CN110334866A - Consider the electrical equipment fault probability forecasting method and system of insulation defect classification and fault correlation - Google Patents
Consider the electrical equipment fault probability forecasting method and system of insulation defect classification and fault correlation Download PDFInfo
- Publication number
- CN110334866A CN110334866A CN201910602683.0A CN201910602683A CN110334866A CN 110334866 A CN110334866 A CN 110334866A CN 201910602683 A CN201910602683 A CN 201910602683A CN 110334866 A CN110334866 A CN 110334866A
- Authority
- CN
- China
- Prior art keywords
- probability
- power equipment
- neural networks
- insulation defect
- electrical equipment
- 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.)
- Granted
Links
Classifications
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention discloses a kind of electrical equipment fault probability forecasting methods for considering insulation defect classification and fault correlation comprising step: (1) acquiring the PRPS spectrum data of power equipment and pre-process to it；(2) based on by pretreated PRPS spectrum data extraction local discharge characteristic；(3) local discharge characteristic is inputted into trained convolutional neural networks, trained convolutional neural networks output power equipment has the probability value P (D of certain class insulation defectk)；And local discharge characteristic is also inputted into trained length Memory Neural Networks in short-term, trained length in short-term Memory Neural Networks output power equipment in DkUnder conditions of break down probability P (F | Dk)；(4) the final probability of malfunction P (F) of power equipment is obtained based on following formula:In addition, the invention also discloses a kind of electrical equipment fault probabilistic forecasting systems.
Description
Technical field
The present invention relates to the failure prediction methods and system more particularly to a kind of electrical equipment fault probability in electric system
Prediction technique and system.
Background technique
Gas insulated combined electrical equipment (GIS) is widely used in electric system, and internal structure is complicated, after breaking down
It is big to overhaul difficulty, be easy to cause heavy losses.The failure cause multiplicity of GIS device, mainly based on insulation defect, and with office
Portion's electric discharge.The severity of its defect is analyzed according to local discharge signal, and carries out risk assessment, is carried out further according to assessment result
Reasonable Strategies of Maintenance is formulated in effective early warning, is of great significance to the safe and reliable operation for guaranteeing equipment.Risk evaluation model
It is the model that join probability carries out forecast assessment to consequence issuable after equipment fault.Risk Calculation be equipment failure rate with
The product of equipment fault consequence.
With the popularization and application of partial discharge electrification detection technology, substation field produces a large amount of Partial Discharge Detection number
According to.Using the detection data that big data platform accumulates, equipment Risk is more accurately assessed.Currently both at home and abroad to risk
The main direction of studying of assessment includes constructing reasonable assessment models and calculating GIS device failure rate.However, being directed to GIS device
The research for carrying out risk assessment or probability of malfunction calculating is less.Calculation method to power transmission and transformation equipment failure rate includes probability artwork
Type, by assuming that probability of malfunction meets between certain distribution characteristics or quantity of state, there are certain associations to shift, utilizes experimental data
Carry out modeling and forecasting etc..
Based on this, it is expected that obtaining a kind of electrical equipment fault probabilistic forecasting for considering insulation defect classification and fault correlation
The relevance of insulation defect classification and equipment fault can be considered in method, this method, so that more accurate probability of malfunction is obtained,
Suitable for the engineer application under big data platform.
Summary of the invention
An object of the present invention is to provide a kind of electrical equipment fault for considering insulation defect classification and fault correlation
Being associated with for insulation defect classification and equipment fault can be considered in probability forecasting method, the electrical equipment fault probability forecasting method
Property, by then being played a game using convolutional neural networks to its local discharge characteristic is extracted by pretreated PRPS spectrum data
Portion's discharge characteristic carries out defect classification, using the conditional failure rate of the long prediction of Memory Neural Networks in short-term power equipment, and according to
The final failure rate of conditional failure rate acquisition power equipment.The electrical equipment fault probability forecasting method based on data due to being driven
It is dynamic, artificial hypothesis Failure probability distribution or structural regime amount relevance are avoided, and due to direct according to neural network model
Probability of malfunction is obtained, thus, there is certain robustness.In addition, the electrical equipment fault probability forecasting method can be for difference
Shelf depreciation classification carries out probability of malfunction prediction, improves probabilistic forecasting effect, is highly suitable for the risk assessment of GIS device.
