CN103163430A - Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network) - Google Patents

Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network) Download PDF

Info

Publication number
CN103163430A
CN103163430A CN2013101071086A CN201310107108A CN103163430A CN 103163430 A CN103163430 A CN 103163430A CN 2013101071086 A CN2013101071086 A CN 2013101071086A CN 201310107108 A CN201310107108 A CN 201310107108A CN 103163430 A CN103163430 A CN 103163430A
Authority
CN
China
Prior art keywords
feeder line
fault
neural network
line selection
energy
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
Application number
CN2013101071086A
Other languages
Chinese (zh)
Inventor
束洪春
高利
段锐敏
朱梦梦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Kunming University of Science and Technology
Original Assignee
Kunming University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Kunming University of Science and Technology filed Critical Kunming University of Science and Technology
Priority to CN2013101071086A priority Critical patent/CN103163430A/en
Publication of CN103163430A publication Critical patent/CN103163430A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Locating Faults (AREA)

Abstract

The invention relates to a resonant grounding system fault line selection method by combining complex wavelets with an ANN (artificial neural network), and belongs to the technical field of power system relay protection. The method includes the steps: immediately starting a fault line selection device and recording faults when a momentary value of bus zero-sequence voltage is out-of-limit to acquire zero-sequence transient current of each feeder line; and transforming and decomposing the zero-sequence transient current of a time window by the aid of the complex wavelets after each feeder line is faulted for 5ms, dividing frequency bands, selecting the frequency band with the maximum energy sum of all feeder lines as a characteristic frequency band according to the maximum energy sum principle, selecting a phase corresponding to the center frequency of the energy sum characteristic frequency band of each feeder line in the characteristic frequency band as a training sample set, determining the number of nodes of an input layer, an output layer and a hidden layer, selecting a transfer function and a learning rule, setting proper neural network parameters, acquiring a fault line selection network by training and adaptively selecting fault lines. A large amount of simulation indicates that the resonant grounding system fault line selection method is accurate and reliable in line selection.

