CN109241944A - A kind of distribution network failure recognition methods based on improvement multi-category support vector machines - Google Patents
A kind of distribution network failure recognition methods based on improvement multi-category support vector machines Download PDFInfo
- Publication number
- CN109241944A CN109241944A CN201811175194.3A CN201811175194A CN109241944A CN 109241944 A CN109241944 A CN 109241944A CN 201811175194 A CN201811175194 A CN 201811175194A CN 109241944 A CN109241944 A CN 109241944A
- Authority
- CN
- China
- Prior art keywords
- support vector
- distribution network
- failure
- phase current
- recognition methods
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06Q10/20—Administration of product repair or maintenance
-
- 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
Abstract
The present invention relates to a kind of based on the distribution network failure recognition methods for improving multi-category support vector machines, it is characterized by comprising following steps: step S1: the main transformer low-pressure side three-phase current and bus residual voltage simulation waveform data of the acquisition latter cycle of failure, as input signal；Step S2: wavelet decomposition processing is carried out to input signal, and reconstructs low frequency component, obtains reconstruction signal；Step S3: using the method for seeking root mean square and Euclidean distance, the feature vector of reconstruction signal is extracted；Step S4: multistage supporting vector model is constructed, and optimized parameter, the Multistage Support Vector Machine model after being trained are found based on Radial basis kernel function；Step S5: the feature vector of reconstruction signal is input to the Multistage Support Vector Machine model after training, obtains failure modes.The present invention is based on multi-category support vector machines are improved, the accuracy rate of failure modes, fault recognition rate are greatly improved.
Description
Technical field
The present invention relates to field of distribution network, and in particular to a kind of based on the distribution network failure for improving multi-category support vector machines
Recognition methods.
Background technique
Modern electric development expands power distribution network scale constantly, structure is increasingly complicated, and all kinds of failures can inevitably occur.Accurately
It identifies simultaneously localization process failure, forced outage coverage can be effectively reduced, improve system run all right.In general, detecting
Failure needs identification of defective type after occurring first, then and selects faulty line, repositions fault section.Different faults type makes
Fault Locating Method have it is different, therefore, accurately identify failure be one of decisive premise of distribution network failure Position Research.
Fault identification can be realized based on stable state or transient state electrical quantity.Wherein, event of the tradition based on power frequency steady-state quantity and setting threshold value
Hinder recognition methods, the factors such as moment, position of failure point, short circuiting transfer resistance are occurred by failure and are affected, are had centainly
Limitation；Based on the fault recognition method of transient state electrical quantity, the characteristic quantity of energy characterization failure type is often first extracted, then is passed through
Classification is completed in pattern-recognition, and general high sensitivity is highly reliable, and is not influenced by arc suppression coil, is realized simple and quick.
As a kind of stronger mode identification method of generalization ability, support vector machines is widely used to distribution network failure class
Type identification.Support vector machines generallys use two classification mode recognition methods, and for that can efficiently accomplish polymorphic type identification, it is more need to extend it
Classification feature.Typical SVM is that a kind of two classification mode recognition methods need to be to its function when being used in the identification of more classification modes
It can be carried out extension.Multi-category support vector machines common method has 1-against-1 method, 1-against-rest method and decision tree method
Deng.The Multistage Support Vector Machine of existing distribution network failure identification mostly uses decision tree method to build, and is easy to appear wrong cumulated downward,
And be affected by different decision paths, it may cause discrimination reduction.It is therefore desirable to be improved to existing method, improve
The accuracy rate of failure modes.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of based on the distribution network failure for improving multi-category support vector machines
Recognition methods improves existing method, improves the accuracy rate of failure modes.
To achieve the above object, the present invention adopts the following technical scheme:
A kind of distribution network failure recognition methods based on improvement multi-category support vector machines, it is characterised in that: including following
Step:
Step S1: the main transformer low-pressure side three-phase current and bus residual voltage simulation waveform number of the acquisition latter cycle of failure
According to as input signal；
Step S2: wavelet decomposition processing is carried out to input signal, and reconstructs low frequency component, obtains reconstruction signal；
Step S3: using the method for seeking root mean square and Euclidean distance, the feature vector of reconstruction signal is extracted；
Step S4: multistage supporting vector model is constructed, and optimized parameter is found based on Radial basis kernel function, after being trained
Multistage Support Vector Machine model；
Step S5: the feature vector of reconstruction signal is input to the Multistage Support Vector Machine model after training, obtains failure
Classification.
Further, the step S1 specifically:
Step S11: 10kV electricity distribution network model is built using PSCAD/EMTDC simulation software；
Step S12: 10 kinds of main transformer low-voltage bus bar residual voltage and main transformer low-pressure side three-phase current are obtained by the model
The simulation waveform data of each cycle before and after failure (AG, BG, CG, ABG, ACG, BCG, AB, AC, BC, ABC)；
Step S13: according to obtained simulation waveform data, extract the latter cycle of failure main transformer low-pressure side three-phase current and
Bus residual voltage simulation waveform data, as input signal.
Further, the step S2 specifically:
Step S21: the good db4 of regularity is chosen as wavelet basis function, Decomposition order is 2 layers；
Step S22: according to wavelet basis function change to input signal carry out 2 layers decomposition, and to the wavelet coefficient of the second layer into
Row reconstruct, obtains reconstruction signal.
Further, the step S3 specifically:
Step S31: the square of three-phase current and residual voltage is sought respectively using the second layer approximation component of reconstruction signal
The alternate Euclidean distance d of root R and three-phase current；
Step S32: according to the alternate Euclidean distance d for the root mean square R and three-phase current for seeking three-phase current and residual voltage,
After being normalized, the feature vector of reconstruction signal is obtained.
Further, the multistage supporting vector model is to improve SVM models of classifying more, which is made of 10 SVM,
It is denoted as SVM1-SVM10.Model overall structure uses decision tree schema, but wherein SVM4~SVM6 uses 1-agai nst-1 method
It votes, failure phase is chosen in a manner of voting adopted.
Compared with the prior art, the invention has the following beneficial effects:
The present invention is based on the present invention is based on multi-category support vector machines are improved, the accurate of failure modes is greatly improved
Rate, fault recognition rate.
Detailed description of the invention
Fig. 1 is the method for the present invention flow chart；
Fig. 2 is the improvement multi-category support vector machines structure chart designed in one embodiment of the invention；
Fig. 3 is the software phantom figure of 10kV power distribution network in one embodiment of the invention.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
Fig. 1 is please referred to, the present invention provides a kind of distribution network failure recognition methods based on improvement multi-category support vector machines,
It is characterized by comprising following steps:
Step S1: the main transformer low-pressure side three-phase current and bus residual voltage simulation waveform number of the acquisition latter cycle of failure
According to as input signal；
Step S2: wavelet decomposition processing is carried out to input signal, and reconstructs low frequency component, obtains reconstruction signal；
Step S3: using the method for seeking root mean square and Euclidean distance, the feature vector of reconstruction signal is extracted；
Step S4: multistage supporting vector model is constructed, and optimized parameter is found based on Radial basis kernel function, after being trained
Multistage Support Vector Machine model；
Step S5: the feature vector of reconstruction signal is input to the Multistage Support Vector Machine model after training, obtains failure
Classification.
In an embodiment of the present invention, a radiant type medium-voltage distribution pessimistic concurrency control is established, main transformer low pressure is obtained by the model
10 kinds of failures (AG, BG, CG, ABG, ACG, BCG, AB, AC, BC, ABC) of bus residual voltage and main transformer low-pressure side three-phase current
The simulation waveform data of each cycle in front and back；According to principle of stacking, when power distribution network breaks down, fault current is when operating normally
Load current is superimposed with fault component electric current under fault condition.Since there are frequency departures for system, after only extracting failure
The fault component of one cycle is as data to be analyzed.
In an embodiment of the present invention, in step s 2, the good db4 of regularity is chosen as wavelet basis function, application
Wavelet transformation carries out 2 layers of decomposition to input signal, and the wavelet coefficient of the second layer is reconstructed.
In an embodiment of the present invention, in step s3,7 fault feature vectors are constructed based on root mean square and Euclidean distanceThe reconstruction signal root mean square R of three-phase current and residual voltage fault component meter
It calculates are as follows:
Wherein, N is the sampling number in a power frequency period.Herein, sample frequency 10kHz, then N=200.
The alternate Euclidean distance d of three-phase current is calculated are as follows:
X=a, b, c；Y=b, c, a
In order to facilitate observation of, characteristic quantity is normalized:
Feature vector is input to and has been trained in perfect Multistage Support Vector Machine model, model is as shown in Figure 2.
Embodiment 1:
In the present embodiment, electrical quantity signal is obtained using the 10kV power distribution network software phantom that simulation software is built,
Wavelet decomposition is carried out to the fault component of feeder line three-phase current and the latter cycle of bus residual voltage failure, and reconstructs the second layer
Approximation component, the root mean square for seeking reconstruction signal and Euclidean distance are input to as feature vector and improve more class Support Vectors
Machine completes distribution network failure classification.Wherein, training sample is 1080, and test sample is 6480.
Distribution network failure classifying step are as follows:
(1) acquisition of feature vector
The above-mentioned technical proposal provided according to the present invention, the main transformer low-pressure side three-phase current of each cycle in interception failure front and back
With the simulation waveform of bus residual voltage, and the fault component of the latter cycle of failure is sought.Selection db4 is wavelet basis function, right
Each waveform carries out 2 layers of wavelet decomposition respectively, and reconstructs the low frequency component of the second layer, based on root mean square and Euclidean distance construction 7
A fault feature vector.
(2) failure modes
The feature vector of the above-mentioned technical proposal provided according to the present invention, input fault disaggregated model is
Building improves more classification SVM models as shown in Fig. 2, the model is made of 10 SVM, is denoted as SVM1-SVM10, mould
Type overall structure uses decision tree schema, but wherein SVM4~SVM6 is voted using 1-agai nst-1 method, with voting adopted side
Formula chooses failure phase.Final output fault type tag number corresponds to 10 kinds of fault types.
Classification results: classification accuracy rate is up to 98.2% or more.
The foregoing is merely presently preferred embodiments of the present invention, all equivalent changes done according to scope of the present invention patent with
Modification, is all covered by the present invention.
Claims (5)
1. a kind of based on the distribution network failure recognition methods for improving multi-category support vector machines, it is characterised in that: including following step
It is rapid:
Step S1: the main transformer low-pressure side three-phase current and bus residual voltage simulation waveform data of the acquisition latter cycle of failure are made
For input signal；
Step S2: wavelet decomposition processing is carried out to input signal, and reconstructs low frequency component, obtains reconstruction signal；
Step S3: using the method for seeking root mean square and Euclidean distance, the feature vector of reconstruction signal is extracted；
Step S4: constructing multistage supporting vector model, and find optimized parameter based on Radial basis kernel function, more after being trained
Grade supporting vector machine model；
Step S5: the feature vector of reconstruction signal is input to the Multistage Support Vector Machine model after training, obtains failure modes.
2. a kind of distribution network failure recognition methods based on improvement multi-category support vector machines according to claim 1,
It is characterized in that: the step S1 specifically:
Step S11: 10kV electricity distribution network model is built using PSCAD/EMTDC simulation software;
Step S12: before and after the failure for obtaining main transformer low-voltage bus bar residual voltage and main transformer low-pressure side three-phase current by the model
The simulation waveform data of each cycle;
Step S13: according to obtained simulation waveform data, the main transformer low-pressure side three-phase current and bus of the latter cycle of failure are extracted
Residual voltage simulation waveform data, as input signal.
3. a kind of distribution network failure recognition methods based on improvement multi-category support vector machines according to claim 1,
It is characterized in that: the step S2 specifically:
Step S21: the good db4 of regularity is chosen as wavelet basis function, Decomposition order is 2 layers；
Step S22: it is changed according to wavelet basis function and 2 layers of decomposition is carried out to input signal, and weight is carried out to the wavelet coefficient of the second layer
Structure obtains reconstruction signal.
4. a kind of distribution network failure recognition methods based on improvement multi-category support vector machines according to claim 1,
It is characterized in that: the step S3 specifically:
Step S31: using the second layer approximation component of reconstruction signal seek respectively three-phase current and residual voltage root mean square R and
The alternate Euclidean distance d of three-phase current；
Step S32: it according to the alternate Euclidean distance d for the root mean square R and three-phase current for seeking three-phase current and residual voltage, carries out
After normalized, the feature vector of reconstruction signal is obtained.
5. a kind of distribution network failure recognition methods based on improvement multi-category support vector machines according to claim 1,
Be characterized in that: to improve SVM models of classifying, which is made of 10 SVM, is denoted as the multistage supporting vector model more
SVM1-SVM10。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2018110443305 | 2018-09-07 | ||
| CN201811044330 | 2018-09-07 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109241944A true CN109241944A (en) | 2019-01-18 |
Family
ID=65055201
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811175194.3A Pending CN109241944A (en) | 2018-09-07 | 2018-10-10 | A kind of distribution network failure recognition methods based on improvement multi-category support vector machines |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109241944A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112014773A (en) * | 2020-09-04 | 2020-12-01 | 内蒙古电力（集团）有限责任公司呼和浩特供电局 | Method for detecting early fault of small current grounding system cable |
| CN112383052A (en) * | 2020-11-16 | 2021-02-19 | 国网电子商务有限公司 | Power grid fault repairing method and device based on power internet of things |
| CN112462193A (en) * | 2020-11-05 | 2021-03-09 | 重庆邮电大学 | Power distribution network automatic reclosing judgment method based on real-time fault filtering data |
| CN113311219A (en) * | 2021-03-11 | 2021-08-27 | 国网福建省电力有限公司 | Power distribution network temporary overvoltage identification method |
| CN113762230A (en) * | 2021-11-10 | 2021-12-07 | 广东电网有限责任公司佛山供电局 | Cable tunnel accessory facility fault automatic identification method and related device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090157573A1 (en) * | 2006-01-23 | 2009-06-18 | The Trustees Of Columbia University In The City Of New York | System And Method For Grading Electricity Distribution Network Feeders Susceptible To Impending Failure |
| CN106226635A (en) * | 2016-07-14 | 2016-12-14 | 国网福建晋江市供电有限公司 | A kind of distribution feeder fault type recognition method and device |
| CN106249101A (en) * | 2016-06-30 | 2016-12-21 | 湖南大学 | A kind of intelligent distribution network fault identification method |
| CN108279364A (en) * | 2018-01-30 | 2018-07-13 | 福州大学 | Wire selection method for power distribution network single phase earthing failure based on convolutional neural networks |
-
2018
- 2018-10-10 CN CN201811175194.3A patent/CN109241944A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090157573A1 (en) * | 2006-01-23 | 2009-06-18 | The Trustees Of Columbia University In The City Of New York | System And Method For Grading Electricity Distribution Network Feeders Susceptible To Impending Failure |
| CN106249101A (en) * | 2016-06-30 | 2016-12-21 | 湖南大学 | A kind of intelligent distribution network fault identification method |
| CN106226635A (en) * | 2016-07-14 | 2016-12-14 | 国网福建晋江市供电有限公司 | A kind of distribution feeder fault type recognition method and device |
| CN108279364A (en) * | 2018-01-30 | 2018-07-13 | 福州大学 | Wire selection method for power distribution network single phase earthing failure based on convolutional neural networks |
Non-Patent Citations (3)
| Title |
|---|
| MOU-FA GUO: "Deep-Learning-Based Earth Fault Detection Using Continuous Wavelet Transform and Convolutional Neural Network in Resonant Grounding Distribution Systems", 《IEEE SENSORS JOURNAL》 * |
| 张碧慧等: "中压配电线路故障在线监测与识别", 《电气技术》 * |
| 郭谋发: "基于LCD_Hilbert谱奇异值和多级支持向量机的配电网故障识别方法", 《高电压技术》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112014773A (en) * | 2020-09-04 | 2020-12-01 | 内蒙古电力（集团）有限责任公司呼和浩特供电局 | Method for detecting early fault of small current grounding system cable |
| CN112014773B (en) * | 2020-09-04 | 2023-05-02 | 内蒙古电力（集团）有限责任公司呼和浩特供电局 | Method for detecting early fault of small-current grounding system cable |
| CN112462193A (en) * | 2020-11-05 | 2021-03-09 | 重庆邮电大学 | Power distribution network automatic reclosing judgment method based on real-time fault filtering data |
| CN112462193B (en) * | 2020-11-05 | 2023-10-24 | 重庆邮电大学 | Automatic reclosing judgment method for power distribution network based on real-time fault filtering data |
| CN112383052A (en) * | 2020-11-16 | 2021-02-19 | 国网电子商务有限公司 | Power grid fault repairing method and device based on power internet of things |
| CN113311219A (en) * | 2021-03-11 | 2021-08-27 | 国网福建省电力有限公司 | Power distribution network temporary overvoltage identification method |
| CN113311219B (en) * | 2021-03-11 | 2022-11-08 | 国网福建省电力有限公司 | Power distribution network temporary overvoltage identification method |
| CN113762230A (en) * | 2021-11-10 | 2021-12-07 | 广东电网有限责任公司佛山供电局 | Cable tunnel accessory facility fault automatic identification method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109241944A (en) | A kind of distribution network failure recognition methods based on improvement multi-category support vector machines | |
| CN110596533B (en) | Power distribution network single-phase earth fault section positioning method and system | |
| CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
| CN110609200B (en) | Power distribution network earth fault protection method based on fuzzy metric fusion criterion | |
| CN111679158A (en) | Power distribution network fault identification method based on synchronous measurement data similarity | |
| CN106330095B (en) | Collect system fault locating method in a kind of large-sized photovoltaic power station | |
| CN111614066B (en) | Automatic setting method and system for relay protection setting value of power distribution network | |
| CN107037322A (en) | Power distribution network low current grounding localization method based on steady state characteristic | |
| CN110672905A (en) | CNN-based self-supervision voltage sag source identification method | |
| CN114384374A (en) | Fault study and judgment method and device based on edge calculation FTU and fault indicator | |
| Zhu et al. | Faulty line identification method based on bayesian optimization for distribution network | |
| Ni et al. | Faulty feeder detection of single phase-to-ground fault for distribution networks based on improved K-means power angle clustering analysis | |
| CN104778348A (en) | Voltage sag action characteristic modeling method for low-voltage trip | |
| Sapountzoglou et al. | Fault detection and localization in LV smart grids | |
| Hong et al. | Deep‐belief‐Networks based fault classification in power distribution networks | |
| CN103823158A (en) | Resonant earthed system fault line selection method with invariant moments | |
| CN106646106B (en) | Electric network fault detection method based on outlier's detection technology | |
| CN106646138B (en) | Distribution net work earthing fault localization method based on the conversion of more sample frequency wavelet character energy | |
| CN110212503B (en) | High-resistance grounding protection method suitable for small-resistance grounding system | |
| CN114002550B (en) | Direct-current power distribution network ground fault line selection method and system | |
| CN113092934B (en) | Single-phase earth fault judgment method and system based on clustering and LSTM | |
| Xue et al. | An efficient and robust case sorting algorithm for transient stability assessment | |
| CN110988590B (en) | PCA-SVM model-based distribution network line selection method and system | |
| CN111614083B (en) | Big data analysis method suitable for 400V power supply network topology identification | |
| Zhang et al. | Single-phase-to-ground fault diagnosis based on waveform feature extraction and matrix analysis |
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: 20190118 |