According to foregoing invention purpose, the present invention proposes a kind of power equipment for considering insulation defect classification and fault correlation
Probability of malfunction prediction technique comprising step:
(1) it acquires the PRPS spectrum data of power equipment and it is pre-processed；
(2) based on by pretreated PRPS spectrum data extraction local discharge characteristic；
(3) local discharge characteristic is inputted into trained convolutional neural networks, trained convolutional neural networks are defeated
Power equipment has the probability value P (D of certain class insulation defect outk)；And local discharge characteristic is also inputted into trained length
Short-term memory neural network, trained length in short-term Memory Neural Networks output power equipment in DkUnder conditions of break down
Probability P (F | Dk)；
(4) the final probability of malfunction P (F) of power equipment is obtained based on following formula:
Wherein, n indicates the insulation defect number of types of power equipment, k=1,2,3 ... n；P(Dk) indicate convolutional Neural
The power equipment of network output is the probability value of certain class insulation defect, DkIndicate n class insulation defect；P(F|Dk) indicate that length is remembered in short-term
The power equipment of neural network output is recalled in DkUnder the conditions of the probability that breaks down.
In the electrical equipment fault probability forecasting method of consideration insulation defect classification and fault correlation of the present invention
In, since deep learning has the characteristics that feature extraction, Data Dimensionality Reduction, mark sheet can be obtained by autonomous learning from sample data
It reaches, therefore, can be widely used.Wherein, convolutional neural networks (Convolutional Neural Networks, CNN) by
In it is locally connected, weight is shared, merges the feature in pond, therefore, in the present case using trained convolutional neural networks
Output power equipment has the probability value P (D of certain class insulation defectk).Further, since long Memory Neural Networks (Long in short-term
Short Term Memory, LSTM) network characteristics can learn long-term dependence, therefore, in the present case, pass through training
Length in short-term Memory Neural Networks output power equipment in DkUnder conditions of break down probability P (F | Dk).Different defects generate
Failure rate it is different, therefore, in technical solutions according to the invention, in conjunction with the probability of different insulative defect, utilize probability public
Formula obtains final failure rate.
Electrical equipment fault probability forecasting method of the present invention avoids artificial hypothesis event due to being based on data-driven
Hinder probability distribution or structural regime amount relevance, and due to directly obtaining probability of malfunction according to neural network model, thus, tool
There is certain robustness.
In addition, electrical equipment fault probability forecasting method of the present invention can be carried out for different shelf depreciation classifications
Probability of malfunction prediction, improves probabilistic forecasting effect, is highly suitable for the risk assessment of GIS device.
Further, in electrical equipment fault probability forecasting method of the present invention, in step (1), pretreatment
Linear normalization is carried out including at least to PRPS spectrum data.
Further, in electrical equipment fault probability forecasting method of the present invention, using unsupervised network model
Self-encoding encoder extracts local discharge characteristic.
In above scheme, unsupervised network model self-encoding encoder can carry out pretreatment to PRPS spectrum data and extract part
Discharge characteristic, such as self-encoding encoder hidden layer characteristic are 64, then final extract obtains 64 local discharge characteristics.
Further, in electrical equipment fault probability forecasting method of the present invention, power equipment includes at least GIS
Equipment, and the insulation defect number of types n=4 of GIS device.
In above scheme, it is contemplated that electrical equipment fault is influenced by factors, such as partial discharge position, overvoltage
Correlation, shelf depreciation development will cause influence with temporal correlation and shelf depreciation insulation defect type, and wherein, office
Portion's electric discharge insulation defect type is affected to probability of malfunction, it is therefore preferred that using the insulation defect type of GIS device, GIS
The insulation defect type of equipment includes: suspended discharge, the electric discharge of insulation class, tip corona and four kinds of particulate electric discharge.
Further, in electrical equipment fault probability forecasting method of the present invention, convolutional neural networks include 1
Input layer, 4 convolutional layers, 4 pond layers, 1 full articulamentum and 1 output category layer.
Further, in electrical equipment fault probability forecasting method of the present invention, convolutional neural networks are carried out
Training includes: that acquisition training sample set data are trained, and using cross entropy cost function, is updated using stochastic gradient descent method
Model parameter；Supervision fine tuning, Optimized model parameter have been carried out by back-propagation algorithm；Wherein activation primitive is using unsaturated non-
Linear function；Output category layer uses Softmax classifier.
