |