CN110069886A - The identification of cable initial failure and classification method based on VMD and CNN - Google Patents
The identification of cable initial failure and classification method based on VMD and CNN Download PDFInfo
- Publication number
- CN110069886A CN110069886A CN201910367000.8A CN201910367000A CN110069886A CN 110069886 A CN110069886 A CN 110069886A CN 201910367000 A CN201910367000 A CN 201910367000A CN 110069886 A CN110069886 A CN 110069886A
- Authority
- CN
- China
- Prior art keywords
- layer
- modal
- cnn
- vmd
- signal
- 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
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 48
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 48
- 239000013598 vector Substances 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 20
- 230000000670 limiting effect Effects 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 11
- 238000013507 mapping Methods 0.000 claims description 8
- 238000001228 spectrum Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 4
- 230000003190 augmentative effect Effects 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000012423 maintenance Methods 0.000 abstract description 3
- 238000010276 construction Methods 0.000 abstract 1
- 239000010410 layer Substances 0.000 description 52
- 238000000605 extraction Methods 0.000 description 16
- 230000005284 excitation Effects 0.000 description 7
- 230000009467 reduction Effects 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 230000002829 reductive effect Effects 0.000 description 3
- 230000002238 attenuated effect Effects 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 239000011229 interlayer Substances 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000016776 visual perception Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010891 electric arc Methods 0.000 description 1
- 230000016507 interphase Effects 0.000 description 1
- 230000002427 irreversible effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002459 sustained effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Complex Calculations (AREA)
Abstract
The identification of cable initial failure and classification method based on VMD and CNN, comprising the following steps: Step 1: obtaining analog signal to be measured；Step 2: choosing bandwidth limiting factors α, noise margin τ and mode decomposition number K as parameter and is arranged parameter value；Step 3: carrying out variation mode decomposition to all kinds of analog signals, each mode and its centre frequency are obtained, realizes that frequency band divides；Modal characteristics and construction feature vector are decomposed Step 4: extracting；Step 5: various types of signal feature vector is inputted convolutional neural networks, tune, which is participated in training, to be practiced and obtains classification results；By using this method, accurately cable initial failure and overcurrent disturbance can be distinguished, cable maintenance is completed in time before initial failure becomes permanent fault, maintain the stable operation of power grid.
Description
Technical Field
The invention relates to the technical field of cable early fault identification, in particular to a cable early fault identification and classification method based on VMD and CNN.
Background
The development process of the fault of the cable, which is a main device for information transmission of the power system, is generally divided into three stages: partial discharge period, early failure period, and permanent failure period. In the using process of the cable, due to the defects, corrosion or aging of the insulating layer, a series of partial discharge pulses occur firstly to form electrical branches or water branches, and the electrical branches or water branches can develop into early faults accompanied with electric arcs along with further deterioration; early failures may recur after the first occurrence until they become irreversible permanent failures. The occurrence of early faults in cables is uncertain and the current at the occurrence of a fault is very small and therefore insufficient to cause the safety protection of conventional overcurrent detecting devices. Meanwhile, due to the existence of some similar disturbances, such as overcurrent problems caused by transformer excitation inrush current, constant impedance and capacitance switching faults, most existing cable fault early-stage identification methods have condition limitation, the identification rate is not high, and the cable early-stage fault and the overcurrent disturbance cannot be accurately distinguished and identified under the influence of the overcurrent disturbance.
Disclosure of Invention
The invention aims to solve the problem that early cable faults and over-current disturbance cannot be accurately distinguished and identified, and provides a VMD and CNN-based early cable fault identification and classification method.
The invention is mainly realized by the following technical scheme:
the cable early fault identification and classification method based on the VMD and the CNN comprises the following steps:
step one, acquiring an analog signal to be detected;
selecting a bandwidth limiting factor α, a noise margin tau and a modal decomposition number K as parameters and setting parameter values;
step three, carrying out variation mode decomposition on various analog signals to obtain each mode and the center frequency thereof, and realizing frequency band division;
extracting decomposition modal characteristics and constructing characteristic vectors;
and fifthly, inputting the characteristic vectors of various signals into a convolutional neural network, performing parameter adjustment training and obtaining a classification result.
The cable early self-clearing fault is generally a single-phase earth fault and is easy to cause an inter-phase earth early fault, and typical fault types mainly comprise a half-cycle early fault and a multi-cycle early fault. The early failure characteristics of the cable are summarized as follows: the duration is short or the current amplitude is low, the fault occurs near the voltage peak value and the half-cycle early fault lasts for 1/4 cycles, and when the current crosses the zero point, the fault disappears automatically; the multi-cycle early fault generally lasts for 1-4 cycles, when the electric arc disappears, the fault is automatically eliminated, although the cable early fault has obvious characteristics, if the original fault is not ideal in direct recognition effect, the method adopting the variational mode decomposition in the invention assumes that each mode of decomposition is a limited bandwidth with corresponding central frequency, the optimal decomposition problem is converted into the variational constraint problem, each mode and the central frequency thereof are solved by using an alternative direction multiplier method, the frequency band division is effectively realized, compared with other decomposition algorithms, the variational mode decomposition has firmer theoretical basis and robustness, the variational mode decomposition method separates the input signals from low frequency to high frequency in sequence, and the cable early fault signals and the characteristic information of similar disturbance different frequency bands can be effectively distinguished; then, parameter setting is completed according to signal characteristics, but the decomposed multilayer signals have large data volume and are directly input into a convolutional neural network for identification, so that not only is the difficulty of network parameter selection increased, but also the training time is too long, and in the method, the modal signals are input into the convolutional neural network for identification after characteristic extraction; the multilayer structure characteristics of the convolutional neural network local sensing and weight sharing enable the convolutional neural network local sensing and weight sharing to perform secondary deep feature learning on the basis of the variable modal decomposition feature extraction, more valuable information is mined, the accuracy of cable early fault classification is improved, further, cable early faults and overcurrent disturbance are effectively distinguished and identified, cable maintenance is completed in time before the early faults become permanent faults, and stable operation of a power grid is maintained; in the invention, VMD is English abbreviation of variational modal decomposition, and CNN is English abbreviation of convolutional neural network.
The two parameter values in the second step are respectively bandwidth limiting factors α of 2000, noise margin τ of 0 and modal decomposition number K of 7, the bandwidth limiting factor α is a parameter affecting the bandwidth of the decomposed signal, as α increases, the frequency of each decomposed signal on both sides is attenuated faster with reference to the central frequency, the bandwidth of the decomposed signal is smaller, otherwise, as α is smaller, the signal on both sides of the central frequency is attenuated slower, and the bandwidth of the decomposed signal is larger, therefore, when the frequency range of the input signal is large, α values are small and are about hundreds, when the frequency concentration range of the input signal is small, α values are large and about tens of thousands of times, the frequency spectrum of 5 types of signals such as half-cycle early fault, multi-cycle early fault, transformer excitation surge disturbance, constant impedance and capacitance switching fault is analyzed, the frequency distribution is found to be between 0 and 380Hz, the frequency distribution is characterized by large frequency range, large low-frequency signal content, when the bandwidth limiting factor α is 2000, signal characteristic extraction effect is good, when the noise margin of the input signal is large, the noise margin K of the decomposition parameter is calculated, the number of the K of the decomposed signal is small, and the noise reduction of the noise reduction can be achieved, when K is too many times of the noise reduction factor K, the noise reduction can be achieved, the noise reduction is too large, the number of the noise reduction can be achieved, and the noise reduction can be achieved, and the number of the decomposition signal can be reduced.
Further, the specific process of performing variational modal decomposition on the original signal in the third step is as follows:
step 3.1, pre-decomposing the input signal x (t) into K mode functions uk(t) Hilbert transform for each modality, and uk(t) the real signal becomes an analytic signal:
where δ (t) is a Dirac function, j represents an imaginary number, uk(t) is the kth modal component, representing the convolution operation;
step 3.2, pre-estimating the center frequency of each modal analysis signal and modulating the frequency spectrum to a corresponding fundamental frequency band to realize frequency mixing:
in the formula of omegakIs the center frequency of the kth modal component;
step 3.3, estimating the bandwidth of each modal component, and introducing a constraint condition to calculate the L of the gradient of the demodulation signal in the formula (2), wherein the sum of the estimated bandwidths of each mode is minimum2Norm, of the form:
wherein, { u [ [ u ] ]k}＝{u1,u2…ukDenotes K modal components, { ωk}＝{ω1,ω2…ωkDenotes the center frequencies of the K components;
step 3.4, introducing a Lagrange multiplier lambda (t) and a penalty factor α, constructing an augmented Lagrange function, and converting the formula (3) into an unconstrained variational problem in the following form:
step 3.5, model solution, the process is as follows: alternately updated by continuous iterationAndto obtain the saddle point of the Lagrangian equation (4) mentioned aboveSolutions for Fourier transform, quadratic optimizationCan be expressed as:
obtaining the updating formula of the center frequency and the Lagrange multiplier in the same way:
here, τ represents a time step as a noise reconstruction constraint;
formula (5) satisfiesConditional, iterative process terminating, obtainingAndto pairPerforming Fourier inverse transformation, wherein the real part is the modal component u in the time domain formk(t)。
Further, the decomposition modal characteristics extracted in the fourth step are peak-to-peak value, root-mean-square, center frequency, number of zero-crossing points, modal relative energy ratio and instantaneous amplitude, and in 1-K modes, the characteristic vector of any mode is constructed as FVkPeak-to-peak, root-mean-square, number of zero-crossings, modal relative energy ratio, instantaneous amplitude, center frequency]The K modal vectors are connected end to end, and each signal obtains a 1 multiplied by 6 multiplied by K one-dimensional vector F. The peak-to-peak value describes the size of a signal value floating range, and the peak-to-peak value is selected as a characteristic to mainly identify early faults of the cable from signal amplitudes of different modes; the root mean square is an effective value used for measuring the size of the signal in one period, so that the characteristic can separate a multi-period fault signal from other types of signals. The central frequency is used as an important frequency domain index of signal decomposition, and can well reflect the frequency composition of different disturbance signals; the number of zero-crossing points is used for distinguishing the non-stationary characteristics of signals in different center frequency modes; the modal relative energy ratio describes the contribution rate of each resolved mode to the overall signal; calculating the amplitude envelope of the signal by moving the fixed window width according to the instantaneous amplitude, and distinguishing the signal as a short-time or continuous fault; the central frequency is used as an important frequency domain index of signal decomposition, and can well reflect the frequency composition of different disturbance signals. The variational modal decomposition method separates input signals from low frequency to high frequency in sequence, effectively distinguishes the characteristic information of cable early fault signals and similar disturbance different frequency bands, however, the decomposed multilayer signals have larger data volume, if the decomposed multilayer signals are directly input into a convolutional neural network for identification, not only the difficulty of network parameter selection is increased, but also the training time is too long, the redundant information can be effectively reduced by extracting the characteristics of the variational modal decomposition, and the problem of low calculation efficiency is solved.
Further, the convolutional neural network in the fifth step comprises an input layer, a convolutional layer, a downsampling layer, a full-link layer and an output layer; the input layer obtains one-dimensional vector F information, the convolution layer obtains depth feature mapping of input signals, the down-sampling layer extracts and filters feature information generated by convolution, the full-connection layer splices a plurality of feature vectors and calculates various probabilities, and finally the probabilities are transmitted to the output layer. The convolutional neural network is a deep learning model with a special network structure, simulates a biological visual perception mode structure, reduces the calculated amount in the learning process by adopting convolution kernel parameter sharing and interlayer sparse connection in a hidden layer, and in addition, the down-sampling layer carries out secondary deep feature learning, so that more valuable information is mined, and early fault classification of cables can be more accurately realized.
Further, the two convolutional layers and the two downsampling layers in the convolutional neural network structure alternately appear.
Further, the convolution arithmetic expression is as follows:
f represents the input characteristic vector, p represents a convolution kernel with the size of 1 XG, b is an offset value, and C is the output result of the convolution layer; wherein G is more than or equal to 1 and less than or equal to G, and n is 6 xK-G + 1; (ii) a
The convolution is followed by the use of an activation function to help express more complex feature maps, and the overall process of convolution and activation is as follows:
where f is the activation function, expressed as:
wherein, Cl jFor the result of the jth feature mapping at the l level, FlAnd blRespectively, the input signal and the offset value of the first layer,is the (l +1) th layerThe weight coefficients of the jth convolution kernel.
The convolution layer can realize the feature extraction of an input signal, the convolution layer internally comprises a plurality of convolution kernels, each element forming the convolution kernels is provided with a corresponding weight coefficient and an offset value, the weight coefficient and the offset value are equivalent to neurons of a feedforward neural network, the convolution kernels sweep the input signal according to a certain step length, convolution operation is completed in each part corresponding to the convolution kernels, the offset values are superposed, the convolution result output of each convolution kernel and the current input is used as the input of the next layer of neural network, and an activation function is usually used after convolution and can be used for assisting in expressing more complex feature mapping.
Further, the down-sampling layer formula is expressed as:
assuming that the size of each feature map C of the above formula is 1M, the size of the down-sampled area is 1 d, whereD is the down sampling layer output result.
Mean pooling is used in the downsampling layer to preserve texture information of the signal features at the expense of feature map size.
Furthermore, the output of the down-sampling layer is spliced into a one-dimensional row vector end to end by the full connection layer.
Further, the output layer obtains the classified label by using a normalized exponential function softmax.
Compared with the prior art, the invention has the following advantages and beneficial effects: the invention adopts the variation modal decomposition feature extraction method to successfully extract different feature information under similar signals from the original signals, and meanwhile, the feature extraction of the variation modal decomposition structure can effectively reduce the problems of excessive redundant information and low calculation efficiency; and finally, the multilayer structure characteristics of local perception and weight sharing of the convolutional neural network are realized, so that on the basis of the extraction of the variational modal decomposition characteristics, secondary deep characteristic learning is also carried out, more valuable information is mined, the accuracy of early cable fault classification is improved, early cable faults and overcurrent disturbance are effectively distinguished and identified, cable maintenance is timely completed before the early faults become permanent faults, and the stable operation of a power grid is maintained.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention. In the drawings:
FIG. 1 is a flow chart based on VMD and CNN algorithms;
FIG. 2 is a current waveform diagram of an early fault of a half-cycle cable, an early fault of a multi-cycle cable, a normal signal, a disturbance of transformer excitation inrush current, a constant impedance fault, and a disturbance of capacitance switching;
FIG. 3 is a VMD result and frequency spectrum plot of a) a multicycle early cable failure and b) transformer inrush current disturbance;
FIG. 4 is a display diagram of 7 decomposition modal feature vectors;
FIG. 5 is a diagram of a CNN architecture employed in the present invention;
fig. 6 is a graph showing the influence of CNN iterative training on classification accuracy.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to examples and accompanying drawings, and the exemplary embodiments and descriptions thereof are only used for explaining the present invention and are not meant to limit the present invention.
Example 1:
as shown in fig. 1, the VMD and CNN based cable early failure identification and classification method includes the following steps:
step one, acquiring an analog signal to be detected;
selecting a bandwidth limiting factor α, a noise margin tau and a modal decomposition number K as parameters and setting parameter values;
step three, carrying out variation mode decomposition on various analog signals to obtain each mode and the center frequency thereof, and realizing frequency band division;
extracting decomposition modal characteristics and constructing characteristic vectors;
and fifthly, inputting the characteristic vectors of various signals into a convolutional neural network, performing parameter adjustment training and obtaining a classification result.
In this embodiment, an early fault of a half-cycle cable, an early fault of a multi-cycle cable, a normal signal, a constant impedance fault, a capacitor switching and a transformer excitation inrush current disturbance are selected as analog signals.
In the embodiment, the 6 types of analog signals to be detected are obtained through PSCARD/EMTDC software simulation. As shown in fig. 2, (a) in fig. 2 is an early fault of a half-cycle cable, (b) is an early fault of a multi-cycle cable, (c) is a normal signal, (d) is a disturbance of a transformer excitation inrush current, (e) is a constant impedance fault, and (f) is a disturbance of capacitance switching. As can be seen from fig. 2, all sample fault positions have randomness and the like signal waveforms are slightly different.
In order to better realize the variable mode decomposition process, parameters in the second step in the embodiment are respectively a bandwidth limiting factor α which is 2000 and a noise margin tau which is 0, and the mode decomposition number is K7. the bandwidth limiting factor α is a parameter which influences the bandwidth of a decomposed signal, when the frequency range of an input signal is very large, α has a small value and is about hundreds of times, when the frequency range of the input signal is very small, the value of α is large and is about tens of thousands of times, when the frequency range of the input signal is very large, the inventor analyzes 5 types of signal spectrums such as half-cycle early fault, multi-cycle early fault, transformer excitation surge flow disturbance, constant impedance and capacitance fault, and the like, finds that the frequency is distributed between 0 and 380Hz, the characteristics are characterized by large frequency range and large content of low-frequency signals, therefore, when the bandwidth limiting factor α is 2000, the signal characteristic extraction effect is good, the noise margin tau influences the constraint strength of Lagrange multiplier during reconstruction, generally, if the accurate reconstruction is not needed, the number of the decomposition factor can be set to 0, when the bandwidth limiting factor is 0, the decomposition number of the inventor can meet the requirements of the decomposition of the signal at the early stage, when K, the decomposition of the cable is too large, the decomposition result of the K, the cable can be observed when K, the decomposition of the cable is too many layers, the cable is too many, the fault is not too many, the fault is observed, and the fault is too many more, the fault.
In this embodiment, the specific process of performing variational modal decomposition on the original signal in the third step is as follows:
step 3.1, pre-decomposing the input signal x (t) into K ═ 7 mode functions uk(t) (k ═ 1, …,7), Hilbert transform is performed for each modality, and u is transformedk(t) the real signal becomes an analytic signal:
where δ (t) is a Dirac function, j represents an imaginary number, uk(t) is the kth modal component, representing the convolution operation;
step 3.2, pre-estimating the center frequency of each modal analysis signal and modulating the frequency spectrum to a corresponding fundamental frequency band to realize frequency mixing:
in the formula of omegakIs the center frequency of the kth modal component;
step 3.3, estimating the bandwidth of each modal component, and introducing a constraint condition to calculate the L of the gradient of the demodulation signal in the formula (2), wherein the sum of the estimated bandwidths of each mode is minimum2Norm, of the form:
wherein, { u [ [ u ] ]k}＝{u1,u2…u7Denotes 7 modal components, { ωk}＝{ω1,ω2…ω7Denotes the center frequencies of 7 modal components;
step 3.4, introducing a Lagrange multiplier lambda (t) and a penalty factor α, constructing an augmented Lagrange function, and converting the formula (3) into an unconstrained variational problem in the following form:
step 3.5, model solution, the process is as follows: alternately updated by continuous iterationAndto obtain the saddle point of the Lagrangian equation (4) mentioned aboveSolutions for Fourier transform, quadratic optimizationCan be expressed as:
obtaining the updating formula of the center frequency and the Lagrange multiplier in the same way:
here, τ represents a time step as a noise reconstruction constraint;
formula (5) satisfiesConditional, iterative process terminating, obtainingAndto pairPerforming Fourier inverse transformation, wherein the real part is the modal component u in the time domain formk(t)。
As shown in fig. 3, the present embodiment shows the VMD results and frequency spectra of a) early fault of multicycle cable and b) disturbance of transformer excitation inrush current, and the analysis process of other signals is similar. As can be seen in fig. 3, the waveforms of each pattern corresponding to different signals are very different.
The feature extraction of the variational modal decomposition can effectively reduce excessive redundant information and solve the problem of low computational efficiency, the decomposition modal features extracted in the fourth step in this embodiment are peak-to-peak values, root-mean-square values, center frequencies, zero-crossing points, modal relative energy ratios and instantaneous amplitudes, among 1 to K modalities, the feature vector of any modality is constructed as FVk ═ peak values, root-mean-square values, zero-crossing points, modal relative energy ratios, instantaneous amplitudes, center frequencies, K modal vectors are connected end to end, each signal obtains a 1 × 6 × K one-dimensional vector F, in this embodiment, K ═ 7, each signal obtains a 1 × 42 one-dimensional vector F.
Peak To Peak (PTP): PTP max (u)k[n])-min(uk[n])
uk[n]Representing a kth decomposition mode signal containing N sampling points, wherein N is more than or equal to 1 and less than or equal to N;
root Mean Square (RMS) is used to measure the magnitude of a periodic signal;
center frequency (ω)k): the VMD process is implemented directly in the frequency domain, so ω is chosenkAs a feature, the need for additional computation of frequency domain information for each decomposition modality is reduced;
the zero crossing number (ZC) is used for distinguishing the non-stationary characteristics of signals under different center frequency modes;
the Mode Relative Energy Ratio (MRER) describes the contribution rate of each resolved mode to the overall signal;
the Instantaneous Amplitude (IA) is used to distinguish short-term or sustained faults of the signal;
here, w is a window of length 20; m is 1, 2 … (N-w + 1).
The features of the 7 decomposition modes u 1-u 2 based on the above features are shown in fig. 4, and it can be seen from fig. 4 that the same feature of different signals in most feature vectors is clearly distinguished in 7 modes.
In this embodiment, the convolutional neural network in the fifth step includes an input layer, a convolutional layer, a downsampling layer, a full-link layer, and an output layer; the input layer obtains one-dimensional vector F information, the convolution layer obtains depth feature mapping of input signals, the down-sampling layer extracts and filters feature information generated by convolution, the full-connection layer splices a plurality of feature vectors and calculates various probabilities, and finally the probabilities are transmitted to the output layer. The convolutional neural network is a deep learning model with a special network structure, simulates a biological visual perception mode structure, reduces the calculated amount in the learning process by adopting convolution kernel parameter sharing and interlayer sparse connection in a hidden layer, and in addition, the down-sampling layer carries out secondary deep feature learning, so that more valuable information is mined, and early fault classification of cables can be more accurately realized.
In this embodiment, the convolutional neural network structure has two convolutional layers and two downsampling layers alternately.
In this embodiment, the convolution arithmetic expression is:
f represents the input characteristic vector, p represents a convolution kernel with the size of 1 XG, b is an offset value, and C is the output result of the convolution layer; wherein G is more than or equal to 1 and less than or equal to G, and n is 6 multiplied by 7-G + 1;
the convolution is followed by the use of an activation function to help express more complex feature maps, and the overall process of convolution and activation is as follows:
where f is the activation function, expressed as:
wherein, Cl jFor the result of the jth feature mapping at the l level, FlAnd blRespectively, the input signal and the offset value of the first layer,is the weight coefficient of the jth convolution kernel of the (l +1) th layer.
The convolution layer can realize the feature extraction of an input signal, the convolution layer internally comprises a plurality of convolution kernels, each element forming the convolution kernels is provided with a corresponding weight coefficient and an offset value, the weight coefficient and the offset value are equivalent to neurons of a feedforward neural network, the convolution kernels sweep the input signal according to a certain step length, convolution operation is completed in each part corresponding to the convolution kernels, the offset values are superposed, the convolution result output of each convolution kernel and the current input is used as the input of the next layer of neural network, and an activation function is usually used after convolution and can be used for assisting in expressing more complex feature mapping.
In this embodiment, the downsampling hierarchical formula is expressed as:
assuming that the size of each feature map C of the above formula is 1M, the size of the down-sampled area is 1 d, whereD isAnd outputting the result by a down-sampling layer.
Mean pooling is used in the downsampling layer to preserve texture information of the signal features at the expense of feature map size.
In this embodiment, the full connection layer splices the outputs of the downsampling layer end to end into a one-dimensional row vector.
In this embodiment, the output layer obtains the classified label by using a normalized exponential function softmax. The Softmax loss function can be expressed as:
pvthe score function is expressed, i.e. the probability that the sample x belongs to the v-th class is calculated. The Softmax function gives a greater probability of a correct classification and a lesser probability of an incorrect classification. And secondly, attributing the samples to the class with the highest probability, thereby realizing classification.
In this embodiment, a specific CNN structure model is shown in fig. 5, and table 1 shows CNN network structure parameters. The training parameters for CNN mainly include learning rate lr, number of samples Bs per training and number of times Ne per sample training. The learning rate lr is an important parameter in deep learning, and determines whether the objective function can converge to a local minimum value in an appropriate time. When lr is set too small, the convergence process will become very slow. When lr is set too large, the gradient may oscillate around the minimum and may not converge. In this embodiment, the inventor sets lr at different unit levels, observes a mean square error curve after comparison training, and finally sets lr equal to 1; the magnitude of the number of samples Bs per training determines the gradient value and the frequency of weight updates. When Bs is set to the entire training sample set, although the backpropagated gradient is very accurate, the computation is time consuming and tends to fall into local optima. When Bs is set to a small number of samples, the gradient calculation is inaccurate, which results in network training not converging. The number Ne of training samples is selected to be closely related to the size of Bs, and when Ne is equal to 1, it represents that each training sample batch is sent into CNN to complete a forward calculation and back propagation process. However, in training, it is not enough to train multiple sample sets iteratively once, and the fitting needs to be repeated for convergence. Through the experimental analysis of the inventor, the efficiency and the precision of network calculation are comprehensively considered, and finally the Bs is 60 and the Ne is 150. After parameter adjustment is completed, the whole training set is iterated for 30000 times. The training accuracy is shown in fig. 6.
TABLE 1 CNN structural model
The following sections compare and analyze the results of the entire implementation. The number of class 6 signal samples is shown in table 2:
TABLE 2 sample number distribution
In order to verify the effectiveness of the VMD feature extraction process, in this embodiment, a CNN classifier with the same structure is used, and the result of direct classification of the extracted features of the present invention and the original signal is compared from the viewpoint of time and accuracy, as shown in table 3, table 3 is a classification effect comparison table after the VMD feature extraction.
TABLE 3 comparison table of classification effect after VMD feature extraction
It can be seen from table 3 that although the classification accuracy of the signal after VMD feature extraction is slightly lower than that of the direct classification by 2.5%, the time is significantly shortened by 2.2 hours, and in comparison, the VMD feature extraction method has certain application values in the aspects of training speed and recognition accuracy.
In order to verify the superiority of the CNN classifier, the CNN classifier and the results of four traditional classifiers, namely the decision tree, the BP neural network, the bayes, and the support vector machine, are evaluated in this embodiment (as shown in table 4):
TABLE 4 comparison of results for different classifiers
From table 4, the classification accuracy of the decision tree, K-nearest neighbor, BP neural network, and support vector machine is 85.31%, 93.45%, 79.1%, and 80.74%, respectively. Compared with the other three traditional algorithms, the K-nearest neighbor classification effect is better and can reach 93.45%, but a certain gap is remained compared with the method. The adoption of the VMD and the CNN can accurately classify and identify the early faults of the cable while shortening the signal extraction time
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. The cable early fault identification and classification method based on the VMD and the CNN is characterized by comprising the following steps of:
step one, acquiring an analog signal to be detected;
selecting a bandwidth limiting factor α, a noise margin tau and a modal decomposition number K as parameters and setting parameter values;
step three, carrying out variation mode decomposition on various analog signals to obtain each mode and the center frequency thereof, and realizing frequency band division;
extracting decomposition modal characteristics and constructing characteristic vectors;
and fifthly, inputting the characteristic vectors of various signals into a convolutional neural network, performing parameter adjustment training and obtaining a classification result.
2. The VMD and CNN-based cable early failure identification and classification method according to claim 1, wherein the parameters in the second step are bandwidth limiting factor α -2000, noise margin τ -0, and mode decomposition number K-7.
3. The VMD and CNN-based cable early failure identification and classification method according to claim 1, wherein the specific process of performing variational modal decomposition on the original signal in the third step is as follows:
step 3.1, pre-decomposing the input signal x (t) into K mode functions uk(t) Hilbert transform for each modality, and uk(t) the real signal becomes an analytic signal:
where δ (t) is a Dirac function, j represents an imaginary number, uk(t) is the kth modal component, representing the convolution operation;
step 3.2, pre-estimating the center frequency of each modal analysis signal and modulating the frequency spectrum to a corresponding fundamental frequency band to realize frequency mixing:
in the formula of omegakIs the center frequency of the kth modal component;
step 3.3, estimating the bandwidth of each modal component, and introducing a constraint condition to calculate the L of the gradient of the demodulation signal in the formula (2), wherein the sum of the estimated bandwidths of each mode is minimum2Norm, of the form:
wherein, { u [ [ u ] ]k}＝{u1,u2…ukDenotes K modal components, { ωk}＝{ω1,ω2…ωkDenotes the center frequencies of the K components;
step 3.4, introducing a Lagrange multiplier lambda (t) and a penalty factor α, constructing an augmented Lagrange function, and converting the formula (3) into an unconstrained variational problem in the following form:
step 3.5, model solution, the process is as follows: alternately updated by continuous iterationAndto obtain the saddle point of the Lagrangian equation (4) mentioned aboveSolutions for Fourier transform, quadratic optimizationCan be expressed as:
obtaining the updating formula of the center frequency and the Lagrange multiplier in the same way:
here, τ represents a time step as a noise reconstruction constraint;
formula (5) satisfiesConditional, iterative process terminating, obtainingAndto pairPerforming Fourier inverse transformation, wherein the real part is the modal component u in the time domain formk(t)。
4. The VMD and CNN-based cable early failure identification and classification method according to claim 1, wherein the decomposed modal features extracted in the fourth step are peak-to-peak value, root-mean-square, center frequency, number of zero-crossing points, modal relative energy ratio and instantaneous amplitude, and the feature vector of any modality among 1-K modalities is constructed as FVkPeak-to-peak, root-mean-square, number of zero-crossings, modal relative energy ratio, instantaneous amplitude, center frequency]The K modal vectors are connected end to end, and each signal obtains a 1 multiplied by 6 multiplied by K one-dimensional vector F.
5. The VMD and CNN based cable early failure identification and classification method according to claim 1, wherein the convolutional neural network in the fifth step comprises an input layer, a convolutional layer, a downsampling layer, a full connection layer and an output layer; the input layer obtains one-dimensional vector F information, the convolution layer obtains depth feature mapping of input signals, the down-sampling layer extracts and filters feature information generated by convolution, the full-connection layer splices a plurality of feature vectors and calculates various probabilities, and finally the probabilities are transmitted to the output layer.
6. The VMD and CNN-based cable early failure identification and classification method of claim 5, wherein two convolutional layers and two downsampling layers alternate in the convolutional neural network structure.
7. The VMD and CNN-based cable early failure identification and classification method according to claim 6, wherein the convolution algorithm is expressed as:
f represents the input characteristic vector, p represents a convolution kernel with the size of 1 XG, b is an offset value, and C is the output result of the convolution layer; wherein G is more than or equal to 1 and less than or equal to G, and n is 6 xK-G + 1;
the convolution is followed by the use of an activation function to help express more complex feature maps, and the overall process of convolution and activation is as follows:
where f is the activation function, expressed as:
wherein,for the result of the jth feature mapping at the l level, FlAnd blRespectively, the input signal and the offset value of the first layer,is the weight coefficient of the jth convolution kernel of the (l +1) th layer.
8. The VMD and CNN-based cable early failure identification and classification method of claim 7, wherein the down-sampling hierarchy is expressed as:
assuming that the size of each feature map C of the above formula is 1M, the size of the down-sampled area is 1 d, whereD is the down sampling layer output result.
9. The VMD and CNN-based cable early failure identification and classification method of claim 8, wherein the full connectivity layer splices the outputs of the down-sampled layer end-to-end into a one-dimensional row vector.
10. The VMD and CNN-based cable early failure identification and classification method of claim 9, wherein the output layer obtains the classified label using a normalized exponential function softmax.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910367000.8A CN110069886A (en) | 2019-05-05 | 2019-05-05 | The identification of cable initial failure and classification method based on VMD and CNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910367000.8A CN110069886A (en) | 2019-05-05 | 2019-05-05 | The identification of cable initial failure and classification method based on VMD and CNN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110069886A true CN110069886A (en) | 2019-07-30 |
Family
ID=67369862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910367000.8A Pending CN110069886A (en) | 2019-05-05 | 2019-05-05 | The identification of cable initial failure and classification method based on VMD and CNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110069886A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110686899A (en) * | 2019-09-21 | 2020-01-14 | 天津大学 | Internal combustion engine noise source identification method |
| CN110766264A (en) * | 2019-07-31 | 2020-02-07 | 贵州电网有限责任公司 | Transmission cable health state monitoring and evaluating method |
| CN111130053A (en) * | 2020-01-08 | 2020-05-08 | 华南理工大学 | Power distribution network overcurrent protection method based on deep reinforcement learning |
| CN111141519A (en) * | 2019-12-27 | 2020-05-12 | 西安交通大学 | Rolling bearing fault quantitative evaluation method |
| CN111259949A (en) * | 2020-01-13 | 2020-06-09 | 山东超越数控电子股份有限公司 | Fault identification model construction method, model and identification method for aircraft environmental control system |
| CN111413588A (en) * | 2020-03-31 | 2020-07-14 | 陕西省地方电力（集团）有限公司咸阳供电分公司 | Power distribution network single-phase earth fault line selection method |
| CN111504646A (en) * | 2020-06-05 | 2020-08-07 | 合肥工业大学 | Weak signal fault feature classification method and system for early failure of bearing |
| CN111664365A (en) * | 2020-06-07 | 2020-09-15 | 东北石油大学 | Oil and gas pipeline leakage detection method based on improved VMD and 1DCNN |
| CN112051480A (en) * | 2020-08-06 | 2020-12-08 | 国网安徽省电力有限公司电力科学研究院 | Neural network power distribution network fault diagnosis method and system based on variational modal decomposition |
| CN112101144A (en) * | 2020-08-27 | 2020-12-18 | 云南电网有限责任公司昭通供电局 | Self-adaptive method for improving transformer vibration signal processing precision |
| CN112580699A (en) * | 2020-12-04 | 2021-03-30 | 中国船舶重工集团公司第七一一研究所 | Sensor data anomaly detection method and application thereof |
| CN113093050A (en) * | 2021-03-31 | 2021-07-09 | 中国矿业大学 | Cable early fault identification method and system based on cable grounding wire current time-frequency characteristics |
| CN113203914A (en) * | 2021-04-08 | 2021-08-03 | 华南理工大学 | Underground cable early fault detection and identification method based on DAE-CNN |
| CN113740671A (en) * | 2020-10-22 | 2021-12-03 | 国网河北省电力有限公司雄安新区供电公司 | Fault arc identification method based on VMD and ELM |
| WO2022067562A1 (en) * | 2020-09-29 | 2022-04-07 | 西门子股份公司 | Method and device for diagnosing fault arc, and computer-readable storage medium |
| CN115166585A (en) * | 2022-09-08 | 2022-10-11 | 石家庄科林电气股份有限公司 | Ground fault detection anti-misjudgment method and device and electronic equipment |
| WO2022227491A1 (en) * | 2021-04-27 | 2022-11-03 | 京东方科技集团股份有限公司 | Display defect detection apparatus and detection method therefor, and display defect detection system |
| CN117929952A (en) * | 2024-03-21 | 2024-04-26 | 国网(山东)电动汽车服务有限公司 | Novel arc fault detection method for electric automobile charging pile |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060112028A1 (en) * | 2004-11-24 | 2006-05-25 | Weimin Xiao | Neural Network and Method of Training |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN108896296A (en) * | 2018-04-18 | 2018-11-27 | 北京信息科技大学 | A kind of wind turbine gearbox method for diagnosing faults based on convolutional neural networks |
| CN109633271A (en) * | 2019-01-17 | 2019-04-16 | 长沙理工大学 | The traveling wave Time-Frequency Analysis Method being distributed based on variation mode decomposition and Eugene Wigner Weir |
-
2019
- 2019-05-05 CN CN201910367000.8A patent/CN110069886A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060112028A1 (en) * | 2004-11-24 | 2006-05-25 | Weimin Xiao | Neural Network and Method of Training |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
| CN108896296A (en) * | 2018-04-18 | 2018-11-27 | 北京信息科技大学 | A kind of wind turbine gearbox method for diagnosing faults based on convolutional neural networks |
| CN109633271A (en) * | 2019-01-17 | 2019-04-16 | 长沙理工大学 | The traveling wave Time-Frequency Analysis Method being distributed based on variation mode decomposition and Eugene Wigner Weir |
Non-Patent Citations (2)
| Title |
|---|
| 刘长良 等: "基于变分模态分解和模糊C均值聚类的滚动轴承故障诊断", 《中国电机工程学报》 * |
| 曾学琼: "基于卷积神经网络的变速器故障分类识别研究", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑)》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110766264A (en) * | 2019-07-31 | 2020-02-07 | 贵州电网有限责任公司 | Transmission cable health state monitoring and evaluating method |
| CN110686899B (en) * | 2019-09-21 | 2021-01-29 | 天津大学 | Internal combustion engine noise source identification method |
| CN110686899A (en) * | 2019-09-21 | 2020-01-14 | 天津大学 | Internal combustion engine noise source identification method |
| CN111141519A (en) * | 2019-12-27 | 2020-05-12 | 西安交通大学 | Rolling bearing fault quantitative evaluation method |
| CN111130053A (en) * | 2020-01-08 | 2020-05-08 | 华南理工大学 | Power distribution network overcurrent protection method based on deep reinforcement learning |
| CN111130053B (en) * | 2020-01-08 | 2021-02-19 | 华南理工大学 | Power distribution network overcurrent protection method based on deep reinforcement learning |
| CN111259949A (en) * | 2020-01-13 | 2020-06-09 | 山东超越数控电子股份有限公司 | Fault identification model construction method, model and identification method for aircraft environmental control system |
| CN111413588A (en) * | 2020-03-31 | 2020-07-14 | 陕西省地方电力（集团）有限公司咸阳供电分公司 | Power distribution network single-phase earth fault line selection method |
| CN111504646A (en) * | 2020-06-05 | 2020-08-07 | 合肥工业大学 | Weak signal fault feature classification method and system for early failure of bearing |
| CN111504646B (en) * | 2020-06-05 | 2022-10-14 | 合肥工业大学 | Weak signal fault feature classification method and system for early failure of bearing |
| CN111664365A (en) * | 2020-06-07 | 2020-09-15 | 东北石油大学 | Oil and gas pipeline leakage detection method based on improved VMD and 1DCNN |
| CN111664365B (en) * | 2020-06-07 | 2022-01-28 | 东北石油大学 | Oil and gas pipeline leakage detection method based on improved VMD and 1DCNN |
| CN112051480A (en) * | 2020-08-06 | 2020-12-08 | 国网安徽省电力有限公司电力科学研究院 | Neural network power distribution network fault diagnosis method and system based on variational modal decomposition |
| CN112101144A (en) * | 2020-08-27 | 2020-12-18 | 云南电网有限责任公司昭通供电局 | Self-adaptive method for improving transformer vibration signal processing precision |
| WO2022067562A1 (en) * | 2020-09-29 | 2022-04-07 | 西门子股份公司 | Method and device for diagnosing fault arc, and computer-readable storage medium |
| CN113740671A (en) * | 2020-10-22 | 2021-12-03 | 国网河北省电力有限公司雄安新区供电公司 | Fault arc identification method based on VMD and ELM |
| CN112580699A (en) * | 2020-12-04 | 2021-03-30 | 中国船舶重工集团公司第七一一研究所 | Sensor data anomaly detection method and application thereof |
| CN113093050A (en) * | 2021-03-31 | 2021-07-09 | 中国矿业大学 | Cable early fault identification method and system based on cable grounding wire current time-frequency characteristics |
| CN113093050B (en) * | 2021-03-31 | 2023-07-07 | 中国矿业大学 | Cable early fault identification method and system based on time-frequency characteristics of cable grounding wire current |
| CN113203914A (en) * | 2021-04-08 | 2021-08-03 | 华南理工大学 | Underground cable early fault detection and identification method based on DAE-CNN |
| WO2022227491A1 (en) * | 2021-04-27 | 2022-11-03 | 京东方科技集团股份有限公司 | Display defect detection apparatus and detection method therefor, and display defect detection system |
| CN115166585A (en) * | 2022-09-08 | 2022-10-11 | 石家庄科林电气股份有限公司 | Ground fault detection anti-misjudgment method and device and electronic equipment |
| CN115166585B (en) * | 2022-09-08 | 2022-11-29 | 石家庄科林电气股份有限公司 | Ground fault detection anti-misjudgment method and device and electronic equipment |
| CN117929952A (en) * | 2024-03-21 | 2024-04-26 | 国网(山东)电动汽车服务有限公司 | Novel arc fault detection method for electric automobile charging pile |
| CN117929952B (en) * | 2024-03-21 | 2024-05-28 | 国网(山东)电动汽车服务有限公司 | Novel arc fault detection method for electric automobile charging pile |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110069886A (en) | The identification of cable initial failure and classification method based on VMD and CNN | |
| CN109271975A (en) | A kind of electrical energy power quality disturbance recognition methods based on big data multi-feature extraction synergetic classification | |
| Luo et al. | Stacked auto-encoder based fault location in VSC-HVDC | |
| Yılmaz et al. | An improved automated PQD classification method for distributed generators with hybrid SVM-based approach using un-decimated wavelet transform | |
| CN103344875A (en) | Classification line selection method for single-phase earth fault of resonance earthing system | |
| CN108761287B (en) | Transformer partial discharge type identification method | |
| CN103728535A (en) | Extra-high-voltage direct-current transmission line fault location method based on wavelet transformation transient state energy spectrum | |
| CN108154223B (en) | Power distribution network working condition wave recording classification method based on network topology and long time sequence information | |
| CN110068759A (en) | A kind of fault type preparation method and device | |
| Ayyagari | Artificial neural network based fault location for transmission lines | |
| CN112255500A (en) | Power distribution network weak characteristic fault identification method based on transfer learning | |
| CN110543921A (en) | cable early fault identification method based on waveform learning | |
| CN111537853A (en) | Intelligent detection method for partial discharge of switch cabinet based on multi-source heterogeneous data analysis | |
| Ray et al. | Application of extreme learning machine for underground cable fault location | |
| CN114414942A (en) | Power transmission line fault identification classifier, identification method and system based on transient waveform image identification | |
| Liu et al. | Dual-channel convolutional network-based fault cause identification for active distribution system using realistic waveform measurements | |
| Uddin et al. | On the protection of power system: Transmission line fault analysis based on an optimal machine learning approach | |
| CN112946425A (en) | Fault positioning method for mining travelling wave time-frequency domain characteristics by utilizing deep learning | |
| Narasimhulu et al. | LWT based ANN with ant lion optimizer for detection and classification of high impedance faults in distribution system | |
| Ekici et al. | Comparison of different regression models to estimate fault location on hybrid power systems | |
| CN117630569A (en) | Low-current single-phase earth fault multi-criterion fusion line selection method based on GRU neural network | |
| CN114397569A (en) | Circuit breaker fault arc detection method based on VMD parameter optimization and sample entropy | |
| CN116540025A (en) | Fault detection method based on transfer learning and residual error network | |
| Saad et al. | Single-phase auto-reclosing scheme using particle filter and convolutional neural network | |
| Zhang et al. | Fault location for overhead and cable hybrid transmission lines based on the difference quantities of forward propagating waves and backward propagating waves in HVDC 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: 20190730 |
|
| RJ01 | Rejection of invention patent application after publication |