Description

A kind of malfunction route selection method for resonant grounded system that utilizes multiple small echo and ANN combination
Technical field
The present invention relates to the Relay Protection Technology in Power System field, specifically a kind of resonant earthed system failure line selection new method of utilizing multiple small echo and ANN combination.
Background technology
The power distribution network broad covered area, and directly provide the electricity consumption service in the face of the user for it.Account for 80% of distribution network failure according to the statistics singlephase earth fault.The power distribution network resonant earthed system is the neutral by arc extinction coil grounding system, belongs to small current neutral grounding system.Single-phase grounded malfunction in grounded system of low current can affect the healthy phases voltage-to-ground and cause its rising, and voltage raises to produce the insulation of grid equipment and destroys; Intermittent arcing ground particularly, can cause arc overvoltage, this voltage is by destroying system insulation and then developing into alternate or the multipoint earthing short circuit, cause system overvoltage, thereby damage equipment, destroy system safety operation, therefore must find accurately, fast faulty line and in time fault be got rid of.During neutral by arc extinction coil grounding system generation singlephase earth fault, carry out route selection if use power frequency steady-state quantity colony amplitude comparison phase comparing method, due to the impact that is subjected to the arc suppression coil compensation effect, ground current is faint, may cause the route selection mistake.And after fault, the fault transient state current amplitude much larger than steady-state current, and is not affected by arc suppression coil, and the selection method Billy who therefore utilizes transient has more advantage with the selection method of steady-state quantity.But no matter utilize steady-state quantity or transient route selection, all exist some fault signature apparent in view, some fault signature is fuzzyyer, disturbing factor is larger on some fault signature impact, some fault signature is affected the problems such as less, therefore utilize single route selection criterion also the situation of falsely dropping can occur.
Multiple small echo is that a series of female small echos are the basis function of plural number, and its Wavelet Transform Parameters is also plural number, can obtain simultaneously signal amplitude information and phase information thus.Multiple small echo series commonly used has multiple Gauss, multiple Shannon, multiple Morlet and complex frequency B spline wavelets.That multiple Gauss wavelet has is nonopiate, the character of biorthogonal, non-tight and Perfect Reconstruction.The advantage of multiple small echo maximum is that it can extract the phase information of signal.
The ANN artificial neural network is a kind of self-adaptation nonlinear dynamic system, and he is interconnected by a large amount of simple neurons and forms.The principal character of artificial neural network is: large-scale information parallel processing capability and distributed information storage function, extremely strong self-study, association, fault-tolerant ability and good adaptivity, self-organization are the nonlinear system of many inputs and many outputs.
Summary of the invention
The objective of the invention is to propose a kind of malfunction route selection method for resonant grounded system that utilizes multiple small echo and ANN combination, overcome the deficiency of existing malfunction route selection method for resonant grounded system.
The present invention utilizes the malfunction route selection method for resonant grounded system of multiple small echo and ANN combination, carries out as follows:
(1) when bus residual voltage instantaneous value is out-of-limit, namely
Figure 341900DEST_PATH_IMAGE002
The time, fault line selection device starts immediately and records ripple, obtains each feeder line zero sequence transient current, all feeder line zero sequence transient currents of window when recording after fault 5ms; Wherein, Be the bus rated voltage,
Figure 660065DEST_PATH_IMAGE004
=0.15;
(2) each zero sequence transient current is carried out the multiple Gauss Wavelet Transform in 20 rank, decomposing the number of plies is 256 layers; Choose the decomposition result that decomposition scale is the 4-203 layer, divide according to frequency band of 20 yardsticks, obtain the energy of 10 frequency bands of each feeder line; According to energy and maximum principle, choose all feeder line energy in 10 frequency bands and maximum frequency band as feature band; Ask for the energy of each feeder line in feature band
Figure 18366DEST_PATH_IMAGE005
, extract phase place corresponding to each feeder line feature band centre frequency;
(3) design BP neural network, the BP neural network is divided into three layers, and topological structure is *
Figure 542068DEST_PATH_IMAGE006
*
Figure 467299DEST_PATH_IMAGE007
Wherein, ground floor is that input layer contains
Figure 127825DEST_PATH_IMAGE006
Individual node, the energy of each feeder line that will try to achieve by multiple wavelet transformation
Figure 189322DEST_PATH_IMAGE005
With phase place corresponding to each feeder line feature band centre frequency
Figure 411356DEST_PATH_IMAGE008
As sample attribute; The second layer is that hidden layer contains
Figure 823882DEST_PATH_IMAGE006
Individual node, adopting the logsig function is logarithm sigmoid transport function; The 3rd layer contains for output layer
Figure 789564DEST_PATH_IMAGE007
Individual node, adopting the purelin function is pure linear transfer function; mBe resonant earthed system feeder line quantity,
(4) carry out ANN (Artificial Neural Network) training, the following fault signature of analog simulation: choose respectively the trouble spot at 1/3l, the 1/2l of every feeder line, 2/3l place and bus place, fault resistance is chosen for respectively arc furnace load and constant transition resistance 20 Ω, 100 Ω, 500 Ω, the initial phase angle of fault is got respectively 0 °, 30 °, 60 °, 90 °, and load is chosen for respectively firm power load and arc furnace load; By step (1), (2) described method, extract respectively the fault transient amount
Figure 98503DEST_PATH_IMAGE005
With
Figure 434544DEST_PATH_IMAGE008
As sample set, train in input BP neural network after normalized, obtain satisfactory failure line selection network, the error performance target is 1
Figure 655DEST_PATH_IMAGE010
(5) with the fault transient amount of power distribution network to be selected With The sample attribute input neural network carries out renormalization to the Output rusults of neural network and processes, obtain 1 *
Figure 925382DEST_PATH_IMAGE007
Failure line selection network output matrix
Figure 295184DEST_PATH_IMAGE013
[
Figure 123463DEST_PATH_IMAGE014
, ,
Figure 965571DEST_PATH_IMAGE016
...,
Figure 76747DEST_PATH_IMAGE017
,
Figure 87428DEST_PATH_IMAGE018
]; Wherein,
Figure 993067DEST_PATH_IMAGE019
Be resonant earthed system feeder line quantity;
(6) according to failure line selection network output matrix YCarry out failure line selection: when
Figure 89199DEST_PATH_IMAGE020
≈ 1, and other value approximates at 0 o'clock, judges the
Figure 2013101071086100002DEST_PATH_IMAGE021
The bar feeder line breaks down,
Figure 738487DEST_PATH_IMAGE021
Be the feeder line numbering; When
Figure 39893DEST_PATH_IMAGE018
≈ 1, and other value approximates at 0 o'clock, judges that bus breaks down;
Figure 116433DEST_PATH_IMAGE019
Be resonant earthed system feeder line quantity.
Principle of the present invention is:
One, the extraction of characteristic quantity
1, when power distribution network bus residual voltage instantaneous value is out-of-limit, namely
Figure 637544DEST_PATH_IMAGE001
Figure 824943DEST_PATH_IMAGE002
The time, fault line selection device starts immediately and records ripple, obtains each feeder line zero sequence transient current, all feeder line zero sequence transient currents of window when recording after fault 5ms; Wherein,
Figure 810217DEST_PATH_IMAGE003
Be the bus rated voltage,
Figure 556193DEST_PATH_IMAGE004
=0.15;
2, one of transient characteristic quantity is selected: because the amplitude (or energy) of transient state component in feature band of each feeder line distributes and PHASE DISTRIBUTION all exists mapping relations with the distribution characteristics of the transient state component of fault feeder, for the impact of reduce disturbance signal, the frequency band of choosing all feeder line energy and maximum according to energy and maximum principle is feature band; In order to strengthen the reliability of route selection, simultaneously in the selected characteristic frequency band phase place corresponding to the energy of each feeder line and feature band centre frequency as the characteristic quantity of route selection.
3, multiple wavelet transformation: the advantage of multiple small echo maximum is that it can extract amplitude information and the phase information of signal simultaneously, this is to be the basis function of plural number because multiple small echo is a series of female small echos, its Wavelet Transform Parameters is also plural number, can obtain simultaneously signal amplitude information and phase information thus.Multiple small echo series commonly used has multiple Gauss, multiple Shannon, multiple Morlet and complex frequency B spline wavelets.
Being transformed to of continuous complex wavelet:
（1）
Figure 618007DEST_PATH_IMAGE023
（2）
In formula, SBe the flexible parameter of the yardstick of wavelet transformation, uBe the time migration parameter, t is the time, and C is constant.
The simple information of multiple wavelet transformation comprises respectively real part (RWT), imaginary part (IWT), amplitude (MWT) and phase place (PHWT), and amplitude and phase place are defined as:
Figure 457787DEST_PATH_IMAGE024
（3）
Figure 876130DEST_PATH_IMAGE025
（4）
Continuous wavelet decomposes the corresponding relation of the number of plies and frequency:
Figure 106255DEST_PATH_IMAGE026
（5）
Wherein, aBe to decompose the number of plies, △ is sampling interval; F cBe wavelet center frequency, the Hz of unit; F aThe pseudo frequency corresponding with decomposing the number of plies, the Hz of unit.Adopt the multiple Gauss wavelet in 20 rank herein.
4, characteristic quantity calculates: for model shown in Figure 1, to after fault during 5ms all feeder line zero sequence transient currents of window carry out the multiple Gauss Wavelet Transform in 20 rank, decomposing the number of plies is 256 layers, choose the decomposition result that decomposition scale is the 4-203 layer, divide according to frequency band of 20 yardsticks, obtain the energy of 10 frequency bands of each feeder line.
According to energy and maximum principle, choose all feeder line energy in 10 frequency bands and maximum frequency band as feature band.Ask for the energy of each feeder line in feature band
Figure 963352DEST_PATH_IMAGE005
, extract phase place corresponding to each feeder line feature band centre frequency
Figure 93857DEST_PATH_IMAGE008
, wherein
Figure 745418DEST_PATH_IMAGE027
Be resonant earthed system feeder line quantity.
The realization of two, route selection function
1, neural network: neural network is to interconnect by a large amount of simple neurons the self-adaptation nonlinear dynamic system that forms.The principal character of artificial neural network is: large-scale parallel processing and the storage of distributed information; Extremely strong self-study, association and fault-tolerant ability; Good self-adaptation and self-organization; The nonlinear system of many inputs, many outputs.
2, the design of BP neural network: for system shown in Figure 1 and selected characteristic quantity design BP neural network, being divided into is three layers, and topological structure is
Figure 728418DEST_PATH_IMAGE006
*
Figure 123627DEST_PATH_IMAGE006
*
Figure 610103DEST_PATH_IMAGE007
, wherein ground floor is that input layer contains Individual node, the energy of each feeder line that will try to achieve by multiple wavelet transformation
Figure 902861DEST_PATH_IMAGE005
The phase place corresponding with the feature band centre frequency
Figure 537980DEST_PATH_IMAGE008
As sample attribute.The second layer is that hidden layer contains
Figure 941279DEST_PATH_IMAGE006
Individual node, adopting the logsig function is logarithm sigmoid transport function; The 3rd layer for output layer contains 7 nodes, and adopting the purelin function is pure linear transfer function.
3, ANN (Artificial Neural Network) training: be set as follows fault signature: choose respectively the trouble spot at 1/3l, the 1/2l of every feeder line, 2/3l place and bus place; Fault resistance is chosen for respectively arc furnace load and constant transition resistance (20 Ω, 100 Ω, 500 Ω); The initial phase angle of fault is got respectively 0 °, and 30 °, 60 °, 90 °; Load is chosen for respectively firm power load and arc furnace load, extracts respectively the fault transient amount by A, B method
Figure 137905DEST_PATH_IMAGE011
With
Figure 892235DEST_PATH_IMAGE012
