CN108596470A - A kind of power equipments defect text handling method based on TensorFlow frames - Google Patents
A kind of power equipments defect text handling method based on TensorFlow frames Download PDFInfo
- Publication number
- CN108596470A CN108596470A CN201810355793.7A CN201810355793A CN108596470A CN 108596470 A CN108596470 A CN 108596470A CN 201810355793 A CN201810355793 A CN 201810355793A CN 108596470 A CN108596470 A CN 108596470A
- Authority
- CN
- China
- Prior art keywords
- text
- power equipments
- term vector
- convolution
- defect
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 28
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 7
- 230000000052 comparative effect Effects 0.000 claims abstract description 3
- 230000011218 segmentation Effects 0.000 claims description 8
- 238000005192 partition Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000001537 neural effect Effects 0.000 claims description 2
- 238000003062 neural network model Methods 0.000 claims 1
- 230000008569 process Effects 0.000 abstract description 7
- 238000012423 maintenance Methods 0.000 abstract description 3
- 238000007689 inspection Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000446 fuel Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000035699 permeability Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000009377 nuclear transmutation Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of power equipments defect text handling methods based on Tensorflow frames, this method first pre-processes electric power defect text, then term vector is converted text to using word2vec algorithms, improved convolutional neural networks model is finally built under TensorFlow frames again, and the term vector after input processing is trained network, obtains grader;After the completion of training, new power equipments defect text is inputted into grader, by actual result and output Comparative result, if rate of accuracy reached is trained and completed to setting accuracy;Otherwise continue to train after parameter adjustment.The present invention can be maked an inspection tour (maintenance) member with supplementary power devices and make more accurate judgement;Meanwhile the present invention may also help in power industry and effectively manage a large amount of empirical data, contribute to the succession to Knowledge Assets and effectively management.
Description
Technical field
The present invention relates to electric system big data technical fields, and in particular to a kind of electric power based on TensorFlow frames
Equipment deficiency text handling method.
Background technology
With the rapid development of intelligent grid, contacting more and more closely for electric system and Internet technology is generated
Explosive growth is also presented in data, constitutes the current academic electric power big data paid close attention to jointly with industrial quarters.It is long in electric system
In phase production practices, a large amount of valuable knowledge experiences are produced, for example senior check man can be by listening transformer sound to judge
Its whether operational excellence etc., how effective exploitation and to manage these " Knowledge Assets ", cause generally weighing for State Grid Corporation of China
Depending on.In addition to " old band new ", on the spot other than the management measures such as training, if machine learning techniques can be applied to a large amount of empirical datas point
Analysis, it will help the succession to Knowledge Assets and effectively management.
In empirical data, mostly important is the judgement to power equipments defect or failure, and maintenance person understands record electricity and sets
The various information and dates of received shipment row, and according to theoretical and a large amount of experiences it is concluded that.So by machine learning techniques to inspection
The power equipments defect or failure text that the person of repairing is recorded are analyzed and are judged, it will help the judgement of maintenance person and experience
Succession.
TensorFlow frames are the second generation artificial intelligence learning systems that Google is researched and developed based on DistBelief, are
One of common frame of artificial intelligence field, performance is strong, degree of optimization is high, applies it in electric power data analysis, and combine
Certain improve of practical progress will obtain more significant effect.
Invention content
Based on the above issues, it is an object of the invention to propose a kind of to handle power equipments defect or failure text
With the method for classification.
The technical solution adopted in the present invention is as follows:At a kind of power equipments defect text based on Tensorflow frames
Reason method, this method first pre-process electric power defect text, then convert text to word using word2vec algorithms
Vector finally builds improved convolutional neural networks model, and the term vector after input processing under TensorFlow frames again
Network is trained, grader is obtained;After the completion of training, new power equipments defect text is inputted into grader, it will be real
Border result and output Comparative result, if rate of accuracy reached is trained and completed to setting accuracy;Otherwise continue to instruct after parameter adjustment
Practice.
Further, described pre-process includes
(1) three classification processing are carried out to power equipments defect severity;
(2) stop words is carried out to defect text and punctuate is handled;
(3) ICTCLAS5.0 Words partition systems are used to carry out word segmentation processing to step (2) treated text.
Further, described that three classification processing are carried out to power equipments defect severity, specially:It is manually tight to defect
It is " general ", " important " and " urgent " three classes that the classification results of weight degree, which arrange,.
Further, the definition of the stop words is:Unrelated sentence is judged with power equipments defect, to eliminate to nerve
The influence of noise of network.
Further, the word segmentation processing segments text using the Words partition system ICTCLAS5.0 that increases income of the Chinese Academy of Sciences
Processing.