Further, in electrical equipment fault probability forecasting method of the present invention, to long short-term memory nerve net
It includes: that acquisition training sample set data are trained that network, which is trained, is finely adjusted using back-propagation algorithm, Optimized model ginseng
Number.
Correspondingly, another object of the present invention is to provide a kind of electric power for considering insulation defect classification and fault correlation
Insulation defect classification and equipment can be considered by the electrical equipment fault probabilistic forecasting system in probability of equipment failure forecasting system
The relevance of failure is highly suitable for the risk assessment of GIS device to obtain ideal probabilistic forecasting effect.
According to foregoing invention purpose, the present invention proposes a kind of power equipment for considering insulation defect classification and fault correlation
Probability of malfunction forecasting system comprising:
Data acquisition device acquires the PRPS spectrum data of power equipment and pre-processes to it；
Unsupervised network model self-encoding encoder, it is special based on shelf depreciation is extracted by pretreated PRPS spectrum data
Sign；
Probability of malfunction prediction module comprising convolutional neural networks module, long Memory Neural Networks module and failure in short-term
Probability output module, wherein convolutional neural networks module receives the local discharge characteristic of input, and output power equipment has certain class
Probability value P (the D of insulation defectk)；Long Memory Neural Networks module in short-term receives input local discharge characteristic, output power equipment
In DkUnder conditions of break down probability P (F | Dk)；The probability of malfunction module based on following formula output power equipment most
Whole probability of malfunction P (F):
Wherein, n indicates the insulation defect number of types of power equipment, k=1,2,3 ... n；P(Dk) indicate convolutional Neural
The power equipment of network output is the probability value of certain class insulation defect, DkIndicate n class insulation defect；P(F|Dk) indicate that length is remembered in short-term
The power equipment of neural network output is recalled in DkUnder the conditions of the probability that breaks down.
Further, in electrical equipment fault probabilistic forecasting system of the present invention, power equipment includes at least GIS
Equipment, and the insulation defect number of types n=4 of GIS device.
Further, in electrical equipment fault probabilistic forecasting system of the present invention, convolutional neural networks module packet
Include 1 input layer, 4 convolutional layers, 4 pond layers, 1 full articulamentum and 1 output category layer.
The of the present invention electrical equipment fault probability forecasting method for considering insulation defect type and fault correlation and
System have the advantages described below and the utility model has the advantages that
Electrical equipment fault probability forecasting method of the present invention avoids artificial hypothesis event due to being based on data-driven
Hinder probability distribution or structural regime amount relevance, and due to directly obtaining probability of malfunction according to neural network model, thus, tool
There is certain robustness.
In addition, electrical equipment fault probability forecasting method of the present invention can be carried out for different shelf depreciation classifications
Probability of malfunction prediction, improves probabilistic forecasting effect, is highly suitable for the risk assessment of GIS device.
In addition, electrical equipment fault probabilistic forecasting system of the present invention also has above advantages and beneficial to effect
Fruit.
Detailed description of the invention
Fig. 1 is the electrical equipment fault probabilistic forecasting side of consideration insulation defect classification and fault correlation of the present invention
The schematic illustration of method.
Fig. 2 schematically shows the power equipment event for considering insulation defect classification and fault correlation of the present invention
Hinder probability forecasting method in some embodiments to the training process schematic diagram of convolutional neural networks.
Fig. 3 schematically shows the power equipment event for considering insulation defect classification and fault correlation of the present invention
Hinder probability system in some embodiments to the training process schematic diagram of long Memory Neural Networks in short-term.
Specific embodiment
Below in conjunction with Figure of description and specific embodiment to consideration insulation defect classification and event of the present invention
The electrical equipment fault probability forecasting method and system for hindering relevance make further explanation, however the explanation and illustration
Improper restriction is not constituted to technical solution of the present invention.
Fig. 1 is the electrical equipment fault probabilistic forecasting side of consideration insulation defect classification and fault correlation of the present invention
The schematic illustration of method.Fig. 2 schematically shows the electricity of consideration insulation defect classification and fault correlation of the present invention
Power probability of equipment failure prediction technique is in some embodiments to the training process schematic diagram of convolutional neural networks.Fig. 3 signal
Property show the of the present invention electrical equipment fault probability system for considering insulation defect classification and fault correlation one
To the training process schematic diagram of long Memory Neural Networks in short-term in a little embodiments.