As sample set, train in input BP neural network after normalized, obtain satisfactory failure line selection network, the error performance target is 1
Figure 566930DEST_PATH_IMAGE010
4, route selection criterion: with the fault transient amount of power distribution network to be selected
Figure 90315DEST_PATH_IMAGE005
With The sample attribute input neural network, the Output rusults of neural network obtains faulty line after renormalization is processed.
The output quantity of this failure line selection network is 1 *
Figure 433888DEST_PATH_IMAGE007
Matrix:
Figure 410810DEST_PATH_IMAGE013
[
Figure 991964DEST_PATH_IMAGE014
,
Figure 327130DEST_PATH_IMAGE015
,
Figure 993735DEST_PATH_IMAGE016
...,
Figure 72549DEST_PATH_IMAGE017
,
Figure 508210DEST_PATH_IMAGE018
]; When
Figure 14277DEST_PATH_IMAGE020
≈ 1, and other value approximates at 0 o'clock, judges the
Figure 627834DEST_PATH_IMAGE021
The bar feeder line breaks down,
Figure 448022DEST_PATH_IMAGE021
Be the feeder line numbering.When
Figure 800506DEST_PATH_IMAGE018
≈ 1, and other value approximates at 0 o'clock, judges that bus breaks down.
The present invention compared with prior art has following advantage:
1, this method has overcome and uses single route selection criterion to cause the problem of route selection mistake;
2, amplitude and the phase place of the disposable extraction zero-sequence current of multiple wavelet transformation energy of the method use, simplified computation process;
3, the method as the route selection criterion, has reduced the impact of undesired signal with the characteristic quantity in feature band; Feeder line transient state energy and phase place corresponding to feature band centre frequency have been increased the reliability of route selection simultaneously as the ANN training sample set.
4, the method ANN(artificial neural network) when training considered the impacts such as fault angle, transition resistance and load characteristic, so the route selection effect do not affect by above-mentioned factor, accurately and reliably, belongs to intelligent route selection method.
Along with the transformation and upgrade of power distribution network, cable is used in a large number, causes the distribution distributed capacitance to increase, and causes the ground connection capacity current to surpass the regulation of operating standard, requires the neutral point must be through grounding through arc.The present invention is applicable to the resonant earthed system by line-the cable joint line forms.
Description of drawings
Fig. 1 is resonant earthed system singlephase earth fault realistic model of the present invention;
Fig. 2 is the energy under the embodiment of the present invention 1 each frequency band of all feeder lines;
Fig. 3 is the energy that the embodiment of the present invention 1 is levied each feeder line in frequency band;
Fig. 4 is the embodiment of the present invention 1 each feeder line feature band centre frequency corresponding phase;
Fig. 5 is the embodiment of the present invention 1 neural network structure.
Embodiment
Below in conjunction with the drawings and specific embodiments, the invention will be further described.
Embodiment 1: 110kV/35kV resonant earthed system singlephase earth fault realistic model as shown in Figure 1, and it has 6 feeder lines, and the Z-shaped transformer neutral point is by arc suppression coil resistance in series ground connection.Overhead feeder
Figure 680738DEST_PATH_IMAGE028
=15km,
Figure 554890DEST_PATH_IMAGE029
=18km, =30km ， Xian – cable mixing feeder line
Figure 323443DEST_PATH_IMAGE031
=17km, its overhead feeder 12km, cable 5km, cable feeder line
Figure 171314DEST_PATH_IMAGE032
=6km,
Figure 34227DEST_PATH_IMAGE033
=8km.Wherein, overhead feeder is JS1 bar type, and LGJ-70 type wire, span 80m, cable feeder line are YJV23-35/95 type cable.G in this electrical network is infinitely great power supply; T is main-transformer, and no-load voltage ratio is 110 kV/35kV, and connection set is
Figure 694754DEST_PATH_IMAGE034
/ d11; It is the zigzag transformer; L is arc suppression coil; R is the damping resistance of arc suppression coil.Feeder line adopts overhead transmission line, overhead line-cable hybrid line and three kinds of circuits of cable line.
Suppose feeder line
Figure 243864DEST_PATH_IMAGE028
Apart from bus 5km place's generation singlephase earth fault, fault moment is 0.023ms, and transition resistance is 20 Ω, and sample frequency is 10kHz.According to preceding method, when choosing 5ms after fault, the zero-sequence current data of each feeder line of window are carried out multiple wavelet transformation, try to achieve the energy and as shown in Figure 2 under each frequency band of all feeder lines. Corresponding frequency band is 2.75 ~ 0.478kHz, Corresponding frequency band is 0.458 ~ 0.256kHz, other frequency bands can the like.Know frequency band by Fig. 2
Figure 475759DEST_PATH_IMAGE037
Corresponding energy and maximum are feature band under this fault condition according to energy and selected this frequency band of maximum principle.The calculated characteristics frequency band
Figure 931011DEST_PATH_IMAGE037
The energy of interior each feeder line
Figure 267052DEST_PATH_IMAGE011
And each feeder line feature band centre frequency corresponding phase
Figure 770846DEST_PATH_IMAGE012
, as shown in Figure 3, Figure 4.Fault feeder and perfect feeder line at feature band
Figure 806935DEST_PATH_IMAGE037
Under energy and phase place corresponding to feature band centre frequency have obvious difference.The above explanation by special case uses energy and the phase place corresponding to feature band centre frequency of each feeder line under multiple wavelet transformation extraction feature band to carry out the route selection accurate and effective as characteristic quantity.
With vectorial P=[
Figure 370772DEST_PATH_IMAGE038
,
Figure 757891DEST_PATH_IMAGE039
, ,
Figure 392189DEST_PATH_IMAGE041
, ,
Figure 673446DEST_PATH_IMAGE043
,
Figure 784621DEST_PATH_IMAGE044
, ,
Figure 199477DEST_PATH_IMAGE046
, , ,
Figure 747767DEST_PATH_IMAGE049
] sample attribute builds the BP neural network as input layer, contains 12 nodes; It is logarithm sigmoid transport function that the hidden layer of network adopts the logsig function, contains 12 nodes; It is pure linear transfer function that output layer adopts the purelin function, contains 7 nodes.Neural network structure figure as shown in Figure 5.Be set as follows the various faults feature: choose respectively the trouble spot at 1/3l, the 1/2l of every feeder line, 2/3l place and bus place; Fault resistance is chosen for respectively arc furnace load and constant transition resistance (20 Ω, 100 Ω, 500 Ω); The initial phase angle of fault is got respectively 0 °, and 30 °, 60 °, 90 °; Load is chosen for respectively firm power load and arc furnace load, and extracts respectively the fault transient amount
Figure 886624DEST_PATH_IMAGE011
With
Figure 407735DEST_PATH_IMAGE012
As sample set, train in input BP neural network after normalized, obtain satisfactory failure line selection neural network, the error performance target is made as 1
Figure 657451DEST_PATH_IMAGE010
, the neural network output vector is
Figure 78943DEST_PATH_IMAGE013
[
Figure 388702DEST_PATH_IMAGE014
,
Figure 397109DEST_PATH_IMAGE015
,
Figure 388199DEST_PATH_IMAGE016
...,
Figure 227979DEST_PATH_IMAGE050
].The Output rusults of neural network obtains faulty line after renormalization is processed.The route selection criterion is, when
Figure 646322DEST_PATH_IMAGE020
≈ 1, and other value approximates at 0 o'clock, judges the The bar feeder line breaks down,
Figure 232079DEST_PATH_IMAGE021
Be the feeder line numbering.When
Figure 926365DEST_PATH_IMAGE050
≈ 1, and other value approximates at 0 o'clock, judges that bus breaks down.
Now suppose feeder line
Figure 515609DEST_PATH_IMAGE028
Singlephase earth fault occurs, and fault distance bus distance is respectively 2km, 6 km, 14 km, corresponding transition resistance and the initial phase angle of fault are respectively 150 Ω, 500 Ω, 100 Ω and 0 °, 15 °, 55 °.Zero sequence transient current under each fault condition is carried out the multiple Gauss Wavelet Transform in 20 rank, and decomposing the number of plies is 256 layers; Choose the decomposition result that decomposition scale is the 4-203 layer, divide according to frequency band of 20 yardsticks, obtain the energy of 10 frequency bands of each feeder line; According to energy and maximum principle, choose all feeder line energy in 10 frequency bands and maximum frequency band as feature band; Ask for the energy of each feeder line in feature band
Figure 560926DEST_PATH_IMAGE005
, extract phase place corresponding to each feeder line feature band centre frequency
Figure 893818DEST_PATH_IMAGE008
With the fault transient amount of trying to achieve
Figure 442611DEST_PATH_IMAGE011
With
Figure 468336DEST_PATH_IMAGE012
In the failure line selection neural network that the above-mentioned training of sample attribute input obtains, the neural network output vector of trying to achieve [
Figure 104908DEST_PATH_IMAGE014
,
Figure 711470DEST_PATH_IMAGE015
,
Figure 970413DEST_PATH_IMAGE016
...,
Figure 396847DEST_PATH_IMAGE050
] result as shown in table 1.As shown in Table 1
Figure 399438DEST_PATH_IMAGE014
≈ 1, and other value approximates 0, judges thus feeder line
Figure 860506DEST_PATH_IMAGE028
Break down, judged result is consistent with hypothesis, and route selection is correct.
Table 1 feeder line
Figure 24771DEST_PATH_IMAGE028
Route selection result when breaking down
Embodiment 2: for resonant earthed system as shown in Figure 1, suppose feeder line Singlephase earth fault occurs, and fault distance bus distance is respectively 2km, 5 km, 9.5km, corresponding transition resistance and the initial phase angle of fault are respectively 50 Ω, 250 Ω, 300 Ω and 25 °, 40 °, 75 °.Zero sequence transient current under each fault condition is carried out the multiple Gauss Wavelet Transform in 20 rank, and decomposing the number of plies is 256 layers; Choose the decomposition result that decomposition scale is the 4-203 layer, divide according to frequency band of 20 yardsticks, obtain the energy of 10 frequency bands of each feeder line; According to energy and maximum principle, choose all feeder line energy in 10 frequency bands and maximum frequency band as feature band; Ask for the energy of each feeder line in feature band
Figure 558892DEST_PATH_IMAGE005
, extract phase place corresponding to each feeder line feature band centre frequency
Figure 97321DEST_PATH_IMAGE008
With the fault transient amount of trying to achieve
Figure 560664DEST_PATH_IMAGE011
With
Figure 842740DEST_PATH_IMAGE012
In the failure line selection neural network that sample attribute input embodiment 1 training obtains, the neural network output vector of trying to achieve
Figure 340718DEST_PATH_IMAGE013
[
Figure 283004DEST_PATH_IMAGE014
,
Figure 499221DEST_PATH_IMAGE015
,
Figure 319410DEST_PATH_IMAGE016
...,
Figure 671894DEST_PATH_IMAGE050
] result as shown in table 2.As shown in Table 2
Figure 286546DEST_PATH_IMAGE052
≈ 1, and other value approximates 0, judges thus feeder line
Figure 662164DEST_PATH_IMAGE031
Break down, judged result is consistent with hypothesis, and route selection is correct.
Table 2 feeder line
Figure 348360DEST_PATH_IMAGE031
Route selection result when breaking down
Figure 991569DEST_PATH_IMAGE053
Embodiment 3: for resonant earthed system as shown in Figure 1, suppose bus generation singlephase earth fault, and the transition resistance of fault and the corresponding initial phase angle of fault are respectively 100 Ω, 200 Ω, 500 Ω and 15 °, 35 °, 75 °.Zero sequence transient current under each fault condition is carried out the multiple Gauss Wavelet Transform in 20 rank, and decomposing the number of plies is 256 layers; Choose the decomposition result that decomposition scale is the 4-203 layer, divide according to frequency band of 20 yardsticks, obtain the energy of 10 frequency bands of each feeder line; According to energy and maximum principle, choose all feeder line energy in 10 frequency bands and maximum frequency band as feature band; Ask for the energy of each feeder line in feature band
Figure 105018DEST_PATH_IMAGE005
, extract phase place corresponding to each feeder line feature band centre frequency
Figure 967932DEST_PATH_IMAGE008
With the fault transient amount of trying to achieve
Figure 129923DEST_PATH_IMAGE011
With
Figure 456999DEST_PATH_IMAGE012
The failure line selection neural network that sample attribute input embodiment 1 training obtains, the neural network output vector of trying to achieve
Figure 679033DEST_PATH_IMAGE013
[
Figure 262199DEST_PATH_IMAGE014
,
Figure 290198DEST_PATH_IMAGE015
,
Figure 409463DEST_PATH_IMAGE016
...,
Figure 864715DEST_PATH_IMAGE050
] result as shown in table 3.As shown in Table 3
Figure 702221DEST_PATH_IMAGE050
≈ 1, and other value approximates 0, judges that thus bus breaks down, and judged result is consistent with hypothesis, and route selection is correct.
Route selection result when table 3 bus breaks down
Figure 268332DEST_PATH_IMAGE054
The above is illustrated embodiments of the present invention by reference to the accompanying drawings, but the present invention is not limited to above-mentioned embodiment, in the ken that those skilled in the art possess, can also make a variety of changes under the prerequisite that does not break away from aim of the present invention.

Claims (1)

1. one kind is utilized the malfunction route selection method for resonant grounded system of answering small echo and ANN combination, it is characterized in that by following process implementation:
(1) when bus residual voltage instantaneous value is out-of-limit, namely
Figure 141376DEST_PATH_IMAGE001
Figure 798753DEST_PATH_IMAGE002
The time, fault line selection device starts immediately and records ripple, obtains each feeder line zero sequence transient current, all feeder line zero sequence transient currents of window when recording after fault 5ms; Wherein,
Figure 108512DEST_PATH_IMAGE003
Be the bus rated voltage,
Figure 615454DEST_PATH_IMAGE004
=0.15;
(2) each zero sequence transient current is carried out the multiple Gauss Wavelet Transform in 20 rank, decomposing the number of plies is 256 layers; Choose the decomposition result that decomposition scale is the 4-203 layer, divide according to frequency band of 20 yardsticks, obtain the energy of 10 frequency bands of each feeder line; According to energy and maximum principle, choose all feeder line energy in 10 frequency bands and maximum frequency band as feature band; Ask for the energy of each feeder line in feature band
Figure 606544DEST_PATH_IMAGE005
, extract phase place corresponding to each feeder line feature band centre frequency
(3) design BP neural network, the BP neural network is divided into three layers, and topological structure is
Figure 864667DEST_PATH_IMAGE007
*
Figure 360371DEST_PATH_IMAGE007
*
Figure 951889DEST_PATH_IMAGE008
Wherein, ground floor is that input layer contains
Figure 82394DEST_PATH_IMAGE007
Individual node, the energy of each feeder line that will try to achieve by multiple wavelet transformation
Figure 999534DEST_PATH_IMAGE005
With phase place corresponding to each feeder line feature band centre frequency As sample attribute; The second layer is that hidden layer contains Individual node, adopting the logsig function is logarithm sigmoid transport function; The 3rd layer contains for output layer
Figure 864219DEST_PATH_IMAGE008
Individual node, adopting the purelin function is pure linear transfer function; mBe resonant earthed system feeder line quantity,
Figure 624365DEST_PATH_IMAGE009
(4) carry out the ANN training, the following fault signature of analog simulation: choose respectively the trouble spot at 1/3l, the 1/2l of every feeder line, 2/3l place and bus place, fault resistance is chosen for respectively arc furnace load and constant transition resistance 20 Ω, 100 Ω, 500 Ω, the initial phase angle of fault is got respectively 0 °, 30 °, 60 °, 90 °, and load is chosen for respectively firm power load and arc furnace load; By step (1), (2) described method, extract respectively the fault transient amount
Figure 891398DEST_PATH_IMAGE005
With
Figure 526516DEST_PATH_IMAGE006
As sample set, train in input BP neural network after normalized, obtain satisfactory failure line selection network, the error performance target is 1
(5) with the fault transient amount of power distribution network to be selected
Figure 126442DEST_PATH_IMAGE011
With The sample attribute input neural network, the Output rusults of neural network obtains faulty line after renormalization is processed, and the output quantity of failure line selection network is 1 *
Figure 555466DEST_PATH_IMAGE008
Matrix, [ ,
Figure 124223DEST_PATH_IMAGE015
,
Figure 399346DEST_PATH_IMAGE016
...,
Figure 980500DEST_PATH_IMAGE017
,
Figure 253350DEST_PATH_IMAGE018
]; Wherein,
Figure 982271DEST_PATH_IMAGE019
Be resonant earthed system feeder line quantity;
The output quantity matrix of the failure line selection network that (6) obtains according to step (5) carries out failure line selection; When
Figure 998769DEST_PATH_IMAGE020
≈ 1, and other value approximates at 0 o'clock, judges the
Figure 2013101071086100001DEST_PATH_IMAGE021
The bar feeder line breaks down,
Figure 870648DEST_PATH_IMAGE021
Be the feeder line numbering; When
Figure 642295DEST_PATH_IMAGE018
≈ 1, and other value approximates at 0 o'clock, judges that bus breaks down; Be resonant earthed system feeder line quantity.
CN2013101071086A 2013-03-29 2013-03-29 Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network) Pending CN103163430A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2013101071086A CN103163430A (en) 2013-03-29 2013-03-29 Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2013101071086A CN103163430A (en) 2013-03-29 2013-03-29 Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network)