Further, described to build improved convolutional neural networks model under TensorFlow frames, it is specific as follows:
First, convolution operation is carried out to the term vector of input using the filter of multiple and different sizes in convolutional layer, and
And each filter constantly slides after convolution, to realize the traversal to entire input data;
After convolution, since filter is of different sizes, multiple convolution results can be obtained, these results are subjected to Max
Pool maximum pondizations operate, and the input maximum of convolutional layer is turned to long feature vector, that is, completes the operation of pond layer;
In full articulamentum, the intersection entropy loss of each class is calculated, and operate to the addition regularization of cross entropy loss result,
Wherein it is defined as using the cost function L for intersecting entropy loss:Wherein regularization is punished
:Each meaning of parameters:N is the sample size of training set, and x is sample size, and y is desired output (practical knot
Fruit), a is grader output valve, and λ is regularization coefficient, and w is weight parameter;Result will finally be obtained by being used for the god that classify more
The softmax classifier calculateds exported through network go out to belong to the probability of every class result.
Further, described that convolution behaviour is carried out to the term vector of input using the filter of multiple and different sizes in convolutional layer
Make, size is chosen for 3,4 and 5 words.
Beneficial effects of the present invention are as follows:
1) convolution operation is carried out to the term vector of input using the filter of multiple and different sizes in convolutional layer, size is chosen
For 3,4 and 5 words, to promote the receptive field size of convolutional layer, the definition of wherein receptive field is:The each layer of convolutional neural networks
The area size that pixel on the characteristic pattern of output maps on the original image.
2) result of convolutional layer is subjected to the operation of Max Pool maximum pondizations, obtains a long feature vector;To reduce
The computation complexity on upper layer reduces input size so that neural network can be absorbed in most important element;
3) the intersection entropy loss for calculating each class, to overcome the problems, such as that square error cost function update weight is excessively slow.And it adds
Regularization operates, and to prevent the generation of over-fitting, and improves the raising generalization ability of network.
4) result will be obtained by softmax classifier calculateds and goes out to belong to the probability of every class result;Softmax functions will patrol
Volume return that two graders are extensive has arrived polytypic situation, probability of outcome is bigger, illustrates that the possibility for belonging to the fault type is got over
Greatly.
5) all codes are completed under Tensorflow frames, and various hyper parameters is allowed to configure.
Description of the drawings
Fig. 1 is present invention training grader flow chart;
Fig. 2 is that the present invention improves convolutional neural networks process layer internal structure chart;
Fig. 3 is grader accuracy detection flow chart of the present invention.
Specific implementation mode
To make the objectives, technical solutions, and advantages of the present invention clearer, it will be set below with the electric power in substation
For standby, clear, complete description is carried out to technical scheme of the present invention in conjunction with attached drawing.
Obviously, described example is only a part of the embodiment of the present invention.Based on the embodiment of the present invention, this field
Those of ordinary skill's obtained every other embodiment under the premise of not making inventive improvements, belongs to institute of the present invention
The range of protection.
First, power equipments defect degree in substation is classified, is divided into " general ", " important " and " urgent " three
Class, foundation are Zhejiang Electric Power Company .Q/ZDJ 44-2005 power grid power transformation primary equipment defect language specification [S] .2005.
Stop words and punctuate in artificial removal's substation equipment defect record, such as " Seeking Truth change ", " bottle transmutation ", " city
The non-effective information such as west change ";The definition of wherein stop words is:Judge unrelated sentence with power equipments defect, such as:Substation
Title, floor manager's name etc., to eliminate the influence of noise to neural network;
The Words partition system ICTCLAS5.0 that increases income for reusing the Chinese Academy of Sciences carries out word segmentation processing to Chinese text, will be existing
Text dividing is conveniently constructed term vector at independent word, for example incite somebody to action " main transformer body conservator fuel outlet valve permeability drips for 3 minutes one,
Temperature is normal " become after word segmentation processing " main transformer+ontology+conservator+fuel outlet valve+drop of permeability+3 minutes+one+temperature+normal ";
Result after word segmentation processing is obtained into term vector by word2vec term vector transformation models, so far completes attached drawing
Input layer module in 1 provides the data convenient for processing for process layer;Wherein word2vec is selected to carry out term vector conversion
On the one hand reason is that the distance between the term vector that word2vec model conversations obtain represents word semantic and phraseological similar
Property, this is not available for universal model (as NNLM), and another aspect word2vec models are also to be developed by Google, and rear
The continuous Tensorflow frames used have stronger compatibility.
It will obtain the 80% of data and be used as training set, input the convolutional neural networks under Tensorflow frames to grader
It is trained;Make to select the filter of multiple and different sizes to carry out convolution operation to the term vector of input in convolutional layer,
And each filter constantly slides after convolution, to realize the traversal to entire input data, it is contemplated that maximum sentence
Length is chosen for 4 words in this network design, so far completes A and B-stage in Fig. 2 process layers;
After convolution, since filter is of different sizes, multiple convolution results can be obtained, these these results are carried out
Max Pool maximum pondizations operate, and obtain a long feature vector, the as C-stage in attached drawing 2;
In full articulamentum, the intersection entropy loss of each class is calculated, and operate to the addition regularization of cross entropy loss result,
To prevent the generation of over-fitting, over-fitting is prevented, and improves the raising generalization ability of network;
Wherein it is defined as using the cost function for intersecting entropy loss:Wherein canonical
Change penalty term:
Each meaning of parameters:The sample size (80%) of n-- training sets;X-- sample sizes, y-- by " general ", " important " and
" urgent " is converted to vector form as desired output:(1,0,0)、(0,1,0)、(0,0,1);A-- real output values;λ--
Regularization coefficient;W-- weight parameters;
Result will be finally obtained to go out to belong to every class by the softmax classifier calculateds exported for more Classification Neurals
As a result probability so far completes the D stages in attached drawing 2;
After completing process layer, remaining 20% test data is input to the grader obtained by training after processing
In, and classification results will be obtained and be compared with actual result, classification accuracy rate ratio is obtained, to verify whether grader needs
It is further improved, improved procedure is mainly completed by adjusting hyper parameter, enhancing training set data two ways;Specific implementation stream
Journey figure embodies in fig. 3, after inputting text test set, first passes around input processing layer and carries out term vector conversion, later
Or else open close grader of crossing carries out prediction and parameter adjustment, after training result and actual result meet accuracy requirement, then
Complete the training of model.
All codes are completed under Tensorflow frames, and various hyper parameters is allowed to configure.
Claims (7)
1. a kind of power equipments defect text handling method based on Tensorflow frames, which is characterized in that this method is first
Electric power defect text is pre-processed, term vector is then converted text to using word2vec algorithms, is finally existed again
Improved convolutional neural networks model is built under TensorFlow frames, and the term vector after input processing instructs network
Practice, obtains grader;After the completion of training, new power equipments defect text is inputted into grader, by actual result and output
Comparative result, if rate of accuracy reached is trained and completed to setting accuracy;Otherwise continue to train after parameter adjustment.
2. according to the method described in claim 1, it is characterized in that, the pretreatment includes
(1) three classification processing are carried out to power equipments defect severity;
(2) stop words is carried out to defect text and punctuate is handled;
(3) ICTCLAS5.0 Words partition systems are used to carry out word segmentation processing to step (2) treated text.
3. according to the method described in claim 1, it is characterized in that, described carry out three classification to power equipments defect severity
Processing, specially:It is " general ", " important " and " urgent " three classes manually to be arranged to the classification results of defect severity.
4. according to the method described in claim 1, it is characterized in that, the definition of the stop words is:Sentence with power equipments defect
Disconnected unrelated sentence, to eliminate the influence of noise to neural network.
5. according to the method described in claim 1, it is characterized in that, the word segmentation processing using the Chinese Academy of Sciences Words partition system of increasing income
ICTCLAS5.0 carries out word segmentation processing to text.
6. according to the method described in claim 1, it is characterized in that, described build improved convolution under TensorFlow frames
Neural network model, it is specific as follows:
First, convolution operation is carried out to the term vector of input using the filter of multiple and different sizes in convolutional layer, and every
A filter constantly slides after convolution, to realize the traversal to entire input data;
After convolution, since filter is of different sizes, multiple convolution results can be obtained, these results are subjected to Max Pool most
Great Chiization operates, and the input maximum of convolutional layer is turned to long feature vector, that is, completes the operation of pond layer;
In full articulamentum, the intersection entropy loss of each class is calculated, and to the addition regularization operation of cross entropy loss result, wherein
It is defined as using the cost function L for intersecting entropy loss:Wherein regularization penalty term:Each meaning of parameters:N is the sample size of training set, and x is sample size, and y is desired output (actual result), a
For grader output valve, λ is regularization coefficient, and w is weight parameter;Result will finally be obtained by being used for more Classification Neurals
The softmax classifier calculateds of output go out the probability for belonging to every class result.
