CN110501603A - Utilize the forever rich direct-current commutation failure method for diagnosing faults of EMD and neural network - Google Patents
Utilize the forever rich direct-current commutation failure method for diagnosing faults of EMD and neural network Download PDFInfo
- Publication number
- CN110501603A CN110501603A CN201910672418.XA CN201910672418A CN110501603A CN 110501603 A CN110501603 A CN 110501603A CN 201910672418 A CN201910672418 A CN 201910672418A CN 110501603 A CN110501603 A CN 110501603A
- Authority
- CN
- China
- Prior art keywords
- neural network
- entropy
- commutation failure
- emd
- sample
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 23
- 230000005540 biological transmission Effects 0.000 claims abstract description 18
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 13
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 230000009466 transformation Effects 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 22
- 238000012360 testing method Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000013480 data collection Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims 1
- 238000003745 diagnosis Methods 0.000 claims 1
- 238000002474 experimental method Methods 0.000 abstract description 2
- 230000035772 mutation Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 17
- 230000005611 electricity Effects 0.000 description 5
- 230000004888 barrier function Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 101100499229 Mus musculus Dhrsx gene Proteins 0.000 description 1
- 241000282376 Panthera tigris Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000005404 monopole Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)
Abstract
The present invention relates to a kind of forever rich direct-current commutation failure method for diagnosing faults using EMD and neural network, belong to HVDC transmission system technical field of relay protection.Commutation failure is a kind of most common failure in DC transmission system, will lead to Current Voltage mutation, brings and seriously threaten to the safe and stable operation of direct current system.This method is extracted inverter side DC current signal and carries out phase-model transformation, the Aerial mode component of current signal is decomposed with empirical mode decomposition (EMD) method again to obtain n intrinsic mode function (IMF) component, by the approximate entropy and Sample Entropy of seeking IMF component, and will be combined as feature vector with neural network after the normalization of obtained entropy, it establishes neural network classifier and is used to identify commutation failure and line fault.Obtained by a large amount of emulation experiments: the method for diagnosing faults, which can be identified quickly and effectively, to be out of order.
Description
Technical field
The present invention relates to a kind of forever rich direct-current commutation failure method for diagnosing faults using EMD and neural network, belong to height
Straightening stream transmission system relaying technical field.
Background technique
The superior functions such as D.C. high voltage transmission is high with its voltage class, conveying distance is remote, transmission capacity is big, solve China
The problem of load and the energy are unevenly distributed.Rich DC transmission engineering is first of China's construction direct current transportation work inside the province forever
Journey.Rich DC transmission system sending end Yongren converter station is located at the Yongren County fierce tiger township of Chuxiong Prefecture city, Yunnan Province northeast 112km forever, by
End Funing converter station is located at the town Li Da to the east of mountain of papers city, Zhuang-Miao Autonomous Prefecture of Wenshan, Yunnan Province, and specified transmission power is bipolar
3000MW, monopole 1500MW, rated current 3000A, rated direct voltage are ± 500kV, and route power transmission distance is about
577km.It can satisfy the installed capacity demand that kwan-yin rock power station sends out to Guangxi 3,000,000 kilowatts in the wet season, and
Dry season can satisfy the power demand in Yunnan Province mountain of papers area.Rich DC transmission system is built up for by rich brown forever
Hydroelectric resources send outside for it is significant.It is for weak exchange since its inverter side (Guangxi side) power supply is weaker, when its inversion top-cross
When streaming system breaks down, it is easy to cause DC transmission system commutation failure.Commutation failure as DC transmission system one
Kind most common failure can cause DC current and DC voltage to mutate, bring seriously to the safe and stable operation of direct current system
It threatens.The prior art still has problems to effective identification of commutation failure, causes to have taken in time in commutation failure
Effect measure.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of forever rich direct-current commutation failure events using EMD and neural network
Hinder diagnostic method, can effectively identify commutation failure, it is significant to DC transmission system.
The technical solution adopted by the present invention is that: it is a kind of to be examined using the forever rich direct-current commutation failure failure of EMD and neural network
Disconnected method, includes the following steps:
A, inverter side AC system single-phase earthing, inverter side AC system line to line fault, inverter side exchange system are extracted respectively
It unites under six kinds of two phase ground, inverter side AC system three-phase ground, system normal condition and direct current transmission line fault differences
Inverter side DC current signal, and the decomposition of phase mould is carried out to the signal of extraction, the Aerial mode component for then choosing fault current carries out
EMD is decomposed, and obtains relatively stable 6 intrinsic mode functions IMF component from high frequency to low frequency；
B, the approximate entropy and Sample Entropy of IMF2~IMF5 component are calculated, and calculated result is normalized, will be returned
One, which changes later entropy, separately constitutes approximate entropy feature vectorWith sample entropy feature vectorWherein,It is the approximate entropy for calculating IMF2, IMF3, IMF4, IMF5 component
Entropy after normalization,It is after the Sample Entropy of calculating IMF2, IMF3, IMF4, IMF5 component normalizes
Entropy；
C, approximate entropy feature vector is chosen respectivelyIn a part as training sample, another part be test sample and
Sample entropy feature vectorIn a part as training sample, another part is test sample；
D, neural network classifier is established, carries out fault identification using approximate entropy as feature vector and respectively with Sample Entropy
As the fault identification of feature vector, content is identified are as follows: neural network classifier output (0 0 1) when normal condition, route event
Neural network classifier output (1 0 0) when barrier, neural network classifier output (0 1 0) when commutation failure.
Specifically, 6 IMF components of gained are to use inverter side DC current as original signal in step A, carry out phase to it
Modular transformation has been obtained after the Aerial mode component of electric current is carried out signal decomposition with EMD as relatively stable 5 from high frequency to low frequency
IMF component and a surplus R amount to 6 components, since first IMF component is the radio-frequency component comprising noise, therefore choose
Tetra- components of IMF2-IMF5 are analyzed.
Specifically, it is max min algorithm to the method that calculated result is normalized in step B, calculates
Formula are as follows:In formula, i=1,2 ..., m, j=1,2 ..., n, n and m are positive integer,For the data set after normalization, Di×jFor raw sample data collection, min (Dj) be original sample minimum value, max (Dj)
For the maximum value of original sample.
Preferably, the selection of training sample, test sample is selected at random in feature vector after normalization in step C
It takes.
Preferably, the neural network classifier established in step D is Elman neural network classifier.
Preferably, the neural network uses adaptive learning rate algorithm, learning rate 0.01, and maximum frequency of training takes
It is 5000, convergence precision is set as 10-2。
The beneficial effects of the present invention are: the present invention can effectively identify commutation failure, to take corresponding commutation to lose in time
Braking measure is lost to provide the foundation.
Detailed description of the invention
Fig. 1 is overall flow figure of the invention.
Fig. 2 is Elman neural network structure figure, in Fig. 2: 1 is input quantity, and 2 is accept layer, and 3 be output layer, 4,5,6 points
Not Wei input layer, hidden layer, accept layer, 7,8,9 be respectively normal condition (0 0 1), line fault (1 0 0), commutation failure
(0 1 0)。
Fault-current signal obtains result after EMD is decomposed when Fig. 3 is inverter side single-phase earthing.Ordinate indicates electricity in figure
Amplitude, unit A are flowed, abscissa indicates the number of sampled point, and abscissa unit is a.The first row of first row a line to the end
Subgraph respectively indicates the waveform diagram of IMF1, IMF2 component after primary fault current signal, decomposition；The first row of secondary series is to most
A line subgraph respectively indicates the waveform diagram of IMF3, IMF4, IMF5 component after decomposing afterwards；Third column subgraph indicates remaining after decomposing
The waveform diagram of component.
Fault-current signal obtains result after EMD is decomposed when Fig. 4 is inverter side line to line fault.Ordinate indicates electricity in figure
Amplitude, unit A are flowed, abscissa indicates the number of sampled point, and abscissa unit is a.The first row of first row a line to the end
Subgraph respectively indicates the waveform diagram of IMF1, IMF2 component after primary fault current signal, decomposition；The first row of secondary series is to most
A line subgraph respectively indicates the waveform diagram of IMF3, IMF4, IMF5 component after decomposing afterwards；Third column subgraph indicates remaining after decomposing
The waveform diagram of component.
Fault-current signal obtains result after EMD is decomposed when Fig. 5 is inverter side two phase ground.Ordinate indicates electricity in figure
Amplitude, unit A are flowed, abscissa indicates the number of sampled point, and abscissa unit is a.The first row of first row a line to the end
Subgraph respectively indicates the waveform diagram of IMF1, IMF2 component after primary fault current signal, decomposition；The first row of secondary series is to most
A line subgraph respectively indicates the waveform diagram of IMF3, IMF4, IMF5 component after decomposing afterwards；Third column subgraph indicates remaining after decomposing
The waveform diagram of component.
Fault-current signal obtains result after EMD is decomposed when Fig. 6 is inverter side three-phase ground.Ordinate indicates electricity in figure
Amplitude, unit A are flowed, abscissa indicates the number of sampled point, and abscissa unit is a.The first row of first row a line to the end
Subgraph respectively indicates the waveform diagram of IMF1, IMF2 component after primary fault current signal, decomposition；The first row of secondary series is to most
A line subgraph respectively indicates the waveform diagram of IMF3, IMF4, IMF5 component after decomposing afterwards；Third column subgraph indicates remaining after decomposing
The waveform diagram of component.
Fig. 7 is that current signal obtains result after EMD is decomposed under normal condition.Ordinate indicates current amplitude in figure, single
Position is A, and abscissa indicates the number of sampled point, and abscissa unit is a.The first row of first row to the end distinguish by a line subgraph
The waveform diagram of IMF1, IMF2 component after indicating primary fault current signal, decomposition；A line is sub to the end for the first row of secondary series
Figure respectively indicates the waveform diagram of IMF3, IMF4, IMF5 component after decomposing；Third column subgraph indicates the wave of residual components after decomposing
Shape figure.
Fault-current signal obtains result after EMD is decomposed when Fig. 8 is direct current transmission line fault.Ordinate indicates in figure
Current amplitude, unit A, abscissa indicate the number of sampled point, and abscissa unit is a.The first row of first row is to last one
Row subgraph respectively indicates the waveform diagram of IMF1, IMF2 component after primary fault current signal, decomposition；The first row of secondary series arrives
Last line subgraph respectively indicates the waveform diagram of IMF3, IMF4, IMF5 component after decomposing；Third column subgraph indicates to remain after decomposing
The waveform diagram of remaining component.
Fig. 9 is the training result figure of Elman+ approximate entropy.Ordinate indicates the corresponding training essence reached of train epochs in figure
Degree, abscissa indicate train epochs.From ordinate 10 in figure-2The straight line that place is drawn indicates default precision, and vertical line expression reaches pre-
If train epochs when precision, curve indicates training curve.Elman+ approximate entropy train epochs can be only achieved default when being 3057
For 0.01 precision.
Figure 10 is the training result figure of Elman+ Sample Entropy.Ordinate indicates the corresponding training essence reached of train epochs in figure
Degree, abscissa indicate train epochs.From ordinate 10 in figure-2The straight line that place is drawn indicates default precision, and vertical line expression reaches pre-
If train epochs when precision, curve indicates training curve.Elman+ Sample Entropy train epochs just have reached pre- when being 2397
It is set as 0.01 precision.Neural network known to analysis chart 9, Figure 10 can reach well default in 30 latter two methods of operation
Training precision, and input when ratio using Sample Entropy as neural network is using approximate entropy as a moment 660 steps of generation when input.
The reason is that the data of its input of Elman+ approximate entropy are most in three kinds of training samples, therefore its frequency of training is also most.
Specific embodiment
In the following with reference to the drawings and specific embodiments, the present invention is described further.
Embodiment 1: as Figure 1-10 shows, the present invention has built rich direct current transportation forever in PSCAD/EMTDC simulation software
The simulation model of system is concluded that when inverter side converter power transformer no-load voltage ratio increases and causes commutation failure and continuously change
The mutually critical no-load voltage ratio of failure；Single-phase earthing, line to line fault occur for inverter side AC system, when two phase ground and three-phase ground failure
Cause the critical resistance of commutation failure and continuous commutation failure；Commutation failure and different frequencies in current signal when line fault occurs
Rate component content is different.
1) when being run under the conditions of high-voltage direct current declared working condition, when inverter side transformer voltage on valve side size be 525kV,
Line side voltage is 204.77kV, no-load voltage ratio sizeIt is found by Multi simulation running: as transformer voltage ratio K
When reduction, Commutation Failure will not be caused；When no-load voltage ratio K is increased to by 2.564When, it will it leads
Cause inverter that a commutation failure occurs；And when no-load voltage ratio continues to increase toWhen, inverter will occur
Continuous commutation failure.
2) when singlephase earth fault occurs, fault resstance is smaller, and fault time is longer, and commutation failure is more easy to happen,
Middle fault close angle is easiest to cause commutation failure when being 90 ° and 270 °.Occur when 90 ° of phase of inverter side AC system A single-phase
Ground fault, failure continue 0.05s, are concluded that by largely emulating when ground fault resistance is between 76.4~129.6
When Ω, a commutation failure will occur for inverter；And when ground fault resistance is greater than 129.6 Ω, inverter will not occur
Commutation failure.
3) when line to line fault occurs for AC system, fault resstance is smaller, and fault time is longer, the easier hair of commutation failure
It is raw, it is easiest to cause commutation failure when wherein fault close angle is 90 ° and 270 °.When 270 ° of inverter side AC system A, B two-phase
Shi Fasheng line to line fault, failure continue 0.05s, are concluded that the event that continuous commutation failure does not occur by largely emulating
Barrier resistance is 115.3 Ω；When fault resstance is between 115.3~343.2 Ω, a commutation failure will occur for inverter；And work as
When ground fault resistance is greater than 343.2 Ω, commutation failure will not occur for inverter.
4) AC system generation double earthfault is similar with two-phase short-circuit fault, and fault resstance is smaller, and fault time gets over
Long, commutation failure is more easy to happen, and is easiest to cause commutation failure when wherein fault close angle is 90 ° and 270 °.Work as inverter side
Two phase ground occurs at 270 ° of AC system A, B two-phase, failure continues 0.05s, is concluded that and does not send out by largely emulating
The fault resstance of raw continuous commutation failure is 63.1 Ω；The Ω when ground fault resistance is between 63.1~79.6, inverter will be sent out
A raw commutation failure；And when ground fault resistance is greater than 79.6 Ω, commutation failure will not occur for inverter.
5) three-phase ground failure is symmetrical fault, occur when three-phase ground short circuit commutation failure and ground resistance size and
Trouble duration has much relations, and the influence of time of failure can almost be ignored.It is 0.5s that the failure generation moment, which is arranged,
Trouble duration 0.05s is concluded that the fault resstance that continuity commutation failure does not occur is by largely emulating
38.2Ω；When ground fault resistance is between 38.2~39.8 Ω, a commutation failure will occur for inverter；And work as ground fault
When resistance is greater than 39.8 Ω, commutation failure will not occur for inverter.
6) inverter side AC system occurs single-phase earthing, line to line fault, two phase ground and three-phase ground and causes commutation mistake
When losing, DC current signal contains higher component in low-frequency range；And DC line is when breaking down, the height of DC current signal
Frequency range contains higher component.And DC current signal is straight when the amplitude of low-frequency range is higher than line fault when commutation failure
Amplitude of the galvanic electricity stream in low-frequency range.
A kind of forever rich direct-current commutation failure method for diagnosing faults using EMD and neural network of the present invention, including walk as follows
It is rapid:
A, inverter side AC system single-phase earthing, inverter side AC system line to line fault, inverter side exchange system are extracted respectively
It unites under six kinds of two phase ground, inverter side AC system three-phase ground, system normal condition and direct current transmission line fault differences
Inverter side DC current signal, and the decomposition of phase mould is carried out to the signal of extraction, the Aerial mode component for then choosing fault current carries out
EMD is decomposed, and obtains relatively stable 6 intrinsic mode functions IMF component from high frequency to low frequency；
B, the approximate entropy and Sample Entropy of IMF2~IMF5 component are calculated, and calculated result is normalized, will be returned
One, which changes later entropy, separately constitutes approximate entropy feature vectorWith sample entropy feature vectorWherein,It is the approximate entropy for calculating IMF2, IMF3, IMF4, IMF5 component
Entropy after normalization,It is after the Sample Entropy of calculating IMF2, IMF3, IMF4, IMF5 component normalizes
Entropy；
C, approximate entropy feature vector is chosen respectivelyIn a part as training sample, another part be test sample and
Sample entropy feature vectorIn a part as training sample, another part is test sample；
D, neural network classifier is established, carries out fault identification using approximate entropy as feature vector and respectively with Sample Entropy
As the fault identification of feature vector, content is identified are as follows: neural network classifier output (0 0 1) when normal condition, route event
Neural network classifier output (1 0 0) when barrier, neural network classifier output (0 1 0) when commutation failure.
Further, 6 IMF components of gained are to use inverter side DC current as original signal in step A, are carried out to it
Phase-model transformation has been obtained after the Aerial mode component of electric current is carried out signal decomposition with EMD as relatively stable 5 from high frequency to low frequency
A IMF component and a surplus R amount to 6 components, since first IMF component is the radio-frequency component comprising noise, therefore select
Tetra- components of IMF2-IMF5 are taken to be analyzed.
It further, is max min algorithm, meter to the method that calculated result is normalized in step B
Calculate formula are as follows:In formula, i=1,2 ..., m, j=1,2 ..., n, n and m are positive whole
Number,For the data set after normalization, Di×jFor raw sample data collection, min (Dj) be original sample minimum value, max
(Dj) be original sample maximum value.
Further, the selection of training sample, test sample is selected at random in feature vector after normalization in step C
It takes.
Further, the neural network classifier established in step D is Elman neural network classifier.
Further, the neural network uses adaptive learning rate algorithm, learning rate 0.01, maximum frequency of training
5000 are taken as, convergence precision is set as 10-2。
It is detected with the transmission system method for diagnosing faults, finds examining for the more commutation failures of the quantity of training sample
Disconnected result is more accurate.Therefore, 60 groups of test samples (totally 240 groups of data) are chosen and obtains diagnostic result and identification under different condition
Rate is shown in Table 1, table 2:
The test result of 1 60 groups of test samples of table
Analytical table 1 it is found that the diagnostic result under different conditions all close to desired output (0/1), but this it appears that
Elman+ Sample Entropy and the result of the result of desired output ratio Elman+ approximate entropy and desired output are closer.
The fault recognition rate of 2 60 groups of test samples of table
According to upper table 2 as can be seen that highest to commutation failure discrimination is Elman+ Sample Entropy, to commutation failure
Discrimination has reached 95.00%, and overall discrimination has also reached 90 or more percent.Testing result accuracy is relatively
Height matches with the testing result of table 1, while also demonstrating the validity of the commutation failure method for diagnosing faults.By a large amount of
Emulation experiment obtains: the method for diagnosing faults, which can be identified quickly and effectively, to be out of order.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (6)
1. a kind of forever rich direct-current commutation failure method for diagnosing faults using EMD and neural network, it is characterised in that: including as follows
Step:
A, inverter side AC system single-phase earthing, inverter side AC system line to line fault, inverter side AC system two are extracted respectively
The mutually inversion under ground connection, inverter side AC system three-phase ground, system normal condition and six kinds of differences of direct current transmission line fault
Side DC current signal, and the decomposition of phase mould is carried out to the signal of extraction, the Aerial mode component for then choosing fault current carries out EMD points
Solution, obtains relatively stable 6 intrinsic mode functions IMF component from high frequency to low frequency；
B, the approximate entropy and Sample Entropy of IMF2~IMF5 component are calculated, and calculated result is normalized, will be normalized
Later entropy separately constitutes approximate entropy feature vectorWith sample entropy feature vectorWherein,It is the approximate entropy for calculating IMF2, IMF3, IMF4, IMF5 component
Entropy after normalization,It is after the Sample Entropy of calculating IMF2, IMF3, IMF4, IMF5 component normalizes
Entropy；
C, approximate entropy feature vector is chosen respectivelyIn a part as training sample, another part is test sample and sample
Entropy feature vectorIn a part as training sample, another part is test sample；
D, establish neural network classifier, carry out respectively fault identification using approximate entropy as feature vector and using Sample Entropy as
The fault identification of feature vector identifies content are as follows: neural network classifier output (0 0 1) when normal condition, when line fault
Neural network classifier exports (1 0 0), neural network classifier output (0 1 0) when commutation failure.
2. the forever rich direct-current commutation failure method for diagnosing faults according to claim 1 using EMD and neural network, special
Sign is: 6 IMF components of gained are to use inverter side DC current as original signal in step A, carry out phase-model transformation to it,
It has been obtained after the Aerial mode component of electric current is carried out signal decomposition with EMD as relatively stable 5 IMF components from high frequency to low frequency
Amount to 6 components with a surplus R, since first IMF component is the radio-frequency component comprising noise, therefore chooses IMF2-
Tetra- components of IMF5 are analyzed.
3. the forever rich direct-current commutation failure method for diagnosing faults according to claim 1 using EMD and neural network, special
Sign is: it is max min algorithm to the method that calculated result is normalized in step B, its calculation formula is:In formula, i=1,2 ..., m, j=1,2 ..., n, n and m are positive integer,For
Data set after normalization, Di×jFor raw sample data collection, min (Dj) be original sample minimum value, max (Dj) it is original
The maximum value of sample.
4. the forever rich direct-current commutation failure method for diagnosing faults according to claim 1 using EMD and neural network, special
Sign is: the selection of training sample, test sample is randomly selected in feature vector after normalization in step C.
5. the forever rich direct-current commutation failure method for diagnosing faults according to claim 1 using EMD and neural network,
Be characterized in that: the neural network classifier established in step D is Elman neural network classifier.
6. utilizing the rich direct-current commutation failure fault diagnosis side forever of EMD and neural network according to claims 1 or 5
Method, it is characterised in that: the neural network uses adaptive learning rate algorithm, learning rate 0.01, and maximum frequency of training takes
It is 5000, convergence precision is set as 10-2。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910672418.XA CN110501603A (en) | 2019-07-24 | 2019-07-24 | Utilize the forever rich direct-current commutation failure method for diagnosing faults of EMD and neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910672418.XA CN110501603A (en) | 2019-07-24 | 2019-07-24 | Utilize the forever rich direct-current commutation failure method for diagnosing faults of EMD and neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110501603A true CN110501603A (en) | 2019-11-26 |
Family
ID=68586823
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910672418.XA Pending CN110501603A (en) | 2019-07-24 | 2019-07-24 | Utilize the forever rich direct-current commutation failure method for diagnosing faults of EMD and neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110501603A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112446430A (en) * | 2020-11-27 | 2021-03-05 | 云南电网有限责任公司电力科学研究院 | Fault identification method for direct-current power transmission system |
| CN112990546A (en) * | 2021-02-08 | 2021-06-18 | 国网上海市电力公司 | Chemical plant power transformer fault prediction method based on particle swarm and neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104748961A (en) * | 2015-03-30 | 2015-07-01 | 中国矿业大学 | Gear fault diagnosis method based on SVD decomposition and noise reduction and correlation EEMD entropy features |
| CN105116208A (en) * | 2015-08-18 | 2015-12-02 | 昆明理工大学 | Method for diagnosing commutation failure of extra-high-voltage DC power transmission system |
| CN107292453A (en) * | 2017-07-24 | 2017-10-24 | 国网江苏省电力公司电力科学研究院 | A kind of short-term wind power prediction method based on integrated empirical mode decomposition Yu depth belief network |
-
2019
- 2019-07-24 CN CN201910672418.XA patent/CN110501603A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104748961A (en) * | 2015-03-30 | 2015-07-01 | 中国矿业大学 | Gear fault diagnosis method based on SVD decomposition and noise reduction and correlation EEMD entropy features |
| CN105116208A (en) * | 2015-08-18 | 2015-12-02 | 昆明理工大学 | Method for diagnosing commutation failure of extra-high-voltage DC power transmission system |
| CN107292453A (en) * | 2017-07-24 | 2017-10-24 | 国网江苏省电力公司电力科学研究院 | A kind of short-term wind power prediction method based on integrated empirical mode decomposition Yu depth belief network |
Non-Patent Citations (1)
| Title |
|---|
| 罗璐: "基于经验模态分解的特高压直流输电***换相失败识别", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112446430A (en) * | 2020-11-27 | 2021-03-05 | 云南电网有限责任公司电力科学研究院 | Fault identification method for direct-current power transmission system |
| CN112990546A (en) * | 2021-02-08 | 2021-06-18 | 国网上海市电力公司 | Chemical plant power transformer fault prediction method based on particle swarm and neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zou et al. | Novel transient-energy-based directional pilot protection method for HVDC line | |
| CN102435921A (en) | Method for determining insulation and lightning impulse withstanding properties of same-tower double-loop power transmission line | |
| CN103760464A (en) | Small current grounding system fault line selecting method based on analytic graph solving and SVM | |
| CN110247420B (en) | Intelligent fault identification method for HVDC transmission line | |
| CN106849150A (en) | A kind of commutation failure Predictive Control System and method based on harmonic voltage detection | |
| CN111985412B (en) | High-voltage direct-current transmission line lightning stroke interference identification method | |
| CN104201671B (en) | A kind of static electric voltage stability appraisal procedure of the three-phase imbalance power distribution network containing wind-powered electricity generation | |
| CN108233435A (en) | Quantitative evaluation index system and comprehensive evaluation method for extra-high voltage access mode | |
| CN101915888A (en) | Extensible fusion identification method for lightening interference of +/-800kV direct current transmission line | |
| Zeng et al. | Faulty feeder selection and segment location method for SPTG fault in radial MMC‐MVDC distribution grid | |
| CN108258662A (en) | A kind of multiterminal flexible direct current distribution line transient protection method and device | |
| CN104617576B (en) | Multi-DC-feed (multi-direct current-feed) AC (alternate current) grid fault calculating method taking DC control characteristics into consideration | |
| CN109917175A (en) | It is a kind of for high anti-back-out when overvoltage method for quick predicting | |
| CN110501603A (en) | Utilize the forever rich direct-current commutation failure method for diagnosing faults of EMD and neural network | |
| Panahi et al. | Fast low frequency fault location and section identification scheme for VSC-based multi-terminal HVDC systems | |
| Zhu et al. | Multiple random forests based intelligent location of single-phase grounding fault in power lines of DFIG-based wind farm | |
| CN106033894B (en) | Method for judging stability of extra-high voltage direct current multi-drop point power grid | |
| CN104392397A (en) | Entropy weight method based short-circuit current suppression scheme evaluation method | |
| Liu et al. | A novel pilot directional protection scheme for HVDC transmission line based on specific frequency current | |
| Xie et al. | Online parameter determination based adaptive single‐phase reclosing scheme for wind‐powered outgoing lines with shunt reactors | |
| CN104979808B (en) | A kind of inverter calculation of penetration level method counted and longitudinal difference protection influences | |
| Niu et al. | A novel fault location and recognition method for low voltage active distribution network | |
| Cui et al. | Generalized regression neural networks based HVDC transmission line fault localization | |
| CN110456200B (en) | Method for identifying internal and external faults of rectifying station area of high-voltage direct-current transmission system | |
| Kheirollahi et al. | Developing a new fault location topology for DC microgrid systems |
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 |
Application publication date: 20191126 |
|
| RJ01 | Rejection of invention patent application after publication |