CN110516742A - A kind of distribution terminal fault distinguishing method and system based on Combination neural network model - Google Patents
A kind of distribution terminal fault distinguishing method and system based on Combination neural network model Download PDFInfo
- Publication number
- CN110516742A CN110516742A CN201910803104.9A CN201910803104A CN110516742A CN 110516742 A CN110516742 A CN 110516742A CN 201910803104 A CN201910803104 A CN 201910803104A CN 110516742 A CN110516742 A CN 110516742A
- Authority
- CN
- China
- Prior art keywords
- training set
- distribution terminal
- time
- neural network
- data
- 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
- 238000009826 distribution Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000003062 neural network model Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 78
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 25
- 238000004519 manufacturing process Methods 0.000 claims abstract description 9
- 230000004069 differentiation Effects 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims abstract description 7
- 230000004913 activation Effects 0.000 claims description 19
- 238000006243 chemical reaction Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 206010048669 Terminal state Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000009514 concussion Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012631 diagnostic technique Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Economics (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application is specifically related to a kind of distribution terminal fault distinguishing method and system based on Combination neural network model, it converts to obtain the time-frequency map of each data by the electric data progress time domain for the time domain for measuring distribution terminal measuring system and constitutes original training set, dilatation is generated to original training set using production confrontation neural network GAN, the new time-frequency map and original training set generated at random according to random number sequence constitutes new training set, provides more more fully training samples for convolutional neural networks CNN;Model training is carried out using convolutional neural networks CNN on the basis of GAN production dilatation, substantially increases differentiation accuracy of the discrimination model to the data of unknown failure type;Method and system provided by the invention can differentiate fault type online, timely feedback the equipment running status of distribution terminal, fault type could be obtained to overhaul of the equipments without arriving device context under off-line state, to improve O&M efficiency and reduce maintenance work amoun.
Description
Technical field
The invention belongs to electric power system power distribution terminal fault diagnostic techniques fields, and in particular to one kind is based on combination nerve net
The distribution terminal fault distinguishing method and system of network model.
Background technique
Open locking, on-pole switch, ring network cabinet etc. of the distribution terminal for medium voltage distribution network in the power system, is distribution
Important monitoring and control equipment in automation, when distribution terminal breaks down its electric data acquired will appear it is abnormal and
It accidentally surveys, therefore the reliability of distribution terminal directly affects the perception and control effect of operation of power networks state.In the prior art to matching
The breakdown judge of electric terminals mostly carries out in the case where entirely ineffective offline, and the defect of this processing means is to arrange in time
Except failure, discovery and maintenance to failure are passively lagged, and are inevitably led to influence consequence of the failure to power distribution network and are deepened, shadow
Range is rung to expand.
Therefore it provides a kind of method and system that can differentiate distribution terminal fault type in time are for ensuring that distribution terminal is set
Standby reliability is particularly important.
Summary of the invention
Based on this, the present invention is intended to provide a kind of distribution terminal fault distinguishing method based on Combination neural network model and
The electric data that distribution terminal acquires is converted into time-frequency figure, is trained structure to combination neural net using time-frequency figure by system
The discrimination model that is out of order is built, the electric data of real-time online acquisition is inputted trained fault distinguishing model and carries out fault type
Differentiate, to solve the technical issues of passively lagging to the differentiation of failure in the prior art.
A kind of distribution terminal fault distinguishing method based on Combination neural network model of the present invention, comprising:
Time-frequency conversion is carried out to the electric data of distribution terminal different faults type, time-frequency map is obtained and constitutes original training
Collection;
Original training set input production confrontation neural network GAN carry out production dilatation, according to random number sequence with
Machine generates new time-frequency map subset, together constitutes with new training set with original training set, new training set inputs convolutional neural networks
CNN carries out convolution sum pond, in the number of full articulamentum output fault type, obtains the fault distinguishing model that training finishes;
The electric data that Input Online acquires in real time carries out fault type differentiation using fault distinguishing model and exports differentiation
As a result.
Preferably, before to the electric data progress time-frequency conversion of distribution terminal different faults type further include:
Electric data measured by measuring system when obtaining distribution terminal failure.
Preferably, electric data further comprises:
Voltage collected, electric current time domain exchange data when distribution terminal failure.
Preferably, time-frequency conversion further comprises:
Short Time Fourier Transform, window function of the following formula as Short Time Fourier Transform
Preferably, the objective function of production confrontation neural network is as follows,
Wherein, G indicates the generator G in GAN, for generating new time-frequency map,
D indicates the arbiter D in GAN, for distinguishing the new time-frequency map of original time-frequency map and generation,
X is characterized sequence, and z is random number sequence;prIndicate initial data time-frequency characteristics atlas;pgIt indicates by the life in GAN
The new data time-frequency characteristics atlas that the G that grows up to be a useful person is generated.
Preferably, new training set input convolutional neural networks CNN includes: in convolutional layer and pond layer progress convolution sum pond
New training set is pre-processed and normalized.
Preferably, new training set input convolutional neural networks CNN carries out convolution sum pond in convolutional layer and pond layer, complete
The number of articulamentum output fault type further comprises:
New training set passes sequentially through the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first connection
Layer, the second full articulamentum carry out feature extraction and dimensionality reduction to new training set, export fault type by the activation primitive of each layer
Number;
Wherein, the first convolutional layer and the second convolutional layer use identical activation primitive, the first pond layer and the second pond layer
Using identical activation primitive, the first full articulamentum and the second full articulamentum use identical activation primitive.
It is preferably, as follows in the convolution process expression formula of the first convolutional layer and the progress of the second convolutional layer,
Wherein,Indicate j-th of output element of l layer,Indicate that a convolution kernel of the convolutional layer, " * " indicate volume
Product operation,Indicate that the convolutional layer biases;F () is activation primitive.
A kind of distribution terminal fault distinguishing system based on Combination neural network model, comprising:
Time-frequency conversion processing module is carried out from time domain to frequency domain for the electric data to distribution terminal different faults type
Transformation, obtain time-frequency map and constitute original training set;
Dilatation module, for carrying out feature learning to original training set to sample size dilatation to original training set,
Obtain new training set;
Model training module carries out model training using new training set and obtains fault distinguishing model;
Output module differentiates result for exporting fault type.
Preferably, being somebody's turn to do the distribution terminal fault distinguishing system based on Combination neural network model further comprises:
Data acquisition module, electric data measured by measuring system when for acquiring distribution terminal failure.
As can be seen from the above technical solutions, the invention has the following advantages that
A kind of distribution terminal fault distinguishing method and system based on Combination neural network model of the present invention, by time domain
Electric data carry out time domain convert to obtain time-frequency map, data image, so that the feature of the data of different faults type
It is more prominent, the Feature capturing to data is facilitated, data-handling efficiency is improved;It is compared in the prior art with distribution terminal
Equipment surface phenomenon of the failure is discrimination standard, and the electric data image conversion for the time domain that the present invention measures measuring system is sufficiently dug
The implicit information potential of electric data has been dug, has differentiated that result is more acurrate objective;It is instructed using production confrontation neural network GAN
Practice collection and generate dilatation, the new time-frequency map and original training set that GAN is generated when extracting equilibrium state constitute new training set, can be
Convolutional neural networks CNN provides more more fully training samples, improves the accuracy of fault distinguishing model;It is generated in GAN
It is trained on the basis of formula dilatation using CNN, it is quasi- to the differentiation of the data of unknown failure type to substantially increase discrimination model
Exactness;Method and system provided by the invention can differentiate fault type online, timely feedback the equipment operation of distribution terminal
State, without under off-line state to device context could to overhaul of the equipments obtain fault type, thus improve O&M efficiency and
Reduce maintenance work amoun.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Distribution terminal fault distinguishing method implementing procedure of Fig. 1 an embodiment of the present invention based on Combination neural network model
Figure
Fig. 2 an embodiment of the present invention is illustrated based on the distribution terminal fault distinguishing system structure of Combination neural network model
Figure
The time-frequency map of electric data in Fig. 3 a an embodiment of the present invention after time-frequency conversion in normal condition
In Fig. 3 b an embodiment of the present invention after time-frequency conversion in big noise failure state electric data when
Frequency map
In the electric data for shaking wear-out failure state after time-frequency conversion in Fig. 3 c an embodiment of the present invention
Time-frequency map
The time-frequency of electric data in Fig. 3 d an embodiment of the present invention after time-frequency conversion in harmonic failure state
Map
The penalty values contrast schematic diagram of difference CNN training method in another embodiment of Fig. 4 present invention
The differentiation accuracy rate contrast schematic diagram of difference CNN training method in another embodiment of Fig. 5 present invention
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1 and Fig. 2 are please referred to, the present embodiment provides a kind of distribution terminal fault distinguishing based on Combination neural network model
Method and system, including data acquisition module 100, time-frequency conversion processing module 110, dilatation module 120, model training module
130, output module 140, the specific implementation following steps of method in the present embodiment:
It is electrical measured by measuring system when obtaining distribution terminal generation different types of faults using data acquisition module 100
Data, the time domain including voltage and current exchange data, and the present embodiment has chosen 4 kinds of fault types and includes normal condition, makes an uproar greatly
Acoustic events, concussion wear-out failure, harmonic failure, each type of sample size are 200, altogether 800 samples;
The electric data input time-frequency conversion module 110 of 4 seed types is carried out time-frequency conversion, the present embodiment uses Fu in short-term
In leaf transformation, converted in MATLAB using analytic function Spectrogram, wherein Short Time Fourier Transform process such as formula
(1) shown in,
Wherein, f (t) is time-domain function to be transformed;Centered on g (t- τ) in the window function at τ moment, the present embodiment
Shown in window function expression formula such as formula (2),
Please refer to Fig. 3 a to Fig. 3 d, it can be seen that have passed through time-frequency conversion, data characteristics ratio more protrudes bright in time domain
It is aobvious, to the Feature capturing of data when this is conducive to model training.
The original training set X that sample size is 800 is obtained after time-frequency conversion, and original training set X is inputted dilatation module
120, deep learning frame of the present embodiment based on Google TensorFlow sets time cutting, the data normalization of frequency map
Deng processing, build the GAN network architecture, and corresponding model parameter is set later, design parameter is as shown in table 1,
1 DCGAN structural parameters table of table
It is to generate new data sample subset M at random according to random number sequence to the dilatation of training set X in the present embodiment,
GAN generates 4 kinds of new each 100 of fault type data sample time-frequency figures, totally 400 time-frequency maps, then the new training set N exported
Sample size be (200+100) × 4=1200, shown in the objective function of GAN such as formula (3),
Wherein, G indicates that generator G, D in GAN indicate that arbiter D, x in GAN are characterized sequence, and z is random number sequence
Column;prIndicate initial data time-frequency characteristics atlas;pgIndicate the new data time-frequency characteristics atlas generated by the generator G in GAN.
New training set N input model training module 130 carry out model training, using CNN to data carry out pretreatment and
Normalization, passes sequentially through the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, second
Full articulamentum, wherein the first convolutional layer and the second convolutional layer use identical activation primitive, the first pond layer and the second pond layer
Using identical activation primitive, the first full articulamentum and the second full articulamentum use identical activation primitive, roll up in the present embodiment
The activation primitive of lamination selects ReLU function, and the activation primitive of full articulamentum selects SoftMax function, the maximum pond of pond layer choosing
Change method;
Shown in the process of convolution such as formula (4),
Wherein,Indicate j-th of output element of l layer,Indicate that a convolution kernel of the convolutional layer, " * " indicate volume
Product operation,Indicate that the convolutional layer biases;F () is activation primitive.
The CNN network structure that the present embodiment is chosen is LeNet-5, and specific structure is as shown in table 2, and takes Adam algorithm
Loss function optimization is carried out, it is 0.0001 that global learning rate, which is arranged, and each batch is 20 samples.
The structural parameters table of 2 CNN of table
Fault distinguishing model is finally obtained by the combined training of GAN and CNN, the electric data of online acquisition is inputted event
Hinder discrimination model, output module 140 will differentiate result output.
Another embodiment is described below, the present embodiment sentences a kind of distribution end failure based on Combination neural network model
Other method is verified, and the present embodiment compares the influence of different capabilities and different dilatation ways to fault distinguishing model.Choose three
A training set, each training set include 4 kinds of data types.
Training set A is original training set, and without GAN dilatation, the sample size of each data type is 200 in sample, then
Total sample size of training set A is 800;
Training set B is the combined training collection that dilatation is generated by GAN, and GAN carries out the data of every kind of data type random
It generates, each data type generates new time-frequency map 100 and opens, then total sample size of training set B is (200+100) × 4=
1200。
Training set C is that random reproduction 400 opens time-frequency figure, this kind of dilatation on the basis of the original training set that sample size is 800
Mode is simple random reproduction, does not generate the data sample with new feature, is constituted new training with original training set with this
Collection, total sample size of training set C are 800+400=1200.
Model training is carried out to training set A, training set B, training set C respectively, data are pre-processed and are normalized, according to
It is secondary to pass through the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, the second full articulamentum,
Wherein, the first convolutional layer and the second convolutional layer use identical activation primitive, the first pond layer and the second pond layer to use identical
Activation primitive, the first full articulamentum and the second full articulamentum use identical activation primitive, and convolutional layer swashs in the present embodiment
Function selection ReLU function living, the activation primitive selection SoftMax function of full articulamentum, pond layer choosing maximum Chi Huafa, most
3 fault distinguishing models for respectively corresponding training set A, training set B, training set C are exported afterwards.
Compare 3 kinds of fault distinguishing models penalty values in CNN training process, the loss function in the present embodiment is using intersection
Shown in entropy function such as formula (5),
Wherein n is fault sample sum;Y is true value;For predicted value,
Referring to FIG. 4, it can be seen that in same sample size, the fault distinguishing model of the method for discrimination of the present embodiment
Penalty values are lower, and compared with not dilatation, penalty values are even more to be greatly reduced;
Referring to FIG. 5, it can be seen that the fault distinguishing model of the method for discrimination of the present embodiment is on differentiating accuracy rate than letter
The discrimination model of single dilatation and non-dilatation will be high, then a kind of distribution based on Combination neural network model provided in this embodiment
Terminal fault method of discrimination has a clear superiority on improving fault identification efficiency and accuracy, can obtain beneficial effect.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although with reference to the foregoing embodiments
Invention is explained in detail, and those skilled in the art is it is understood that it can still remember previous embodiment
The technical solution of load is modified or equivalent replacement of some of the technical features;And these are modified or replaceed, and
The essence of corresponding technical solution is not set to be detached from the spirit and scope of technical solution of the embodiment of the present invention.
Claims (10)
1. a kind of distribution terminal fault distinguishing method based on Combination neural network model characterized by comprising
Time-frequency conversion is carried out to the electric data of distribution terminal different faults type, time-frequency map is obtained and constitutes original training set;
The original training set input production confrontation neural network GAN is carried out production dilatation, according to random number sequence with
Machine generates new time-frequency map subset, together constitutes with new training set with the original training set, the new training set inputs convolution
Neural network CNN carries out convolution sum pond, in the number of full articulamentum output fault type, obtains the fault distinguishing that training finishes
Model;
The testing data that Input Online acquires in real time carries out fault type differentiation using the fault distinguishing model and exports differentiation
As a result.
2. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 1, special
Sign is that the electric data to distribution terminal different faults type carries out before time-frequency conversion further include:
Electric data measured by measuring system when obtaining the distribution terminal failure.
3. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 1, special
Sign is that the electric data further comprises:
Voltage collected, electric current time domain exchange data when distribution terminal failure.
4. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 1, special
Sign is that the time-frequency conversion further comprises:
Short Time Fourier Transform, the following formula of the window function of the Short Time Fourier Transform indicate
5. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 1, special
Sign is that the objective function of the production confrontation neural network is as follows,
Wherein, G indicates the generator G in GAN, for generating new time-frequency map,
D indicates the arbiter D in GAN, for distinguishing the new time-frequency map of original time-frequency map and generation,
X is characterized sequence, and z is random number sequence;prIndicate initial data time-frequency characteristics atlas;pgIt indicates by the generator in GAN
The new data time-frequency characteristics atlas that G is generated.
6. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 1, special
Sign is that the new training set input convolutional neural networks CNN carries out convolution sum pond in convolutional layer and pond layer further include:
The new training set is pre-processed and normalized.
7. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 1, special
Sign is that new training set input convolutional neural networks CNN carries out convolution sum pond in convolutional layer and pond layer, defeated in full articulamentum
The number for the type that is out of order further comprises:
New training set pass sequentially through the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first full articulamentum,
Second full articulamentum carries out feature extraction and dimensionality reduction to the new training set, exports fault type by the activation primitive of each layer
Number;
Wherein, first convolutional layer and second convolutional layer use identical activation primitive, first pond layer and institute
It states the second pond layer and is swashed using identical activation primitive, the first full articulamentum and the second full articulamentum using identical
Function living.
8. a kind of distribution terminal fault distinguishing method based on Combination neural network model according to claim 7, special
Sign is,
It is as follows in the convolution process expression formula that first convolutional layer and second convolutional layer carry out,
Wherein,Indicate j-th of output element of l layer,Indicate that a convolution kernel of the convolutional layer, " * " indicate convolution behaviour
Make,Indicate that the convolutional layer biases;F () is activation primitive.
9. a kind of distribution terminal fault distinguishing system based on Combination neural network model characterized by comprising
Time-frequency conversion processing module carries out the change from time domain to frequency domain for the electric data to distribution terminal different faults type
It changes, obtains time-frequency map and constitute original training set;
Dilatation module, for carrying out feature learning to the original training set to which the sample size to the original training set expands
Hold, obtains new training set;
Model training module carries out model training using the new training set and obtains fault distinguishing model;
Output module differentiates result for exporting fault type.
10. a kind of distribution terminal fault distinguishing system based on Combination neural network model according to claim 9, special
Sign is that the system further comprises:
Data acquisition module, electric data measured by measuring system when for acquiring the distribution terminal failure.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910803104.9A CN110516742A (en) | 2019-08-28 | 2019-08-28 | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910803104.9A CN110516742A (en) | 2019-08-28 | 2019-08-28 | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110516742A true CN110516742A (en) | 2019-11-29 |
Family
ID=68627608
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910803104.9A Pending CN110516742A (en) | 2019-08-28 | 2019-08-28 | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110516742A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112001314A (en) * | 2020-08-25 | 2020-11-27 | 江苏师范大学 | Early fault detection method for variable speed hoist |
CN113111591A (en) * | 2021-04-29 | 2021-07-13 | 南方电网电力科技股份有限公司 | Automatic diagnosis method, device and equipment based on internal fault of modular power distribution terminal |
CN113592181A (en) * | 2021-08-02 | 2021-11-02 | 广西大学 | Small hydropower station output prediction method and system |
CN113687610A (en) * | 2021-07-28 | 2021-11-23 | 国网江苏省电力有限公司南京供电分公司 | Method for protecting terminal information of GAN-CNN power monitoring system |
WO2023098753A1 (en) * | 2021-12-02 | 2023-06-08 | 广东电网有限责任公司江门供电局 | Power distribution terminal fault diagnosis method, system and apparatus, and storage medium |
CN116776130A (en) * | 2023-08-23 | 2023-09-19 | 成都新欣神风电子科技有限公司 | Detection method and device for abnormal circuit signals |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109444604A (en) * | 2018-12-13 | 2019-03-08 | 武汉理工大学 | A kind of DC/DC converter method for diagnosing faults based on convolutional neural networks |
CN210923872U (en) * | 2019-08-28 | 2020-07-03 | 广东电科院能源技术有限责任公司 | Online discrimination system of distribution terminal trouble |
-
2019
- 2019-08-28 CN CN201910803104.9A patent/CN110516742A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109444604A (en) * | 2018-12-13 | 2019-03-08 | 武汉理工大学 | A kind of DC/DC converter method for diagnosing faults based on convolutional neural networks |
CN210923872U (en) * | 2019-08-28 | 2020-07-03 | 广东电科院能源技术有限责任公司 | Online discrimination system of distribution terminal trouble |
Non-Patent Citations (2)
Title |
---|
包萍等: ""不均衡数据集下基于生成对抗网络的改进深度模型故障识别研究"", 《电子测量与仪器学报》, vol. 33, no. 3, pages 176 - 183 * |
李恒等: ""基于短时傅里叶变换和卷积神经网络的轴承故障诊断方法"", 《振动与冲击》, vol. 37, no. 19, pages 124 - 131 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112001314A (en) * | 2020-08-25 | 2020-11-27 | 江苏师范大学 | Early fault detection method for variable speed hoist |
CN113111591A (en) * | 2021-04-29 | 2021-07-13 | 南方电网电力科技股份有限公司 | Automatic diagnosis method, device and equipment based on internal fault of modular power distribution terminal |
CN113111591B (en) * | 2021-04-29 | 2022-06-21 | 南方电网电力科技股份有限公司 | Automatic diagnosis method, device and equipment based on internal fault of modular power distribution terminal |
CN113687610A (en) * | 2021-07-28 | 2021-11-23 | 国网江苏省电力有限公司南京供电分公司 | Method for protecting terminal information of GAN-CNN power monitoring system |
CN113592181A (en) * | 2021-08-02 | 2021-11-02 | 广西大学 | Small hydropower station output prediction method and system |
CN113592181B (en) * | 2021-08-02 | 2023-04-28 | 广西大学 | Small hydro-electric group output prediction method and system |
WO2023098753A1 (en) * | 2021-12-02 | 2023-06-08 | 广东电网有限责任公司江门供电局 | Power distribution terminal fault diagnosis method, system and apparatus, and storage medium |
CN116776130A (en) * | 2023-08-23 | 2023-09-19 | 成都新欣神风电子科技有限公司 | Detection method and device for abnormal circuit signals |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110516742A (en) | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model | |
CN109948597B (en) | High-voltage circuit breaker mechanical fault diagnosis method | |
CN106197999B (en) | A kind of planetary gear method for diagnosing faults | |
CN108832619A (en) | Transient stability evaluation in power system method based on convolutional neural networks | |
CN109917205A (en) | A kind of solenoid valve failure diagnostic device and method based on feature extraction and multi-layer perception (MLP) | |
CN109374270A (en) | A kind of analysis of GIS abnormal vibrations and mechanical fault diagnosis device and method | |
CN107255772B (en) | Semi-supervised voltage sag accident source identification method | |
CN109596942A (en) | A kind of voltage sag reason recognition methods based on depth confidence network | |
CN110070102B (en) | Method for establishing sequence-to-sequence model for identifying power quality disturbance type | |
CN107942210A (en) | The classification of transformer pulse electric current Partial Discharge and denoising method and system | |
CN108562811A (en) | Complicated electrical energy power quality disturbance analysis method based on the memory of two-way shot and long term | |
CN108535635A (en) | A kind of analog circuit intermittent fault diagnostic method based on EEMD and HMM | |
CN106597154B (en) | Transformer fault diagnosis method for improving based on DAG-SVM | |
CN112557966A (en) | Transformer winding looseness identification method based on local mean decomposition and support vector machine | |
Shang et al. | Fault diagnosis method of rolling bearing based on deep belief network | |
CN112083328A (en) | Fault diagnosis method, system and device for high-voltage circuit breaker | |
CN106291162A (en) | A kind of method for diagnosing faults of photovoltaic diode clamp formula three-level inverter | |
Lu et al. | An optimized stacked diagnosis structure for fault diagnosis of wind turbine planetary gearbox | |
Chen et al. | An adversarial learning framework for zero-shot fault recognition of mechanical systems | |
CN208076016U (en) | A kind of GIL On-line Faults monitoring system based on vibration signal support vector machines | |
CN114298080A (en) | Hydro-turbo generator set monitoring method based on throw data mining | |
Shang et al. | Rolling Bearing Fault Diagnosis Methodbased on EEMD and GBDBN | |
Song et al. | State Diagnosis of Elevator Control Transformer over Vibration Signal Based on MEA‐BP Neural Network | |
CN114814391A (en) | Charging pile fault identification method and storage medium | |
Li et al. | A Robust Classification Method for Power Transformer Status Recognition Based on Sound Signals |
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 |