7. according to the method described in claim 1, it is characterized in that, the filter for using multiple and different sizes in convolutional layer
Convolution operation is carried out to the term vector of input, size is chosen for 3,4 and 5 words.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810355793.7A CN108596470A (en) | 2018-04-19 | 2018-04-19 | A kind of power equipments defect text handling method based on TensorFlow frames |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810355793.7A CN108596470A (en) | 2018-04-19 | 2018-04-19 | A kind of power equipments defect text handling method based on TensorFlow frames |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108596470A true CN108596470A (en) | 2018-09-28 |
Family
ID=63611309
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810355793.7A Pending CN108596470A (en) | 2018-04-19 | 2018-04-19 | A kind of power equipments defect text handling method based on TensorFlow frames |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108596470A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109376229A (en) * | 2018-12-04 | 2019-02-22 | 国网江西省电力有限公司信息通信分公司 | A kind of click bait detection method based on convolutional neural networks |
CN109918499A (en) * | 2019-01-14 | 2019-06-21 | 平安科技(深圳)有限公司 | A kind of file classification method, device, computer equipment and storage medium |
CN110083719A (en) * | 2019-03-29 | 2019-08-02 | 杭州电子科技大学 | A kind of industrial product defect detection method based on deep learning |
CN111191447A (en) * | 2019-12-18 | 2020-05-22 | 东软集团股份有限公司 | Equipment defect classification method, device and equipment |
CN111767398A (en) * | 2020-06-30 | 2020-10-13 | 国网新疆电力有限公司电力科学研究院 | Secondary equipment fault short text data classification method based on convolutional neural network |
CN111767397A (en) * | 2020-06-30 | 2020-10-13 | 国网新疆电力有限公司电力科学研究院 | Electric power system secondary equipment fault short text data classification method |
CN111798853A (en) * | 2020-03-27 | 2020-10-20 | 北京京东尚科信息技术有限公司 | Method, device, equipment and computer readable medium for speech recognition |
CN111966825A (en) * | 2020-07-16 | 2020-11-20 | 电子科技大学 | Power grid equipment defect text classification method based on machine learning |
CN112037215A (en) * | 2020-09-09 | 2020-12-04 | 华北电力大学(保定) | Insulator defect detection method and system based on zero sample learning |
CN113111183A (en) * | 2021-04-20 | 2021-07-13 | 通号(长沙)轨道交通控制技术有限公司 | Traction power supply equipment defect grade classification method |
CN113421271A (en) * | 2021-05-17 | 2021-09-21 | 江苏大学 | Natural leather defect segmentation method based on MLP multilayer neural network |
CN113435195A (en) * | 2021-07-01 | 2021-09-24 | 贵州电网有限责任公司 | Defect intelligent diagnosis model construction method based on main transformer load characteristics |
CN114708267A (en) * | 2022-06-07 | 2022-07-05 | 浙江大学 | Image detection processing method for corrosion defect of tower stay wire on power transmission line |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106250934A (en) * | 2016-08-12 | 2016-12-21 | 南方电网科学研究院有限责任公司 | Defect data classification method and device |
CN107024331A (en) * | 2017-03-31 | 2017-08-08 | 中车工业研究院有限公司 | A kind of neutral net is to train motor oscillating online test method |
CN107908716A (en) * | 2017-11-10 | 2018-04-13 | 国网山东省电力公司电力科学研究院 | 95598 work order text mining method and apparatus of word-based vector model |
-
2018
- 2018-04-19 CN CN201810355793.7A patent/CN108596470A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106250934A (en) * | 2016-08-12 | 2016-12-21 | 南方电网科学研究院有限责任公司 | Defect data classification method and device |
CN107024331A (en) * | 2017-03-31 | 2017-08-08 | 中车工业研究院有限公司 | A kind of neutral net is to train motor oscillating online test method |
CN107908716A (en) * | 2017-11-10 | 2018-04-13 | 国网山东省电力公司电力科学研究院 | 95598 work order text mining method and apparatus of word-based vector model |
Non-Patent Citations (2)
Title |
---|
刘梓权等: "基于卷积神经网络的电力设备缺陷文本分类模型研究", 《电网技术》 * |
张天琦: "基于深度学习的行人流量统计算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109376229A (en) * | 2018-12-04 | 2019-02-22 | 国网江西省电力有限公司信息通信分公司 | A kind of click bait detection method based on convolutional neural networks |
CN109918499A (en) * | 2019-01-14 | 2019-06-21 | 平安科技(深圳)有限公司 | A kind of file classification method, device, computer equipment and storage medium |
CN110083719A (en) * | 2019-03-29 | 2019-08-02 | 杭州电子科技大学 | A kind of industrial product defect detection method based on deep learning |
CN111191447B (en) * | 2019-12-18 | 2023-07-14 | 东软集团股份有限公司 | Equipment defect classification method, device and equipment |
CN111191447A (en) * | 2019-12-18 | 2020-05-22 | 东软集团股份有限公司 | Equipment defect classification method, device and equipment |
CN111798853A (en) * | 2020-03-27 | 2020-10-20 | 北京京东尚科信息技术有限公司 | Method, device, equipment and computer readable medium for speech recognition |
CN111767398A (en) * | 2020-06-30 | 2020-10-13 | 国网新疆电力有限公司电力科学研究院 | Secondary equipment fault short text data classification method based on convolutional neural network |
CN111767397A (en) * | 2020-06-30 | 2020-10-13 | 国网新疆电力有限公司电力科学研究院 | Electric power system secondary equipment fault short text data classification method |
CN111966825A (en) * | 2020-07-16 | 2020-11-20 | 电子科技大学 | Power grid equipment defect text classification method based on machine learning |
CN112037215A (en) * | 2020-09-09 | 2020-12-04 | 华北电力大学(保定) | Insulator defect detection method and system based on zero sample learning |
WO2022052181A1 (en) * | 2020-09-09 | 2022-03-17 | 华北电力大学(保定) | Insulator defect detection method and system based on zero-shot learning |
CN112037215B (en) * | 2020-09-09 | 2024-05-28 | 华北电力大学(保定) | Insulator defect detection method and system based on zero sample learning |
CN113111183A (en) * | 2021-04-20 | 2021-07-13 | 通号(长沙)轨道交通控制技术有限公司 | Traction power supply equipment defect grade classification method |
CN113421271A (en) * | 2021-05-17 | 2021-09-21 | 江苏大学 | Natural leather defect segmentation method based on MLP multilayer neural network |
CN113435195A (en) * | 2021-07-01 | 2021-09-24 | 贵州电网有限责任公司 | Defect intelligent diagnosis model construction method based on main transformer load characteristics |
CN113435195B (en) * | 2021-07-01 | 2023-10-03 | 贵州电网有限责任公司 | Defect intelligent diagnosis model construction method based on main transformer load characteristics |
CN114708267A (en) * | 2022-06-07 | 2022-07-05 | 浙江大学 | Image detection processing method for corrosion defect of tower stay wire on power transmission line |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108596470A (en) | A kind of power equipments defect text handling method based on TensorFlow frames | |
Shamshirband et al. | A survey of deep learning techniques: application in wind and solar energy resources | |
CN104077613B (en) | Crowd density estimation method based on cascaded multilevel convolution neural network | |
CN107908671A (en) | Knowledge mapping construction method and system based on law data | |
CN110929918B (en) | 10kV feeder fault prediction method based on CNN and LightGBM | |
CN103679611B (en) | Operation method of city comprehensive emergency intelligent simulation system based on case-based reasoning | |
WO2023142424A1 (en) | Power financial service risk control method and system based on gru-lstm neural network | |
CN107767384A (en) | A kind of image, semantic dividing method based on dual training | |
CN107169598A (en) | A kind of weather prediction method and system based on deep learning | |
CN110009095A (en) | Road driving area efficient dividing method based on depth characteristic compression convolutional network | |
CN109977921A (en) | A kind of transmission line of electricity perils detecting method | |
CN106650725A (en) | Full convolutional neural network-based candidate text box generation and text detection method | |
CN110009030A (en) | Sewage treatment method for diagnosing faults based on stacking meta learning strategy | |
CN109086892A (en) | It is a kind of based on the visual problem inference pattern and system that typically rely on tree | |
CN112557826A (en) | Ship electric power system fault diagnosis method | |
CN112488025A (en) | Double-temporal remote sensing image semantic change detection method based on multi-modal feature fusion | |
Huang et al. | Research on urban modern architectural art based on artificial intelligence and GIS image recognition system | |
CN102564496A (en) | Micro-analysis method for transformer device based on BP nerve network and manual shoal | |
CN107871183A (en) | Permafrost Area highway distress Forecasting Methodology based on uncertain Clouds theory | |
CN106951844A (en) | A kind of Method of EEG signals classification and system based on the very fast learning machine of depth | |
Peng et al. | A forecast model of tourism demand driven by social network data | |
Zhang et al. | A combined traffic flow forecasting model based on graph convolutional network and attention mechanism | |
Wang et al. | Weather condition-based hybrid models for multiple air pollutants forecasting and minimisation | |
Zhou et al. | Fault detection of train height valve based on nanodet-resnet101 | |
CN107563508A (en) | A kind of big data failure prediction method |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180928 |