As shown in Figure 1, simultaneously if necessary referring to figs. 2 and 3, the PRPS according to the i of collected power equipment months schemes
Modal data is pre-processed, and since PRPS map indicates local discharge characteristic with two-dimensional matrix, and the two-dimensional matrix two are tieed up
Degree respectively indicates shelf depreciation phase and period, and the numerical value of each point indicates electric discharge amplitude.In the present embodiment, it can use
PRPS spectrum data of the superfrequency detection system as data collection system acquisition power equipment, 360 degree of acquisition phase, phase point
5 degree of resolution, 50 power frequency periods are acquired altogether, obtain the two-dimensional matrix that format is 72 × 50.Then PRPS spectrum data is carried out
Linear normalization, formula are as follows:
In formula: xminIndicate minimum value in two-dimensional matrix, xmaxIndicate maximum value in two-dimensional matrix, the original in x representing matrix
Beginning data, x ' indicate the data after linear normalization.
It should be noted that whether failure belongs to two classification problems to equipment, therefore, certain following a period of time of equipment is defined
(such as one month) breaks down, then otherwise it is 0 that the device class situation, which is 1,.Therefore, two classification situation drags are exportable
Classification results are 1 probability, i.e. required probability of equipment failure.
Then, it is based on extracting local put by pretreated PRPS spectrum data by unsupervised network model self-encoding encoder
Electrical feature, it is contemplated that power equipment is mainly affected by the shelf depreciation insulation defect type of GIS device, therefore, at this
In case, the power equipment of use includes at least GIS device, and the insulation defect number of types n=4 of GIS device, and above-mentioned scarce
It falls into including suspended discharge, the electric discharge of insulation class, tip corona and totally four kinds of particulate electric discharge.
Wherein, by pretreatment obtain PRPS spectrum data totally 14280,11424 be used as training set data, 2856
For item as test set data, self-encoding encoder hidden layer characteristic is 64, and therefore, final extract obtains 64 local discharge characteristics.
Construct convolutional neural networks, convolutional neural networks include 1 input layer, 4 convolutional layers, 4 pond layers, 1 entirely
Articulamentum and 1 output category layer.
Random initializtion convolutional neural networks carry out study instruction to training set data using the volume machine neural network of building
Practice, using cross entropy cost function, updates model parameter using stochastic gradient descent method.Prison has been carried out by back-propagation algorithm
Superintend and direct fine tuning, Optimized model parameter.Wherein, convolution kernel size is 3 × 3 in model, and activation primitive is using unsaturated non-linear letter
Number f (x)=max (0, x).The size of pond layer is 1 × 2, using maximum down-sampled method.The neuron number of articulamentum is entirely
1024.Output layer selects the Softmax classifier for being adapted to non-linear more classification problems.Local discharge characteristic is inputted by instruction
Experienced convolutional neural networks, trained convolutional neural networks output power equipment have the probability value P of certain class insulation defect
(Dk)。
It is above-mentioned that Fig. 2 can be referred to the training process of convolutional neural networks.As shown in Fig. 2, passing through collection site band electric-examination
Measured data obtains the PRPS spectrum data of power equipment, then carries out linear normalization pretreatment, then passes through unsupervised network
Model self-encoding encoder is based on extracting local discharge characteristic by pretreated PRPS spectrum data, to the convolutional neural networks of building
It is trained, then obtains the volume machine neural network model of training completion, local discharge characteristic is inputted into trained convolution
Neural network, trained convolutional neural networks output power equipment have the probability value P (D of certain class insulation defectk)。
And about can be with specific reference to Fig. 3 to the long building of Memory Neural Networks in short-term, training process.Wherein, for instruction
Practice the acquisition process of data and test data as described above, it passes through the power equipment PRPS spectrum data warp collected
Property normalization pretreatment obtain, be then based on by unsupervised network model self-encoding encoder by pretreated PRPS spectrum data
Extract local discharge characteristic.
Being trained to long Memory Neural Networks in short-term includes: that acquisition training sample set data are trained, using reversed
Propagation algorithm is finely adjusted, Optimized model parameter.Specifically, using i months local discharge characteristics, to four kinds of insulation defects
Probability corresponding to type is predicted to get to four corresponding Conditional failure probabilities.Conditional failure rate refers to that equipment is belonging to
Under the precondition of certain defect, the probability of possible breakdown.Probability of malfunction solution can regard two classification problem of failure as.It is to equipment
No failure carries out two classification, obtains the probability that output category result is 1, the i.e. condition fault of equipment using softmax classifier
Probability.Specifically:
Long memory network model in short-term is constructed, altogether 30 hidden layers, Hidden unit 32.Using the model of building to training
Collect data and carry out learning training, be finely adjusted using back-propagation algorithm, and can be by minimizing error Optimized model ginseng
Number.The label of training data is 1 and 0, wherein certain following a period of time (such as one month) interior device fails of 1 expression, and 0
Indicate that there is no failures for certain following a period of time (such as one month) interior equipment of equipment.
The training set data inputted in long Memory Neural Networks in short-term is to extract obtained local discharge characteristic data composition
Time series, the characteristic quantity of every month is 64, and training dataset can come from substation field shelf depreciation live detection,
4272 sample datas altogether；
Final probability of malfunction P (F) based on following formula output power equipment:
Wherein, the insulation defect number of types of n expression power equipment, k=1,2,3 ... n, in the present embodiment, n=
4；P(Dk) indicate that the power equipment of convolutional neural networks output is the probability value of certain class insulation defect, DkIndicate n class insulation defect；
P(F|Dk) indicate the power equipment of the long output of Memory Neural Networks in short-term in DkUnder the conditions of the probability that breaks down.
In addition, in several embodiments it is contemplated that the electrical equipment fault of insulation defect classification and fault correlation is predicted
System includes:
Data acquisition device acquires the PRPS spectrum data of power equipment and pre-processes to it；
Unsupervised network model self-encoding encoder, it is special based on shelf depreciation is extracted by pretreated PRPS spectrum data
Sign；
Probability of malfunction prediction module comprising convolutional neural networks module, long Memory Neural Networks module and failure in short-term
Probability output module, wherein convolutional neural networks module receives the local discharge characteristic of input, and output power equipment has certain class
Probability value P (the D of insulation defectk)；Long Memory Neural Networks module in short-term receives input local discharge characteristic, output power equipment
In DkUnder conditions of break down probability P (F | Dk)；The probability of malfunction module based on following formula output power equipment most
Whole probability of malfunction P (F):
Wherein, n indicates the insulation defect number of types of power equipment, k=1,2,3 ... n；P(Dk) indicate convolutional Neural
The power equipment of network output is the probability value of certain class insulation defect, DkIndicate n class insulation defect；P(F|Dk) indicate that length is remembered in short-term
The power equipment of neural network output is recalled in DkUnder the conditions of the probability that breaks down.
Find out in summary, electrical equipment fault probability forecasting method of the present invention is kept away due to being based on data-driven
Artificial hypothesis Failure probability distribution or structural regime amount relevance are exempted from, and due to directly obtaining event according to neural network model
Hinder probability, thus, there is certain robustness.
In addition, electrical equipment fault probability forecasting method of the present invention can be carried out for different shelf depreciation classifications
Probability of malfunction prediction, improves probabilistic forecasting effect, is highly suitable for the risk assessment of GIS device.
In addition, electrical equipment fault probabilistic forecasting system of the present invention also has above advantages and beneficial to effect
Fruit.
It should be noted that prior art part is not limited to given by present specification in protection scope of the present invention
Embodiment, all prior arts not contradicted with the solution of the present invention, including but not limited to first patent document, formerly
Public publication, formerly openly use etc., it can all be included in protection scope of the present invention.
In addition, in this case in the combination of each technical characteristic and unlimited this case claim documented combination or
It is combination documented by specific embodiment, all technical characteristics that this case is recorded can be freely combined in any way
Or combine, unless generating contradiction between each other.
It is also to be noted that embodiment enumerated above is only specific embodiments of the present invention.The obvious present invention is not
Above embodiments are confined to, the similar variation or deformation made therewith are that those skilled in the art can be from present disclosure
It immediately arrives at or is easy to just to associate, be within the scope of protection of the invention.
Claims (10)
1. a kind of electrical equipment fault probability forecasting method for considering insulation defect classification and fault correlation, which is characterized in that
Comprising steps of
(1) it acquires the PRPS spectrum data of power equipment and it is pre-processed；
(2) based on by pretreated PRPS spectrum data extraction local discharge characteristic；
(3) local discharge characteristic is inputted into trained convolutional neural networks, trained convolutional neural networks output electricity
Power equipment has the probability value P (D of certain class insulation defectk)；And local discharge characteristic is also inputted into trained length in short-term
Memory Neural Networks, trained length in short-term Memory Neural Networks output power equipment in DkUnder conditions of break down it is general
Rate P (F | Dk)；
(4) the final probability of malfunction P (F) of power equipment is obtained based on following formula:
Wherein, n indicates the insulation defect number of types of power equipment, k=1,2,3 ... n；P(Dk) indicate that convolutional neural networks are defeated
Power equipment out is the probability value of certain class insulation defect, DkIndicate n class insulation defect；P(F|Dk) indicate long short-term memory nerve
The power equipment of network output is in DkUnder the conditions of the probability that breaks down.
2. electrical equipment fault probability forecasting method as described in claim 1, which is characterized in that described pre- in step (1)
Processing, which is included at least, carries out linear normalization to PRPS spectrum data.
3. electrical equipment fault probability forecasting method as described in claim 1, which is characterized in that use unsupervised network model
Self-encoding encoder extracts local discharge characteristic.
4. electrical equipment fault probability forecasting method as described in claim 1, which is characterized in that the power equipment at least wraps
Include GIS device, and the insulation defect number of types n=4 of GIS device.
5. electrical equipment fault probability forecasting method as described in claim 1, which is characterized in that the convolutional neural networks packet
Include 1 input layer, 4 convolutional layers, 4 pond layers, 1 full articulamentum and 1 output category layer.
6. electrical equipment fault probability forecasting method as claimed in claim 5, which is characterized in that carried out to convolutional neural networks
Training includes: that acquisition training sample set data are trained, and using cross entropy cost function, is updated using stochastic gradient descent method
Model parameter；Supervision fine tuning, Optimized model parameter have been carried out by back-propagation algorithm；Wherein activation primitive is using unsaturated non-
Linear function；The output category layer uses Softmax classifier.
7. electrical equipment fault probability forecasting method as described in claim 1, which is characterized in that long short-term memory nerve net
It includes: that acquisition training sample set data are trained that network, which is trained, is finely adjusted using back-propagation algorithm, Optimized model ginseng
Number.
8. a kind of electrical equipment fault probabilistic forecasting system for considering insulation defect classification and fault correlation, which is characterized in that
Include:
Data acquisition device acquires the PRPS spectrum data of power equipment and pre-processes to it；
Unsupervised network model self-encoding encoder, based on by pretreated PRPS spectrum data extraction local discharge characteristic；
Probability of malfunction prediction module comprising convolutional neural networks module, long Memory Neural Networks module and probability of malfunction in short-term
Output module, wherein convolutional neural networks module receives the local discharge characteristic of input, and output power equipment insulate with certain class
Probability value P (the D of defectk)；Long Memory Neural Networks module in short-term receives input local discharge characteristic, and output power equipment is in Dk
Under conditions of break down probability P (F | Dk)；The probability of malfunction module is final based on following formula output power equipment
Probability of malfunction P (F):
Wherein, n indicates the insulation defect number of types of power equipment, k=1,2,3 ... n；P(Dk) indicate that convolutional neural networks are defeated
Power equipment out is the probability value of certain class insulation defect, DkIndicate n class insulation defect；P(F|Dk) indicate long short-term memory nerve
The power equipment of network output is in DkUnder the conditions of the probability that breaks down.
9. electrical equipment fault probabilistic forecasting system as claimed in claim 8, which is characterized in that the power equipment at least wraps
Include GIS device, and the insulation defect number of types n=4 of GIS device.
