CN112147462A - Power transmission line fault identification method based on deep learning - Google Patents
Power transmission line fault identification method based on deep learning Download PDFInfo
- Publication number
- CN112147462A CN112147462A CN202010971262.8A CN202010971262A CN112147462A CN 112147462 A CN112147462 A CN 112147462A CN 202010971262 A CN202010971262 A CN 202010971262A CN 112147462 A CN112147462 A CN 112147462A
- Authority
- CN
- China
- Prior art keywords
- fault
- transmission line
- deep learning
- phase
- current
- 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
- 230000005540 biological transmission Effects 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000013135 deep learning Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims description 15
- 230000016507 interphase Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 239000013598 vector Substances 0.000 claims description 11
- 210000002569 neuron Anatomy 0.000 claims description 10
- 230000007704 transition Effects 0.000 claims description 9
- 238000013136 deep learning model Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 7
- 238000007477 logistic regression Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 235000017166 Bambusa arundinacea Nutrition 0.000 claims description 3
- 235000017491 Bambusa tulda Nutrition 0.000 claims description 3
- 241001330002 Bambuseae Species 0.000 claims description 3
- 208000025274 Lightning injury Diseases 0.000 claims description 3
- 235000015334 Phyllostachys viridis Nutrition 0.000 claims description 3
- 239000011425 bamboo Substances 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 6
- 238000013145 classification model Methods 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000014509 gene expression Effects 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R23/00—Arrangements for measuring frequencies; Arrangements for analysing frequency spectra
- G01R23/16—Spectrum analysis; Fourier analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R27/00—Arrangements for measuring resistance, reactance, impedance, or electric characteristics derived therefrom
- G01R27/02—Measuring real or complex resistance, reactance, impedance, or other two-pole characteristics derived therefrom, e.g. time constant
- G01R27/08—Measuring resistance by measuring both voltage and current
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/52—Testing for short-circuits, leakage current or ground faults
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
A transmission line fault identification method based on deep learning comprises the steps of extracting fault current and voltage signals in a wave recording system; extracting and analyzing the characteristic information of the fault current and voltage signals by adopting a Fourier analysis technology; then calculating fundamental wave components and third harmonic components of zero sequence current and zero sequence voltage; and finally, carrying out specific fault identification according to the classification model. The judging method comprises a linear classification module and a nonlinear classification module, wherein the linear classification is to preliminarily classify data samples through zero-sequence current and zero-sequence voltage; on the basis, according to the data characteristics of the transmission line faults, deep learning is selected to classify the multiple types of fault data, and finally, the transmission line fault identification is realized.
Description
Technical Field
The invention relates to a power transmission line fault identification method based on deep learning, and belongs to the technical field of power lines.
Background
The geographical environment of a power transmission line in China is severe, the coverage area of a power transmission network is wide, some lines can pass through landforms such as mountains, rivers, seas and plains, and are exposed outdoors for a long time, and are often influenced by severe weather such as thunder, strong wind, rain, fog, ice and snow or are damaged by external force, so that tripping accidents are very easy to happen.
After the fault happens, the electric power company can arrange personnel to patrol the line to the range of fault location and positioning to find out specific fault reasons, and then organize the professional personnel to carry out fault first-aid repair. And the transmission line fault often occurs in suburbs, and it may take several hours from tripping to restoring the normal power supply of the line, which will cause great influence on social economy and people's life. If the possible reasons causing the faults are deduced in time according to the relevant fault information after the line is tripped, the line patrol personnel can be guided to search key points, the searching time is saved, and meanwhile, a reasonable scheme for eliminating the faults is formulated, so that the method has great significance for reducing the power failure loss. Meanwhile, the reason of the line fault is determined each time and statistical analysis is carried out, hidden danger and weak links of the line can be found by the operation port, and therefore improvement measures are provided, and the operation and maintenance level of the power grid is improved. The invention can utilize the deep learning technology to identify the transmission line fault.
Disclosure of Invention
The invention aims to solve the technical problem of low efficiency of identifying faults of the existing power transmission line, realize efficient identification of the faults of the power transmission line, and provide a power transmission line fault identification method based on deep learning.
The technical scheme for realizing the invention is as follows: a power transmission line fault identification method based on deep learning comprises the following steps:
(1) and collecting fault recording data of the power transmission side and the power receiving side of the power transmission line, and extracting three-phase current and three-phase voltage signals.
(2) Calculating zero sequence current and zero sequence voltage corresponding to the three-phase current and the three-phase voltage by adopting Fourier transform, and extracting third harmonic content of the zero sequence current and the zero sequence voltage; calculating fault phase current, fundamental wave component and third harmonic component of voltage; and calculating the transition resistance of the power transmission line during the fault by using a lumped parameter method.
(3) The method comprises the steps of preliminarily dividing transmission line faults into two types by judging whether zero-sequence current exists after the transmission line faults; if the zero sequence current exists, the fault is a grounding short circuit fault, and if the zero sequence current does not exist, the fault is an interphase short circuit fault; the grounding short circuit faults comprise a single-phase grounding short circuit and a two-phase grounding short circuit; the interphase short-circuit fault includes a two-phase interphase short-circuit and a three-phase short-circuit.
(4) A phase current difference mutation method is adopted to judge whether the fault is a single-phase grounding short circuit, a two-phase interphase short circuit or a three-phase short circuit.
(5) Through a deep learning network structure, taking the third harmonic content, the time period, the month, the weather condition, the lightning strike condition and the reclosing condition of the transition resistance, the zero sequence current and the zero sequence voltage as original input parameters, carrying out deep learning network sample training after normalization processing of the input parameters, and outputting a result as a fault type found in an actual field; the fault types comprise eight types, namely, forest fire, tree and bamboo discharge, foreign matter short circuit (nonmetallic), foreign matter short circuit (metallic), crane line collision, bird flashover, lightning stroke and pollution flashover.
(6) And (5) repeating the steps (1) to (5), training and optimizing at least 300 groups of data samples collected historically, and obtaining a deep learning model.
(7) Judging the fault type of at least 50 groups of untrained data samples by using a deep learning model, verifying the fault type with an actual fault type, if the expected error rate is less than 5%, indicating that the deep learning model obtained in the step (6) is effective, and executing the step (8); if not, repeating the step (6).
(8) And (3) acquiring the three-phase current signals of the actual power transmission line to be judged after the expected error rate is met, and executing the steps (1) to (5) to judge the fault type of the actual power transmission line.
The invention extracts the third harmonic component from the phase current and phase voltage of the fault to carry out characteristic analysis and verification. In view of the discreteness of the recording sampling data, the invention adopts a Discrete Fourier Transform (DFT) method to decompose the fault phase current and phase voltage sampling data i (N), N is 0,1,2, … N-1 of the fault cycle after the fault, and the data is converted into the expansion of each subharmonic.
Where k is 0,1,2, … N-1 denotes the harmonic order, and N is the number of sampling points of one cycle.
In formula (1), when k is 1, I (1) represents a current fundamental component; when k > 1, I (k) represents the current k harmonic component; in equation (2), when k is 1, U (1) represents a voltage fundamental component; when k > 1, U (k) represents the k harmonic component of the voltage.
Therefore, the calculation formulas for the fault phase current, the fundamental component of the voltage and the third harmonic component are respectively as follows:
respectively calculating the ratio I of the third harmonic component to the fundamental wave3Ratio U of third harmonic component to fundamental wave of voltage3As an evaluation index of the harmonic content of the fault phase, the expressions are respectively as follows:
I3＝I(3)/I(1) (7)
U3＝U(3)/U(1) (8)
in view of consideration of a large number of high-frequency transient components in a fault signal in the initial fault occurrence period, the method selects recording data after one cycle of the fault occurrence time for analysis, and the attenuation of the high-frequency components is almost eliminated, so that the influence of the transient components on required parameters is remarkably reduced.
The method for calculating the transition resistance when the power transmission line has faults comprises the following steps:
FIG. 2 is a schematic diagram of an internal fault of a single-phase line, where the fault point is F, m and n are bus positions at two ends, i.e. the positions of fault recording measurement，ZS、ZRRespectively the system impedance at both ends.
According to the internal fault diagram of the single-phase line, the following two voltage equations can be obtained:
in the formula of UmThe voltage measured at the m terminal is in kV; i ismCurrent measured at the m terminal is in kA; u shapenThe voltage measured at the n terminal is in kV; i isnThe current measured for the n terminal is in kA; z is the impedance of the unit length of the line and the unit omega/km; dLThe total length of the line is km; dmFThe distance from the m end to the fault point F is in unit km; rFThe unit is the transition resistance of a fault point, namely omega; i isFShort-circuit current at the fault point is in kA.
The above expression applies only to single-phase circuits. Symmetric component methods are widely used in asymmetric fault analysis in three-phase power systems.
The following unified expression of the voltage equation can be written for different nets:
in the formula, J is a sequence network number and is determined by the fault type, and the three-phase short circuit J is 1; for two-phase short circuit, J is 1, 2; for a short to ground, J is 1,2, 0.
From the above ranging expression, it can be known that fault ranging can be performed using any sequence quantity. However, of the three positive, negative, and zero sequences, only the positive sequence is applicable to all types of faults. Under the condition of using positive sequence quantity to make distance measurement, for asymmetric short circuit and earth fault, it also can use negative sequence quantity or zero sequence quantity to make distance measurement calculation at the same time, and uses its result and distance measurement result using positive sequence quantity to make comparison so as to make distance measurement result more reliable. It should also be noted that the use of positive sequence ranging is also beneficial in eliminating the effects of parallel line mutual inductance.
Simultaneously solving equations (11) and (12) to eliminate the fault point voltage UFJThe fault distance can be found.
The sequence components of the short-circuit current at the fault point are obtained by the following formula:
the transition resistance is calculated as follows:
the deep learning network structure comprises three parts: an input layer, a hidden layer and an output layer. Depending on the application scenario, the input layer is typically normalized to an input vector between 0 and 1.
And the three hidden layers form the RBM network. And the output layer for classification adopts a Softmax function, and the Regression adopts a Regression function.
1. Input layer
And the normalized input vector is used as an input layer of the deep learning network and forms an RBM (visible layer non-binary) with a first hidden layer. Compared with the traditional neural network, the deep learning network can extract the signal characteristics by increasing the network depth without the problems of gradient decrement and local optimum, and can directly use the extracted characteristic quantity as an input layer after normalization.
2. Hidden layer
The hidden layer of the deep learning network is formed by stacking a plurality of RBMs. The hidden layer of the front layer RBM is stacked in sequence as the visible layer of the rear layer.
The Restricted Boltzmann Machine (RBM) is a binary undirected probability graph model based on an energy model, and the structure of the RBM is shown in fig. 4. Assuming that the visible layer neurons are nvDimension binary state vector v, hidden layer neuron nhA dimension binary state vector h, whose corresponding offset vectors are respectively
Is a weight matrix connecting the visible layer neurons with the hidden layer neurons. For state (v, h), its joint distribution function can be expressed as:
wherein, Z is a distribution function used for probability normalization, and E (v, h) is an energy function of RBM, and is defined as follows:
wherein, i and j respectively represent the ith visible layer neuron and the jth hidden layer neuron. As shown in (17), when the state energy E (v, h) of the RBM network is the lowest, the probability of the state is the highest. The network in this state is also the most stable in the physical sense of the energy function.
Because of the connectionless structure between RBM layers, the visible layer and the hidden layer of the RBM are conditionally independent. By combining the equations (16) and (17) and the above attributes, a more common neural network activation function probability equation can be derived:
wherein sigma is sigmoid function. By virtue of equation (18) above, the RBM can approximate the state of one of the layers, knowing the state of the other layer, using the Gibbs sampling algorithm. Although the structure of the RBM is simpler than that of a general neural network, the solution space of the partition function is quite large
It is also quite complex to solve. Therefore, the training of the RBM employs a contrast divergence algorithm to accelerate the Gibbs sampling process.
3. Output layer
The present invention applies a deep learning network to the problem of identification and classification, and therefore, a Logistic Regression (LR) layer is superimposed on the output layer of the deep learning network. LR is a probabilistic linear classifier, formed by a weight matrix wlrAnd an offset vector blrAnd (4) forming. The probability that the input vector x belongs to the ith class, i.e., the class variable Y ═ i, is written as:
to obtain a more direct classification result, the most probable result can be chosen as output using the Argmax { P } function, i.e.:
yp＝Arg maxi{P(Y＝i|x,wlr,blr)} (20)
the method has the advantages that the method is based on different reason types through research and analysis on the real wave recording data extracted from the fault site. The method comprises the steps of deeply mining two aspects of external meteorological factors and internal factors of fault occurrence respectively, and summarizing characteristic rules in the aspects of weather, seasons, time, reclosing, three-phase current and voltage third harmonic content of a fault line and transition resistance properties. And integrating external meteorological features and internal features obtained by recording, and establishing a nonlinear corresponding relation between fault features and categories by using a deep learning algorithm to realize fault cause identification.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic diagram of a single phase line internal fault;
FIG. 3 is a diagram of a deep learning network architecture with three hidden layers;
fig. 4 is a diagram of a restricted boltzmann machine.
Detailed Description
The embodiments are described in detail below with reference to the accompanying drawings.
In order to verify the validity of the power transmission line fault identification model and the reasonability of parameter setting, wave recording data of a power company in a certain area are collected to form a test set for verification. The method comprises the following specific steps:
step 1, selecting data from a historical database of a wave recording system to adjust parameters of a fault classification method. The algorithm of the embodiment uses the recording current and voltage, so that the transmission line recording current and voltage sequence is obtained by converting the recording file according to a set data conversion rule; then, a current and voltage sequence text is generated, and the acquisition interval of the sequence data is 0.3125ms, namely 3200 equal intervals of data are acquired every second. When parameters are adjusted, data in a proper time period are required to be selected as a training data set and a testing data set, the data of the parameter training set and the testing data set in the embodiment comprise data when power transmission lines of 35kV, 110kV, 220kV and 500kV have faults, the reasonability of parameter setting is guaranteed by the diversity of training samples, short-circuit faults of the power transmission lines which occur from 2012 1 month to 2020 month 3 of a power grid company in a certain area are counted, and 300 groups of typical data are selected as samples to perform training and verification of a classification model. 200 groups of data samples are used as training data; 100 sets of samples were used as test data to verify the validity of the method and the accuracy of the parameters.
Step 2, data preprocessing
And carrying out data preprocessing on the wave recording data in the training set. And decomposing fault current and voltage signals through Fourier transformation.
The fundamental component, the third harmonic component, of the fault current and voltage is calculated.
In the embodiment, a Discrete Fourier Transform (DFT) method is adopted to decompose the fault phase current and phase voltage sample data i (N) of one cycle after the fault, where N is 0,1,2, … N-1, and convert into expansion of each harmonic:
where k is 0,1,2, … N-1 denotes the harmonic order, and N is the number of sampling points of one cycle.
In formula (1), when k is 1, I (1) represents a current fundamental component; when k > 1, I (k) represents the current k harmonic component; in equation (2), when k is 1, U (1) represents a voltage fundamental component; when k > 1, U (k) represents the k harmonic component of the voltage.
Therefore, the calculation formulas for the fault phase current, the fundamental component of the voltage and the third harmonic component are respectively as follows:
respectively calculating the ratio I of the third harmonic component to the fundamental wave3Ratio U of third harmonic component to fundamental wave of voltage3As fault phase harmonicThe evaluation indexes of the quantity are respectively expressed as follows:
I3＝I(3)/I(1) (7)
U3＝U(3)/U(1) (8)
step 3, preliminarily dividing the transmission line faults into two types by judging whether the zero sequence current exists after the transmission line faults exist, wherein if the zero sequence current exists, the transmission line faults are grounding short-circuit faults, and if the zero sequence current does not exist, the transmission line faults are interphase short-circuit faults; the grounding short circuit faults comprise a single-phase grounding short circuit and a two-phase grounding short circuit; the interphase short-circuit fault includes a two-phase interphase short-circuit and a three-phase short-circuit.
And 4, judging whether the fault is a single-phase grounding short circuit, a two-phase interphase short circuit or a three-phase short circuit by using a phase current difference mutation method.
And 5, inputting original data of a fault waveform, fault time and weather environment as training samples into a deep learning network for training, and establishing corresponding eight fault types such as forest fire, tree bamboo discharge, foreign matter short circuit (nonmetallic), foreign matter short circuit (metallic), crane line collision, bird flash, lightning stroke, pollution flash and the like.
Step 6, fault diagnosis process
Through deep research and analysis on fault recording of the power transmission line, the short-circuit channel of the mountain fire fault is formed by an air gap and a flame carbonized particle discharge arc, the flame resistivity is as high as 9.09k omega m, and the overall fault is a typical nonlinear high-resistance value ground fault. The mountain fire fault is a typical single-phase high-resistance earth fault, while the high-resistance earth fault is mostly an arc fault, so that the fault phase current of the mountain fire fault has prominent nonlinear characteristics and contains abundant harmonic waves. After the mountain fire fault trips, reclosing is mostly unsuccessful.
TABLE 1 Transmission line Fault numerical characteristic rule
Step 7, parameter training
The deep learning model is trained through 200 groups of data samples in the training set, so that the classification model can accurately determine the fault type of the power transmission line.
TABLE 2 deep learning model classification results example
Step 8, verifying the validity and accuracy of the method
To further verify the effectiveness of the proposed method of the present invention, the test set may select the recording current data when a typical fault occurs within 23 months of a certain area. The test set selected here contains 100 sets of data. Finally, the data faults in the test set are analyzed one by one, and 97 groups are correctly classified and 3 groups are wrongly classified; the diagnosis accuracy rate is 97%, the misdiagnosis rate is 3%, and the method conforms to the engineering error rate. And ending the fault identification and method verification process of the transmission line of the whole power system.
The present invention is not limited to the above embodiments, and any changes or substitutions that can be easily made by those skilled in the art within the technical scope of the present invention are also within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (6)
1. A power transmission line fault identification method based on deep learning is characterized by comprising the following steps:
(1) collecting fault recording data of transformer substations at the power transmission side and the power receiving side of the power transmission line, and extracting three-phase current and three-phase voltage signals;
(2) calculating fault phase current, fundamental wave component and third harmonic component of voltage by Fourier transform; calculating the transition resistance of the transmission line when the transmission line fails by using a lumped parameter method;
(3) the method comprises the steps of preliminarily dividing transmission line faults into two types by judging whether zero-sequence current exists after the transmission line faults; if the zero sequence current exists, the fault is a grounding short circuit fault, and if the zero sequence current does not exist, the fault is an interphase short circuit fault; the grounding short circuit faults comprise a single-phase grounding short circuit and a two-phase grounding short circuit; the interphase short-circuit fault comprises a two-phase interphase short-circuit and a three-phase short-circuit;
(4) judging whether the fault is a single-phase grounding short circuit, a two-phase interphase short circuit or a three-phase short circuit by using a phase current difference mutation method;
(5) through a deep learning network structure, taking the third harmonic content, the time period, the month, the weather condition, the lightning strike condition and the reclosing condition of the transition resistance, the zero sequence current and the zero sequence voltage as original input parameters, carrying out deep learning network sample training after normalization processing of the input parameters, and outputting a result as a fault type found in an actual field;
(6) repeating the steps (1) to (5), training and optimizing at least 300 groups of data samples collected historically to obtain a deep learning model;
(7) judging the fault type of at least 50 groups of untrained data samples by using a deep learning model, verifying the fault type with an actual fault type, if the expected error rate is less than 5%, indicating that the deep learning model obtained in the step (6) is effective, and executing the step (8); if not, repeating the step (6);
(8) and (3) acquiring the three-phase current signals of the actual power transmission line to be judged after the expected error rate is met, and executing the steps (1) to (5) to judge the fault type of the actual power transmission line.
2. The method for identifying the power transmission line fault based on the deep learning of claim 1, wherein the deep learning network structure comprises an input layer, a hidden layer and an output layer, and a logistic regression layer is superposed on the output layer of the deep learning network; the logistic regression layer is probabilistic linearClassifier based on a weight matrix wlrAnd an offset vector blrComposition is carried out; the probability that the input vector x belongs to the ith class, i.e., the class variable Y ═ i, is written as:
to obtain a more direct classification result, the most probable result is chosen as output using the Argmax { P } function, i.e.:
yp＝Argmaxi{P(Y＝i|x,wlr,blr)}
wherein, ypIs the output layer result of the deep learning network.
3. The method for identifying the transmission line fault based on the deep learning of claim 1, wherein the calculation formula of the fault phase current, the fundamental component of the voltage and the third harmonic component is as follows:
wherein U (1) represents a voltage fundamental component; u (3) represents the voltage third harmonic component; i (1) represents the current fundamental component; i (3) represents the current third harmonic component; n is the number of sampling points of one cycle.
4. The method for identifying the transmission line fault based on the deep learning of claim 1, wherein the transition resistance is calculated according to the following formula:
wherein, UFJIs the fault point voltage; i isFJIs the short circuit current at the fault point;
the sequence components of the short-circuit current at the fault point are as follows:
5. the method as claimed in claim 1, wherein the fault types include eight types, namely, forest fire, tree and bamboo discharge, foreign object short circuit (non-metallic), foreign object short circuit (metallic), crane line collision, bird flashover, lightning stroke and pollution flashover.
6. The method for identifying the transmission line fault based on the deep learning of claim 2, wherein the hidden layer is formed by stacking a plurality of RBMs. The hidden layer of the front layer RBM is used as the visible layer of the rear layer and is sequentially stacked; because the RBM has no connection structure, the visible layer and the hidden layer of the RBM are independent of each other;
the neural network activation function probability formula is:
wherein sigma is a sigmoid function; i and j respectively represent the ith visible layer neuron and the jth hidden layer neuron; v is a visible layer neuron of nvA binary state vector of dimensions; h is hidden layer neuron nhA binary state vector of dimensions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010971262.8A CN112147462A (en) | 2020-09-16 | 2020-09-16 | Power transmission line fault identification method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010971262.8A CN112147462A (en) | 2020-09-16 | 2020-09-16 | Power transmission line fault identification method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112147462A true CN112147462A (en) | 2020-12-29 |
Family
ID=73892886
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010971262.8A Pending CN112147462A (en) | 2020-09-16 | 2020-09-16 | Power transmission line fault identification method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112147462A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113033096A (en) * | 2021-03-24 | 2021-06-25 | 应急管理部四川消防研究所 | Deep learning-based building risk early warning analysis method |
| CN113092934A (en) * | 2021-03-23 | 2021-07-09 | 武汉大学 | Single-phase earth fault judgment method and system based on clustering and LSTM |
| CN113377835A (en) * | 2021-06-09 | 2021-09-10 | 国网河南省电力公司电力科学研究院 | Distribution network line power failure identification method based on long-short term memory deep learning network |
| CN113391163A (en) * | 2021-06-01 | 2021-09-14 | 国家电网公司华中分部 | Fault recording information based on-line parameter identification method and device for circuit |
| CN113762412A (en) * | 2021-09-26 | 2021-12-07 | 国网四川省电力公司电力科学研究院 | Power distribution network single-phase earth fault identification method, system, terminal and medium |
| CN114113911A (en) * | 2021-12-06 | 2022-03-01 | 国网山东省电力公司电力科学研究院 | Fault waveform-based fault type discrimination method and discrimination system |
| CN115932484A (en) * | 2023-02-15 | 2023-04-07 | 重庆大学 | Method and device for identifying and ranging faults of power transmission line and electronic equipment |
| CN116560894A (en) * | 2023-07-10 | 2023-08-08 | 天之翼(苏州)科技有限公司 | Unmanned aerial vehicle fault data analysis method, server and medium applying machine learning |
| CN117216485A (en) * | 2023-11-09 | 2023-12-12 | 国网山东省电力公司电力科学研究院 | Objective weighting-based power transmission wave-recording bird damage fault judging method and system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104267313A (en) * | 2014-09-25 | 2015-01-07 | 广东易事特电源股份有限公司 | Micro grid system feeder line fault type distinguishing and fault point positioning method |
| CN108089099A (en) * | 2017-12-18 | 2018-05-29 | 广东电网有限责任公司佛山供电局 | The diagnostic method of distribution network failure based on depth confidence network |
| CN109239527A (en) * | 2018-09-07 | 2019-01-18 | 福州大学 | Distribution network failure recognition methods based on depth confidence network |
| CN109270407A (en) * | 2018-11-16 | 2019-01-25 | 国网山东省电力公司电力科学研究院 | Extra high voltage direct current transmission line failure cause discrimination method based on Multi-source Information Fusion |
| CN110376497A (en) * | 2019-08-12 | 2019-10-25 | 国网四川电力服务有限公司 | Low-voltage distribution system series fault arc identification method based on full-phase deep learning |
| CN110687395A (en) * | 2019-09-23 | 2020-01-14 | 南京理工大学 | Fault line selection method for power distribution network with distributed power supply based on deep belief network |
| CN111537841A (en) * | 2020-06-30 | 2020-08-14 | 上海交通大学 | Optimization method and system suitable for ground fault type identification |
-
2020
- 2020-09-16 CN CN202010971262.8A patent/CN112147462A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104267313A (en) * | 2014-09-25 | 2015-01-07 | 广东易事特电源股份有限公司 | Micro grid system feeder line fault type distinguishing and fault point positioning method |
| CN108089099A (en) * | 2017-12-18 | 2018-05-29 | 广东电网有限责任公司佛山供电局 | The diagnostic method of distribution network failure based on depth confidence network |
| CN109239527A (en) * | 2018-09-07 | 2019-01-18 | 福州大学 | Distribution network failure recognition methods based on depth confidence network |
| CN109270407A (en) * | 2018-11-16 | 2019-01-25 | 国网山东省电力公司电力科学研究院 | Extra high voltage direct current transmission line failure cause discrimination method based on Multi-source Information Fusion |
| CN110376497A (en) * | 2019-08-12 | 2019-10-25 | 国网四川电力服务有限公司 | Low-voltage distribution system series fault arc identification method based on full-phase deep learning |
| CN110687395A (en) * | 2019-09-23 | 2020-01-14 | 南京理工大学 | Fault line selection method for power distribution network with distributed power supply based on deep belief network |
| CN111537841A (en) * | 2020-06-30 | 2020-08-14 | 上海交通大学 | Optimization method and system suitable for ground fault type identification |
Non-Patent Citations (3)
| Title |
|---|
| 余达: ""基于深度学习的风力发电系统故障在线诊断研究"", 《中国优秀硕士学位论文全文数据库》, 15 December 2018 (2018-12-15), pages 1 - 96 * |
| 张然: ""基于深度置信网络的电能质量扰动检测与分类"", 《广东电力》, 24 June 2020 (2020-06-24), pages 92 - 98 * |
| 陈子明: ""基于嵌入式并行运算架构深度...电力线路非正常状态分析研究"", 《中国优秀博士学位论文全文数据库》, 15 June 2020 (2020-06-15), pages 1 - 113 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113092934A (en) * | 2021-03-23 | 2021-07-09 | 武汉大学 | Single-phase earth fault judgment method and system based on clustering and LSTM |
| CN113092934B (en) * | 2021-03-23 | 2022-05-13 | 武汉大学 | Single-phase earth fault judgment method and system based on clustering and LSTM |
| CN113033096A (en) * | 2021-03-24 | 2021-06-25 | 应急管理部四川消防研究所 | Deep learning-based building risk early warning analysis method |
| CN113391163A (en) * | 2021-06-01 | 2021-09-14 | 国家电网公司华中分部 | Fault recording information based on-line parameter identification method and device for circuit |
| CN113377835A (en) * | 2021-06-09 | 2021-09-10 | 国网河南省电力公司电力科学研究院 | Distribution network line power failure identification method based on long-short term memory deep learning network |
| CN113762412A (en) * | 2021-09-26 | 2021-12-07 | 国网四川省电力公司电力科学研究院 | Power distribution network single-phase earth fault identification method, system, terminal and medium |
| CN114113911A (en) * | 2021-12-06 | 2022-03-01 | 国网山东省电力公司电力科学研究院 | Fault waveform-based fault type discrimination method and discrimination system |
| CN115932484A (en) * | 2023-02-15 | 2023-04-07 | 重庆大学 | Method and device for identifying and ranging faults of power transmission line and electronic equipment |
| CN115932484B (en) * | 2023-02-15 | 2023-07-18 | 重庆大学 | Power transmission line fault identification and fault location method and device and electronic equipment |
| CN116560894A (en) * | 2023-07-10 | 2023-08-08 | 天之翼(苏州)科技有限公司 | Unmanned aerial vehicle fault data analysis method, server and medium applying machine learning |
| CN116560894B (en) * | 2023-07-10 | 2023-09-01 | 天之翼(苏州)科技有限公司 | Unmanned aerial vehicle fault data analysis method, server and medium applying machine learning |
| CN117216485A (en) * | 2023-11-09 | 2023-12-12 | 国网山东省电力公司电力科学研究院 | Objective weighting-based power transmission wave-recording bird damage fault judging method and system |
| CN117216485B (en) * | 2023-11-09 | 2024-01-30 | 国网山东省电力公司电力科学研究院 | Objective weighting-based power transmission wave-recording bird damage fault judging method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112147462A (en) | Power transmission line fault identification method based on deep learning | |
| CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
| CN109307824B (en) | Clustering-based power distribution network single-phase earth fault section positioning method | |
| CN103344875A (en) | Classification line selection method for single-phase earth fault of resonance earthing system | |
| CN110247420B (en) | Intelligent fault identification method for HVDC transmission line | |
| CN110609200A (en) | Power distribution network earth fault protection method based on fuzzy metric fusion criterion | |
| CN111999591B (en) | Method for identifying abnormal state of primary equipment of power distribution network | |
| CN107462810A (en) | A kind of fault section location method suitable for active power distribution network | |
| CN113447758A (en) | Single-phase earth fault distance measuring method for multi-branch current collection circuit of wind power plant | |
| CN115932484B (en) | Power transmission line fault identification and fault location method and device and electronic equipment | |
| Wu et al. | Transmission line fault cause identification method based on transient waveform image and MCNN-LSTM | |
| Mahela et al. | A protection scheme for distribution utility grid with wind energy penetration | |
| CN112255493A (en) | Multi-criterion comprehensive voting power distribution network fault line selection method | |
| CN113567810B (en) | Method, device and system for positioning arc grounding fault section of power distribution network | |
| CN110927539B (en) | Power distribution network fault section positioning method based on single-ended transient energy spectrum similarity | |
| Dehghani et al. | Distribution feeder classification based on self organized maps (case study: Lorestan province, Iran) | |
| CN114325236B (en) | Power distribution network fault identification method and system based on frequency spectrum entropy and random forest | |
| Ding et al. | Research on power grid fault diagnosis method based on PMU data and convolutional neural network | |
| Wei et al. | Single-Pole Ground Fault Location Method for Multi-terminal Flexible DC Transmission Lines Based on LSSVM | |
| CN115809761B (en) | Voltage quality analysis method and system based on low-voltage transformer area | |
| CN109861178A (en) | 220kV power grid complicated protection action determination method based on SCADA steady state information | |
| CN113625099B (en) | Layered positioning method for fault section of power distribution network | |
| Sun et al. | Transmission Line Fault Diagnosis Method Based on Improved Multiple SVM Model | |
| Liu et al. | Fuzzy Inference Model for Recognition of Single-Phase-to-Ground Fault Caused by Insulation Deterioration in Medium-Voltage Distribution Networks | |
| Xu et al. | Research on single-phase fault line selection method of distribution network based on knowledge graph |
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 |