Publications (1)

Publication Number Publication Date
CN103163430A true CN103163430A (en) 2013-06-19

Family

ID=48586682

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2013101071086A Pending CN103163430A (en) 2013-03-29 2013-03-29 Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network)

Country Status (1)

Country Link
CN (1) CN103163430A (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103424668A (en) * 2013-08-05 2013-12-04 昆明理工大学 Arc light ground fault continuous route selection method utilizing principal component analysis of zero-sequence current of feeder line and evidence theoretical integration
CN103454559A (en) * 2013-09-02 2013-12-18 国家电网公司 Power distribution network single-phase earth fault zone positioning method and positioning device
CN103631985A (en) * 2013-09-02 2014-03-12 国家电网公司 Simulation impedance model of electric arc furnace piecewise linearity
CN104569684A (en) * 2015-01-14 2015-04-29 上海和伍新材料科技有限公司 Fault electric arc detection method based on electric arc spectrum signals
CN104865499A (en) * 2015-05-11 2015-08-26 昆明理工大学 Super-high voltage direct-current power transmission line region internal and external fault identification method
CN106443334A (en) * 2016-09-18 2017-02-22 昆明理工大学 Zero sequence current difference polarity comparison based power distribution network fault line selection 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
CN108663600A (en) * 2018-05-09 2018-10-16 广东工业大学 A kind of method for diagnosing faults, device and storage medium based on power transmission network
CN109061397A (en) * 2018-10-11 2018-12-21 南方电网科学研究院有限责任公司 A kind of line fault area recognizing method
CN112240965A (en) * 2019-07-17 2021-01-19 南京南瑞继保工程技术有限公司 Grounding line selection device and method based on deep learning algorithm
CN113866572A (en) * 2021-09-06 2021-12-31 西安理工大学 Direct-current fault arc detection and positioning method under condition of access of multiple power electronic devices
CN114814468A (en) * 2022-06-20 2022-07-29 南京工程学院 Intelligent line selection method considering high-proportion DG access for single-phase earth fault of power distribution network
CN115291039A (en) * 2022-08-09 2022-11-04 贵州大学 Single-phase earth fault line selection method of resonance earthing system

Citations (3)

* Cited by examiner, † Cited by third party
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
CN101545943A (en) * 2009-05-05 2009-09-30 昆明理工大学 Method for fault line selection of cable-wire mixed line of electric distribution network by using wavelet energy relative entropy
CN101924354A (en) * 2010-04-19 2010-12-22 昆明理工大学 Artificially neural network routing method for distribution network failure by using S-transforming energy sampling property

Patent Citations (3)

* Cited by examiner, † Cited by third party
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
CN101545943A (en) * 2009-05-05 2009-09-30 昆明理工大学 Method for fault line selection of cable-wire mixed line of electric distribution network by using wavelet energy relative entropy
CN101924354A (en) * 2010-04-19 2010-12-22 昆明理工大学 Artificially neural network routing method for distribution network failure by using S-transforming energy sampling property

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
束洪春等: "基于零序电流小波变换系数均方根值的故障选线ANN方法", 《电力科学与技术学报》 *

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103424668B (en) * 2013-08-05 2016-02-24 昆明理工大学 A kind of continuous selection method of arc grounding fault utilizing the principal component analysis (PCA) of feeder line zero-sequence current and evidence theory fusion
CN103424668A (en) * 2013-08-05 2013-12-04 昆明理工大学 Arc light ground fault continuous route selection method utilizing principal component analysis of zero-sequence current of feeder line and evidence theoretical integration
CN103454559A (en) * 2013-09-02 2013-12-18 国家电网公司 Power distribution network single-phase earth fault zone positioning method and positioning device
CN103631985A (en) * 2013-09-02 2014-03-12 国家电网公司 Simulation impedance model of electric arc furnace piecewise linearity
CN103454559B (en) * 2013-09-02 2015-10-28 国家电网公司 A kind of one-phase earthing failure in electric distribution network Section Location and locating device
CN104569684A (en) * 2015-01-14 2015-04-29 上海和伍新材料科技有限公司 Fault electric arc detection method based on electric arc spectrum signals
CN104569684B (en) * 2015-01-14 2017-08-25 上海和伍复合材料有限公司 A kind of fault arc detection method based on arc spectrum signal
CN104865499A (en) * 2015-05-11 2015-08-26 昆明理工大学 Super-high voltage direct-current power transmission line region internal and external fault identification method
CN106443334B (en) * 2016-09-18 2019-04-09 昆明理工大学 A kind of distribution network fault line selection method based on zero-sequence current difference Polarity comparision
CN106443334A (en) * 2016-09-18 2017-02-22 昆明理工大学 Zero sequence current difference polarity comparison based power distribution network fault line selection 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
CN108279364B (en) * 2018-01-30 2020-01-14 福州大学 Power distribution network single-phase earth fault line selection method based on convolutional neural network
CN108663600A (en) * 2018-05-09 2018-10-16 广东工业大学 A kind of method for diagnosing faults, device and storage medium based on power transmission network
CN108663600B (en) * 2018-05-09 2020-11-10 广东工业大学 Fault diagnosis method and device based on power transmission network and storage medium
CN109061397A (en) * 2018-10-11 2018-12-21 南方电网科学研究院有限责任公司 A kind of line fault area recognizing method
CN109061397B (en) * 2018-10-11 2020-07-28 南方电网科学研究院有限责任公司 Line fault area identification method
CN112240965A (en) * 2019-07-17 2021-01-19 南京南瑞继保工程技术有限公司 Grounding line selection device and method based on deep learning algorithm
CN113866572A (en) * 2021-09-06 2021-12-31 西安理工大学 Direct-current fault arc detection and positioning method under condition of access of multiple power electronic devices
CN113866572B (en) * 2021-09-06 2024-05-28 西安理工大学 DC fault arc detection and positioning method under multi-power electronic device access condition
CN114814468A (en) * 2022-06-20 2022-07-29 南京工程学院 Intelligent line selection method considering high-proportion DG access for single-phase earth fault of power distribution network
CN114814468B (en) * 2022-06-20 2022-09-20 南京工程学院 Intelligent line selection method considering high-proportion DG access for single-phase earth fault of power distribution network
CN115291039A (en) * 2022-08-09 2022-11-04 贵州大学 Single-phase earth fault line selection method of resonance earthing system

Similar Documents

Publication Publication Date Title
CN103163430A (en) Resonant grounding system fault line selection method by combining complex wavelets with ANN (artificial neural network)
Yang et al. New ANN method for multi-terminal HVDC protection relaying
CN103424669B (en) A kind of selection method utilizing fault feeder zero-sequence current matrix principal component analysis (PCA) first principal component
CN103018627B (en) Adaptive fault type fault line detection method for non-effectively earthed system
CN103424668B (en) A kind of continuous selection method of arc grounding fault utilizing the principal component analysis (PCA) of feeder line zero-sequence current and evidence theory fusion
CN103267927B (en) A kind of low current neutral grounding system fault route selecting method utilizing power frequency component wavelet coefficient fitting a straight line to detect
CN103474981B (en) A kind of distribution network single-phase ground protection method based on the multistage differential transformation direction of zero-sequence current
CN103217622B (en) Based on the distribution network fault line selection method of multiterminal voltage traveling wave
CN103474980B (en) A kind of power distribution network single-phase ground protection method based on transient power direction
CN103257304A (en) ANN fault line selection method through CWT coefficient RMS in zero-sequence current feature band
CN103474974B (en) A kind of distribution network single-phase ground protection method based on zero-sequence current Sudden Changing Rate fitting a straight line direction
CN101964515B (en) Method for converting boundary element by extra-high voltage direct current transmission line mode voltage S
CN101924354B (en) Artificially neural network routing method for distribution network failure by using S-transforming energy sampling property
CN105759167A (en) Wavelet neural network-based distribution network single-phase short circuit line selection method
CN110783896B (en) Distribution network single-phase grounding protection method based on weak fault active and passive joint detection
Azeroual et al. Fault location and detection techniques in power distribution systems with distributed generation: Kenitra City (Morocco) as a case study
CN102135591A (en) Resonant grounding power grid single-phase ground fault db wavelet transient component line selection method
CN102928731A (en) Power distribution network fault line selection method using zero-sequence current full quantity Hough transformation
CN104614638A (en) Grounding line selection method for small current system
CN103163417A (en) Unreal grounding identification method based on short time window and high-low frequency transient state energy ratio
CN103293448B (en) Identification method of single-phase ground fault and virtual grounding based on semi-cycle energy ratio
Khalili et al. A fault location and detection technique for STATCOM compensated transmission lines using game theory
Lee et al. New fault detection method for low voltage DC microgrid with renewable energy sources
CN103323728B (en) Based on singlephase earth fault and the Xuhanting oral solution recognition methods of whole wave energy Ratios
Nasab et al. A hybrid scheme for fault locating for transmission lines with TCSC

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20130619