10. electrical equipment fault probabilistic forecasting system as claimed in claim 8, which is characterized in that the convolutional neural networks
Module includes 1 input layer, 4 convolutional layers, 4 pond layers, 1 full articulamentum and 1 output category layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910602683.0A CN110334866B (en) | 2019-07-05 | 2019-07-05 | Power equipment fault probability prediction method and system considering insulation defect type and fault correlation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910602683.0A CN110334866B (en) | 2019-07-05 | 2019-07-05 | Power equipment fault probability prediction method and system considering insulation defect type and fault correlation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110334866A true CN110334866A (en) | 2019-10-15 |
| CN110334866B CN110334866B (en) | 2022-11-11 |
Family
ID=68144265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910602683.0A Active CN110334866B (en) | 2019-07-05 | 2019-07-05 | Power equipment fault probability prediction method and system considering insulation defect type and fault correlation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334866B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110907770A (en) * | 2019-11-28 | 2020-03-24 | 深圳供电局有限公司 | Partial discharge pulse feature extraction method and device, computer equipment and medium |
| CN111366820A (en) * | 2020-03-09 | 2020-07-03 | 广东电网有限责任公司电力科学研究院 | Pattern recognition method, device, equipment and storage medium for partial discharge signal |
| CN111881627A (en) * | 2020-08-05 | 2020-11-03 | 哈尔滨工程大学 | Nuclear power device fault diagnosis method and system |
| CN112749509A (en) * | 2020-12-30 | 2021-05-04 | 西华大学 | Intelligent substation fault diagnosis method based on LSTM neural network |
| CN112799330A (en) * | 2021-01-20 | 2021-05-14 | 浙江晶立捷环境科技有限公司 | AI control system and method for industrial wastewater treatment |
| CN113435759A (en) * | 2021-07-01 | 2021-09-24 | 贵州电网有限责任公司 | Primary equipment risk intelligent evaluation method based on deep learning |
| CN115905959A (en) * | 2023-01-06 | 2023-04-04 | 佰聆数据股份有限公司 | Method and device for analyzing relevance fault of power circuit breaker based on defect factor |
| CN117093952A (en) * | 2023-10-19 | 2023-11-21 | 北京松岛菱电电力工程有限公司 | Intelligent control method and system based on operation state of power equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105224782A (en) * | 2014-10-16 | 2016-01-06 | 华北电力大学 | A kind of converting equipment probability of malfunction computing method based on fault mode |
| CN109145961A (en) * | 2018-07-30 | 2019-01-04 | 上海交通大学 | A kind of mode identification method and system of unstructured Partial Discharge Data |
| CN109188211A (en) * | 2018-07-30 | 2019-01-11 | 上海交通大学 | A kind of high voltage equipment insulation method for diagnosing faults and system |
| CN109212392A (en) * | 2018-09-25 | 2019-01-15 | 上海交通大学 | Recognition methods, system and the Caffe convolutional neural networks of direct current cables shelf depreciation defect failure |
-
2019
- 2019-07-05 CN CN201910602683.0A patent/CN110334866B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105224782A (en) * | 2014-10-16 | 2016-01-06 | 华北电力大学 | A kind of converting equipment probability of malfunction computing method based on fault mode |
| CN109145961A (en) * | 2018-07-30 | 2019-01-04 | 上海交通大学 | A kind of mode identification method and system of unstructured Partial Discharge Data |
| CN109188211A (en) * | 2018-07-30 | 2019-01-11 | 上海交通大学 | A kind of high voltage equipment insulation method for diagnosing faults and system |
| CN109212392A (en) * | 2018-09-25 | 2019-01-15 | 上海交通大学 | Recognition methods, system and the Caffe convolutional neural networks of direct current cables shelf depreciation defect failure |
Non-Patent Citations (4)
| Title |
|---|
| LING ZHENG等: ""A Fault Prediction Of Equipment Based On CNN-LSTM Network"", 《2019 IEEE INTERNATIONAL CONFERENCE ON ENERGY INTERNET (ICEI)》 * |
| 代杰杰 等: ""采用 LSTM 网络的电力变压器运行状态预测方法研究"", 《高电压技术》 * |
| 宋辉 等: ""复杂数据源下基于深度卷积网络的局部放电模式识别"", 《高电压技术》 * |
| 宋辉 等: ""运行条件下 GIS局部放电严重程度评估方法"", 《中国电机工程学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110907770A (en) * | 2019-11-28 | 2020-03-24 | 深圳供电局有限公司 | Partial discharge pulse feature extraction method and device, computer equipment and medium |
| CN111366820A (en) * | 2020-03-09 | 2020-07-03 | 广东电网有限责任公司电力科学研究院 | Pattern recognition method, device, equipment and storage medium for partial discharge signal |
| CN111881627A (en) * | 2020-08-05 | 2020-11-03 | 哈尔滨工程大学 | Nuclear power device fault diagnosis method and system |
| CN112749509A (en) * | 2020-12-30 | 2021-05-04 | 西华大学 | Intelligent substation fault diagnosis method based on LSTM neural network |
| CN112799330A (en) * | 2021-01-20 | 2021-05-14 | 浙江晶立捷环境科技有限公司 | AI control system and method for industrial wastewater treatment |
| CN113435759A (en) * | 2021-07-01 | 2021-09-24 | 贵州电网有限责任公司 | Primary equipment risk intelligent evaluation method based on deep learning |
| CN113435759B (en) * | 2021-07-01 | 2023-07-04 | 贵州电网有限责任公司 | Primary equipment risk intelligent assessment method based on deep learning |
| CN115905959A (en) * | 2023-01-06 | 2023-04-04 | 佰聆数据股份有限公司 | Method and device for analyzing relevance fault of power circuit breaker based on defect factor |
| CN117093952A (en) * | 2023-10-19 | 2023-11-21 | 北京松岛菱电电力工程有限公司 | Intelligent control method and system based on operation state of power equipment |
| CN117093952B (en) * | 2023-10-19 | 2024-01-02 | 北京松岛菱电电力工程有限公司 | Intelligent control method and system based on operation state of power equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110334866B (en) | 2022-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334866A (en) | Consider the electrical equipment fault probability forecasting method and system of insulation defect classification and fault correlation | |
| Luo et al. | Data-driven short-term voltage stability assessment based on spatial-temporal graph convolutional network | |
| CN109635928A (en) | A kind of voltage sag reason recognition methods based on deep learning Model Fusion | |
| Dou et al. | Hybrid model for renewable energy and loads prediction based on data mining and variational mode decomposition | |
| CN105426839A (en) | Power system overvoltage classification method based on sparse autocoder | |
| CN110334948A (en) | Power equipment shelf depreciation Severity method and system based on characteristic quantity prediction | |
| Zha et al. | Ultra-short-term power forecast method for the wind farm based on feature selection and temporal convolution network | |
| CN110334865A (en) | A kind of electrical equipment fault rate prediction technique and system based on convolutional neural networks | |
| Wang et al. | Data-driven prediction method for characteristics of voltage sag based on fuzzy time series | |
| CN114021848A (en) | Generating capacity demand prediction method based on LSTM deep learning | |
| CN116842337A (en) | Transformer fault diagnosis method based on LightGBM (gallium nitride based) optimal characteristics and COA-CNN (chip on board) model | |
| Jing et al. | Adjustable piecewise regression strategy based wind turbine power forecasting for probabilistic condition monitoring | |
| Zhang et al. | Power grid fault diagnosis based on a deep pyramid convolutional neural network | |
| Li et al. | A probabilistic data-driven method for response-based load shedding against fault-induced delayed voltage recovery in power systems | |
| Song et al. | A Novel Wind Power Prediction Approach for Extreme Wind Conditions Based on TCN-LSTM and Transfer Learning | |
| Sun et al. | A multi-model-integration-based prediction methodology for the spatiotemporal distribution of vulnerabilities in integrated energy systems under the multi-type, imbalanced, and dependent input data scenarios | |
| Yao et al. | A new condition-monitoring method based on multi-variable correlation learning network for wind turbine fault detection | |
| Zheng et al. | Offshore wind power ramp prediction based on optimal combination model | |
| Zhenhao et al. | Prediction of wind power ramp events based on deep neural network | |
| Gawali et al. | Fault prediction model in wind turbines using deep learning structure with enhanced optimisation algorithm | |
| Jianli et al. | Wind power forecasting by using artificial neural networks and Grubbs criterion | |
| Yang et al. | Research on the Method of Station Load Prediction Based on SVR Optimized by GS-PSO | |
| Shan et al. | Power system fault diagnosis based on logistic regression deep neural network | |
| Bhuyan et al. | Fault Classification in a DG Connected Power System using Artificial Neural Network | |
| Muller | Creating building energy prediction models with convolutional recurrent neural networks |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |