CN103941152B - A kind of cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity - Google Patents
A kind of cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity Download PDFInfo
- Publication number
- CN103941152B CN103941152B CN201410131079.1A CN201410131079A CN103941152B CN 103941152 B CN103941152 B CN 103941152B CN 201410131079 A CN201410131079 A CN 201410131079A CN 103941152 B CN103941152 B CN 103941152B
- Authority
- CN
- China
- Prior art keywords
- fault
- waveform
- similarity
- fault distance
- wavelet
- 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.)
- Active
Links
Landscapes
- Locating Faults (AREA)
Abstract
The present invention is a kind of based on waveform similaritykThe cable mixed line fault distance-finding method of NN algorithm, belongs to Relay Protection Technology in Power System field.The present invention utilizes emulation to obtain zero mould current data under measuring end different faults position and fault condition, it is carried out wavelet decomposition, obtains the wavelet coefficient tissue historical sample data under the 6th yardstick.After there is singlephase earth fault, utilizekNN algorithm, by utilizing correlation analysis to carry out similarity mode the wavelet coefficient waveform under the conditions of different faults in the zero mould current waveform wavelet coefficient waveform after wavelet decomposition of the actual measurement historical sample with analogue simulation, obtaining the fault distance corresponding to the Wave data that first three similarity is the highest, the method further according to the different application recurrence of its weight obtains fault distance.A large amount of emulation show, for singlephase earth fault the method is reliable and precision is higher.
Description
Technical field
The present invention relates to the cable mixed line fault distance-finding method of a kind of k-NN algorithm based on waveform similarity, belong to
Relay Protection Technology in Power System field.
Background technology
Net carries out fault localization after breaking down can alleviate artificial line walking workload, shortens fault correction time, raising
Power supply reliability, minimizing loss of outage and the hidden danger of discovery existence are also acted upon as early as possible, prevent the generation again of fault.
The method of fault localization can be divided into fault analytical method and traveling wave method by range measurement principle.The core of traveling wave method is to measure row ripple at bus
And the propagation time between trouble point calculates line fault distance.Travelling wave ranging generally can be divided into Single Terminal Traveling Wave Fault Location and both-end
Travelling wave ranging two class.Single Terminal Traveling Wave Fault Location need not GPS and realizes the synchronization of data, it is not required that carry out two ends data communication, its
Cost is the half of both-end travelling wave ranging cost, carries out travelling wave ranging by single-ended method, owing to row ripple is anti-in trouble point and bus end
Carry out catadioptric again, reflect that the wavefront of the various character on measurement end time shaft is staggered, to trouble point echo
The demarcation of due in brings great difficulty, and Range finding reliability is difficult to ensure that.And both-end travelling wave ranging only requires two bus ends
The moment that accurately first wavefront of detection arrives, therefore the method by transition resistor electric arc characteristic, line distribution capacitance with
And the impact of load current is less, there is higher reliability for more single-ended method, but it is right to need dual ended data communication and GPS to synchronize
Time equipment, cost of investment is big.
Along with the proportion that the development of urban power distribution network, cable feeder line and cable joint line are shared in power distribution network
Increasing, power distribution network cable line fault has concealed feature mostly, and the lookup of cable fault relatively overhead feeder is more
Difficulty.Cable its particularity compared with aerial line shows: one is that the construction of cable is relative complex, typically by cable core, screen layer,
Many conductor systems composition of the composition such as sheath；Two is that cable is embedded in underground more, the tightst with the earth relation, causes in cable
Electromagnetic transient differ greatly compared with aerial line, the velocity of wave that row ripple is propagated in the cable has the biggest compared with aerial line
Difference.Once break down, power supply reliability can be produced impact undoubtedly.Prior art is all built upon determining abort situation
Carrying out fault location after being in overhead transmission line section or cable sections again, workload is big, and step is complicated, therefore the present invention seeks
Other visual angles are asked to carry out cable joint line singlephase earth fault section location.
Summary of the invention
The technical problem to be solved in the present invention is to propose the cable blend of a kind of k-NN algorithm based on waveform similarity
Road fault distance-finding method, it is not necessary to determining that fault section is in cable run or overhead transmission line, range finding effect is preferable.
The technical scheme is that the cable mixed line fault range finding of a kind of k-NN algorithm based on waveform similarity
Method, utilizes emulation to obtain zero mould current data under measuring end different faults position and fault condition, it is carried out little wavelength-division
Solve, obtain the wavelet coefficient tissue historical sample data under the 6th yardstick；When there is singlephase earth fault, utilize k-NN algorithm,
By by different from the historical sample of analogue simulation for the zero mould current waveform wavelet coefficient waveform after wavelet decomposition of actual measurement
Wavelet coefficient waveform under fault condition utilizes correlation analysis to carry out similarity mode, obtains the waveform that first three similarity is the highest
Fault distance corresponding to data, the method further according to the different application recurrence of its weight obtains fault distance.
Concretely comprise the following steps:
(1) cable joint line actual feeder line parameter can be surveyed according to overhead transmission line top, arrange along the line by certain density
Fault distance and different faults condition, obtain after fault zero mould electric current number in a timing window at measuring end under 1MHz sample rate
According to, and it is carried out 8 layers of wavelet decomposition, choose the wavelet coefficient under the 6th yardstick as historical sample；
(2) by different faults in the measured waveform wavelet coefficient waveform after the wavelet decomposition historical sample with analogue simulation
Under the conditions of wavelet coefficient waveform utilize correlation analysis to carry out similarity mode, i.e. utilize
Calculating both similarities, wherein ρ is correlation coefficient, and x (n) is measured data, and y (n) is sample data, and ρ is the biggest, shows two
Individual waveform is the most similar, obtains the fault distance corresponding to the Wave data that first three similarity is the highest；
(3) the weights D defining fault localization based on waveform similarity is to utilize correlation analysis to measure two signals
Similarity, i.e. correlation coefficient ρ；According to the fault distance representated by the Wave data that first three similarity is the highest and combine its weight
Different application return method carry out fault localization, fault distance xfFormula is
Wherein, x 'fX () is the fault distance representated by Wave data that in historical sample, k similarity is the highest, Dr(x, x') is history
The weight of sample, S (x ') is the set of the Wave data that k similarity is the highest.
The principle of the present invention is: utilize the fault localization of k-NN, and its essence is by finding out the k of a test sample
Neighbour, the impact that this test sample is produced by k arest neighbors gives different weights, obtains test sample by regression algorithm
Fault distance.
The weights D now defining fault localization based on waveform similarity is the phase utilizing correlation analysis to measure two signals
Like spending, the correlation coefficient of x (n) and y (n) is
Wherein ρ is correlation coefficient, i.e. weights D, x (n) is measured data, and y (n) is sample data, and ρ is the biggest, two waveforms
The most similar.
If D is closer to 1, then it represents that test signal is the most similar to training signal.Utilize D as the weight of k arest neighbors, and
The fault distance representated by k arest neighbors is utilized to carry out fault localization, then fault distance xfFor
In formula (2), x 'fX () is the fault distance in training sample representated by k arest neighbors, Dr(x, x') is history sample
This weight, S (x ') is the set of the Wave data that k similarity is the highest.
Assuming that aerial line origin or beginning shown in Fig. 1 can survey cable joint line distance M end 5km, there is singlephase earth fault in 15km,
Its transition resistance is different, is 60 °, is 90 ° during 15km fault during fault initial phase angle 5km fault.Sample rate at 1MHz
At lower measuring end, during 1ms, in window, zero mould current waveform such as Fig. 2 (a) and Fig. 3 (a) is shown.Zero mould current data is carried out 8 layers of small echo
Decomposing, choose the wavelet reconstruction coefficient of the 6th yardstick, wherein, d6 frequency range is 7.8125kHz～15.625kHz.In different events
Under barrier position, shown in wavelet reconstruction time-domain coefficients such as Fig. 2 (b) under measuring end zero mould electric current the 6th yardstick and Fig. 3 (b).
From Fig. 2 and Fig. 3, when the initial phase angle of fault is identical, different faults position, the zero mould electricity that measuring end detects
Stream waveform is different, and same fault distance, under different transition resistances, the similarity of zero mould current waveform is the highest.Therefore, utilize
This characteristic can carry out fault localization, it is achieved the fault section identification problem of cable joint line.Therefore, it can utilize k-NN
Algorithm, by by the wavelet systems under the conditions of measured waveform wavelet coefficient waveform after wavelet decomposition and different faults in Sample Storehouse
Number waveform mates, and obtains the fault distance that first three similarity is the highest, the method returned according to the different application of its weight
Carry out fault localization.
The invention has the beneficial effects as follows:
1. it is positioned at cable or overhead transmission line time without determining fault section, finds range effective；
2. the fault localization utilizing wavelet coefficient similarity to carry out returning is affected by noise little, and method reliability wants height；
3. the method is affected little by system failure initial phase angle and transition resistance, and range measurement precision is higher；
4., merely with single-end information amount, it is not necessary to both sides data syn-chronization, reduce investment outlay.
Accompanying drawing explanation
Fig. 1 is that aerial line origin or beginning can survey cable mixing direct distribution lines system diagram；
Fig. 2 is distance M end 5km, the wavelet coefficient under zero mould current waveform and the 6th yardstick under different transition resistances；
Fig. 3 is distance M end 15km, the wavelet coefficient under zero mould current waveform and the 6th yardstick under different transition resistances.
Detailed description of the invention
Below in conjunction with the accompanying drawings and detailed description of the invention, the invention will be further described.
The cable mixed line fault distance-finding method of a kind of k-NN algorithm based on waveform similarity, utilizes emulation acquisition amount
Surveying zero mould current data under end different faults position and fault condition, it is carried out wavelet decomposition, obtain under the 6th yardstick is little
Wave system array knits historical sample data；When there is singlephase earth fault, utilize k-NN algorithm, by the zero mould electric current by actual measurement
Wavelet coefficient under the conditions of different faults in the historical sample of waveform wavelet coefficient waveform after wavelet decomposition and analogue simulation
Waveform utilizes correlation analysis to carry out similarity mode, obtain fault corresponding to the Wave data that first three similarity is the highest away from
From, the method further according to the different application recurrence of its weight obtains fault distance.
Concretely comprise the following steps:
(1) cable joint line actual feeder line parameter can be surveyed according to overhead transmission line top, arrange along the line by certain density
Fault distance and different faults condition, obtain after fault zero mould electric current number in a timing window at measuring end under 1MHz sample rate
According to, and it is carried out 8 layers of wavelet decomposition, choose the wavelet coefficient under the 6th yardstick as historical sample；
(2) by different faults in the measured waveform wavelet coefficient waveform after the wavelet decomposition historical sample with analogue simulation
Under the conditions of wavelet coefficient waveform utilize correlation analysis to carry out similarity mode, i.e. utilizeCome
Calculating both similarities, wherein ρ is correlation coefficient, and x (n) is measured data, and y (n) is sample data, and ρ is the biggest, shows two
Waveform is the most similar, obtains the fault distance corresponding to the Wave data that first three similarity is the highest；
(3) the weights D defining fault localization based on waveform similarity is to utilize correlation analysis to measure two signals
Similarity, i.e. correlation coefficient ρ；According to the fault distance representated by the Wave data that first three similarity is the highest and combine its weight
Different application return method carry out fault localization, fault distance xfFormula isIts
In, x 'fX () is the fault distance representated by Wave data that in historical sample, k similarity is the highest, Dr(x, x') is history sample
This weight, S (x ') is the set of the Wave data that k similarity is the highest.
Now set the aerial line origin or beginning shown in Fig. 1 and can survey cable mixing direct distribution lines system diagram, wherein in cable joint line
The long 10km of overhead feeder, the long 10km of cable feeder line, perfecting circuit is the long cable of 2km and the long overhead feeder of 25km, installs one at M end
Group traveling wave fault location device.
Now make somebody a mere figurehead origin or beginning for outgoing lines shown in Fig. 1 and can survey topological structure tissue historical sample data.Sample rate is 1MHz,
Zero mould current data in window during 1ms after acquisition fault, fault is provided that
1., from the beginning of at distance M end 0.5km, A-G fault is set every 0.5km；
2. the initial phase angle of fault is respectively 60 ° and 90 °；
3. overhead transmission line section transition resistance is respectively 0 Ω, 30 Ω, 50 Ω and 70 Ω, and cable sections transition resistance is respectively
Be 0 Ω, 10 Ω and 30 Ω.
Utilize measuring end zero mould current data under above-mentioned fault condition, it is carried out 8 layers of wavelet decomposition, chooses the 6th yardstick
Wavelet coefficient tissue historical sample data.
Embodiment 1: it is now assumed that distance M end 8.75km basic routing line generation single-phase earthing metallic fault, the initial phase angle of fault
Being 90 °, fault resistance is 70 Ω, and emulation sample frequency is 1MHz, utilizes the zero mould current wave that M end range unit gathers
In the historical sample of shape, the wavelet coefficient waveform obtained after wavelet function feedback and analogue simulation under the conditions of different faults
Wavelet coefficient waveform utilizes correlation analysis formula (1) to carry out Similarity Measure, obtains the Wave data that first three similarity is the highest
Corresponding fault distance.Define the weights D of fault localization based on waveform similarity for utilizing correlation analysis to measure two
The similarity of signal.According to the fault distance representated by the Wave data that first three similarity is the highest the difference that combines its weight
The method i.e. formula (2) that application returns carries out fault localization, and providing abort situation is distance M end 8.83km at backbone.
Embodiment 2: it is now assumed that distance M end 13.65km basic routing line generation single-phase earthing metallic fault, the initial phase angle of fault
Being 60 °, fault resistance is 30 Ω, and emulation sample frequency is 1MHz, utilizes the zero mould current wave that M end range unit gathers
In the historical sample of shape, the wavelet coefficient waveform obtained after wavelet function feedback and analogue simulation under the conditions of different faults
Wavelet coefficient waveform utilizes correlation analysis formula (1) to carry out Similarity Measure, obtains the Wave data that first three similarity is the highest
Corresponding fault distance.Define the weights D of fault localization based on waveform similarity for utilizing correlation analysis to measure two
The similarity of signal.According to the fault distance representated by the Wave data that first three similarity is the highest the difference that combines its weight
The method i.e. formula (2) that application returns carries out fault localization, and providing abort situation is distance M end 13.83km at backbone.
Above in conjunction with accompanying drawing, the detailed description of the invention of the present invention is explained in detail, but the present invention is not limited to above-mentioned
Embodiment, in the ken that those of ordinary skill in the art are possessed, it is also possible to before without departing from present inventive concept
Put that various changes can be made.
Claims (2)
1. the cable mixed line fault distance-finding method of a k-NN algorithm based on waveform similarity, it is characterised in that: utilize
Emulation obtains zero mould current data under measuring end different faults position and fault condition, it is carried out wavelet decomposition, obtains the 6th
Wavelet coefficient tissue historical sample data under yardstick；When there is singlephase earth fault, utilize k-NN algorithm, by surveying
Zero mould current waveform wavelet coefficient waveform after wavelet decomposition and different faults in the historical sample of analogue simulation under the conditions of
Wavelet coefficient waveform utilize correlation analysis to carry out similarity mode, obtain corresponding to the Wave data that first three similarity is the highest
Fault distance, further according to its weight different application return method obtain fault distance.
The cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity the most according to claim 1,
It is characterized in that concretely comprising the following steps:
(1) cable joint line actual feeder line parameter can be surveyed according to overhead transmission line top, fault is set along the line by certain density
Distance and different faults condition, obtain after fault zero mould current data in a timing window at measuring end under 1MHz sample rate,
And it is carried out 8 layers of wavelet decomposition, choose the wavelet coefficient under the 6th yardstick as historical sample；
(2) by different faults condition in the measured waveform wavelet coefficient waveform after the wavelet decomposition historical sample with analogue simulation
Under wavelet coefficient waveform utilize correlation analysis to carry out similarity mode, i.e. utilize
Calculating both similarities, wherein ρ is correlation coefficient, and x (n) is measured data, and y (n) is sample data, and ρ is the biggest, shows two
Individual waveform is the most similar, obtains the fault distance corresponding to the Wave data that first three similarity is the highest；
(3) the weights D defining fault localization based on waveform similarity is to utilize correlation analysis to measure the similar of two signals
Degree, i.e. correlation coefficient ρ；According to the fault distance representated by the Wave data that first three similarity is the highest and combine its weight not
The method returned with application carries out fault localization, fault distance xfFormula isIts
In, x 'fX () is the fault distance representated by Wave data that in historical sample, k similarity is the highest, Dr(x, x') is history sample
This weight, S (x ') is the set of the Wave data that k similarity is the highest.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410131079.1A CN103941152B (en) | 2014-04-03 | 2014-04-03 | A kind of cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410131079.1A CN103941152B (en) | 2014-04-03 | 2014-04-03 | A kind of cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103941152A CN103941152A (en) | 2014-07-23 |
| CN103941152B true CN103941152B (en) | 2016-09-14 |
Family
ID=51188901
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410131079.1A Active CN103941152B (en) | 2014-04-03 | 2014-04-03 | A kind of cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103941152B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105974264B (en) * | 2016-04-28 | 2019-10-11 | 长沙理工大学 | Fault line selection method based on phase current transient characteristics |
| CN107121617B (en) * | 2017-03-09 | 2019-09-27 | 昆明理工大学 | A kind of DC power transmission line k nearest neighbor distance measuring method using filter branches electric current |
| CN109633373B (en) * | 2018-12-18 | 2023-09-22 | 中国电力科学研究院有限公司 | Method and device for accurately positioning faults in power distribution network |
| CN110703012B (en) * | 2019-09-26 | 2021-10-15 | 国电南瑞科技股份有限公司 | Distributed fault diagnosis method for power transmission line |
| CN111987699B (en) * | 2020-08-18 | 2021-08-27 | 广东电网有限责任公司电力调度控制中心 | Current differential protection data synchronization method of alternating current line and related device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101162838A (en) * | 2007-11-29 | 2008-04-16 | 昆明理工大学 | Low current neutral grounding system fault route selecting method by wavelet package decompose and correlation analysis |
| CN101404408A (en) * | 2008-11-13 | 2009-04-08 | 昆明理工大学 | Distribution network cable-wire mixed line failure route selection method by utilizing relative energy |
| CN102721741A (en) * | 2012-06-18 | 2012-10-10 | 北京科技大学 | Wind power blade damage monitoring and positioning system based on wireless acoustic emission sensor network |
| CN102788926A (en) * | 2012-07-04 | 2012-11-21 | 河南理工大学 | Single-phase ground fault section positioning method of small-current ground system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2487766B1 (en) * | 2011-02-11 | 2015-08-26 | ABB Technology AG | Method and apparatus for detecting earth fault |
-
2014
- 2014-04-03 CN CN201410131079.1A patent/CN103941152B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101162838A (en) * | 2007-11-29 | 2008-04-16 | 昆明理工大学 | Low current neutral grounding system fault route selecting method by wavelet package decompose and correlation analysis |
| CN101404408A (en) * | 2008-11-13 | 2009-04-08 | 昆明理工大学 | Distribution network cable-wire mixed line failure route selection method by utilizing relative energy |
| CN102721741A (en) * | 2012-06-18 | 2012-10-10 | 北京科技大学 | Wind power blade damage monitoring and positioning system based on wireless acoustic emission sensor network |
| CN102788926A (en) * | 2012-07-04 | 2012-11-21 | 河南理工大学 | Single-phase ground fault section positioning method of small-current ground system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103941152A (en) | 2014-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103941152B (en) | A kind of cable mixed line fault distance-finding method of k-NN algorithm based on waveform similarity | |
| Suonan et al. | A novel fault-location method for HVDC transmission lines | |
| CN106959401B (en) | A kind of distribution travelling wave ranging method | |
| CN105353268B (en) | One kind is used for the judgement of transmission line of electricity distribution traveling wave fault and localization method | |
| CN103941153B (en) | A kind of outgoing lines radiation network fault distance-finding method of k-NN algorithm based on waveform similarity | |
| CN101299538B (en) | Cable-aerial mixed line fault travelling wave ranging method | |
| CN107505538B (en) | The asynchronous Fault Locating Method of half-wave power transmission circuit based on line mode voltage traveling wave amplitude attenuation characteristic | |
| CN104865495B (en) | Feeder line fault travelling wave positioning method and system | |
| CN101672883B (en) | Overhead and cable mixed power transmission line fault locating method and device | |
| CN102074937B (en) | Ranging method for adaptive transmission line fault of model | |
| CN108896874B (en) | A kind of fault positioning method for transmission line of end connection short-term road | |
| CN104898021B (en) | A kind of distribution network fault line selection method based on k means cluster analyses | |
| CN105372559B (en) | A kind of limited continuation method of chain type transmission line of electricity one-end fault ranging | |
| CN105548802B (en) | A kind of asynchronous fault distance-finding method in three end of T link based on distribution character along fault traveling wave | |
| CN103592575A (en) | Self-adaptation weighting data fusion fault distance measurement method based on multi-sensor system | |
| CN105182184A (en) | Distribution network cable hybrid direct distribution line fault ranging method based on line side direction traveling wave decomposition | |
| CN103424667B (en) | The distance-finding method that a kind of cable mixed line fault ripple mates to sequential Simulation after test | |
| CN105403812B (en) | It is a kind of based on decomposing along fault traveling wave and the triangle loop grid one-end fault ranging method of distance calibration | |
| CN106019080A (en) | Line-side energy mutation based single-end travelling wave fault location method for double DC circuits on same tower | |
| CN105375452A (en) | Fault transient protection method for electric transmission line | |
| CN105842584B (en) | A kind of T link multiterminal travelling wave ranging method based on distribution character along fault traveling wave | |
| Gaur et al. | New ground fault location method for three-terminal transmission line using unsynchronized current measurements | |
| CN108845233A (en) | Electric distribution network overhead wire single-phase earthing double-end monitor localization method | |
| RU2532760C1 (en) | Method for determining damaged point of branched power transmission line | |
| CN107121617A (en) | A kind of direct current transmission line fault distance-finding method of use filter branches electric current and K k-nearest neighbors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |