CN112051480A - Neural network power distribution network fault diagnosis method and system based on variational modal decomposition - Google Patents
Neural network power distribution network fault diagnosis method and system based on variational modal decomposition Download PDFInfo
- Publication number
- CN112051480A CN112051480A CN202010783864.0A CN202010783864A CN112051480A CN 112051480 A CN112051480 A CN 112051480A CN 202010783864 A CN202010783864 A CN 202010783864A CN 112051480 A CN112051480 A CN 112051480A
- Authority
- CN
- China
- Prior art keywords
- fault
- power distribution
- distribution network
- imf
- fault diagnosis
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 66
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000009826 distribution Methods 0.000 title claims abstract description 52
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 45
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 27
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims description 18
- 238000012216 screening Methods 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 2
- 230000009466 transformation Effects 0.000 abstract description 4
- 238000012549 training Methods 0.000 description 21
- 230000006870 function Effects 0.000 description 14
- 238000012360 testing method Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000004590 computer program Methods 0.000 description 7
- 238000004088 simulation Methods 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 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
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Locating Faults (AREA)
Abstract
The invention discloses a neural network power distribution network fault diagnosis method and system based on variational modal decomposition. And then, according to an Intrinsic Mode Function (IMF) obtained by variable Mode decomposition, selecting an IMF component with more fault characteristics, and extracting the fault characteristics through Hilbert-Huang transformation. And finally, the extracted fault characteristics are used as the input of a convolutional neural network model to realize fault positioning and fault type judgment. The method and the device can realize the fault location of the power distribution network and the fault type judgment, and have higher diagnosis precision compared with other methods. Through the selection of the CNN model and the adjustment of model parameters, the fault diagnosis precision can be obviously improved, and the time consumption of fault diagnosis can be reduced. Compared with other methods, the method can effectively improve the fault precision and has good generalization capability.
Description
Technical Field
The invention relates to the technical field of power distribution network fault diagnosis, in particular to a neural network power distribution network fault diagnosis method and system based on variational modal decomposition.
Background
The normal operation of the power distribution network has important significance for ensuring the power consumption quality of power consumers. With the development of energy and power systems, the grid connection of renewable energy sources, the access of electric equipment and the network topology structure of a power distribution network are more and more complex, and the difficulty is brought to fault diagnosis. When the distribution network breaks down, how to realize fault location and fault type discernment in time, guarantee that the trouble is in time got rid of, supply the power consumption to resume rapidly, distribution network normal operating is the problem that awaits the solution urgently.
When a power distribution network fails, the first thing that changes is the electrical quantity in the power distribution network line. Then, the changed electrical quantity causes a relay protection device of the line to act, thereby realizing fault isolation and line protection. Therefore, there are two main research starting points in the fault location research direction of the power distribution network, namely, a fault diagnosis method for the electric quantity and a fault diagnosis method for the switching quantity. When a fault occurs, the switching value may have the problems of missing report and misoperation, and the electric quantity can essentially reflect the fault condition of the line. Therefore, the electric quantity fault diagnosis method has certain advantages in fault tolerance, accuracy and the like compared with the switching value fault diagnosis method. The invention adopts a fault diagnosis method of electric quantity.
At present, the Artificial intelligence technology is rapidly developed and has been applied to various fields including power distribution Network fault diagnosis, mainly including expert systems, bayesian networks, petri networks, genetic algorithms, Artificial Neural Networks (ANN), and the like. Among them, ANN has excellent adaptability and generalization ability, and thus has been applied and studied. In the existing literature, a radial basis function neural network and fuzzy integral are combined for solving the problem of fault diagnosis of a tie line after a power grid is partitioned; and in some cases, discrete wavelet transform and ANN (artificial neural network) are combined to realize the identification and fault location of the fault section of the power distribution network. Some of the conventional ANNs are adopted for power distribution network fault diagnosis, and some of the conventional ANNs have the defects of dimension disaster, sensitive input data, incapability of sensing plane space characteristics and the like. Some points indicate that deep learning has the advantages of high-level fault extraction capability, multi-source time sequence signals, high-dimensional and nonlinear data adaptability and the like in the field of fault diagnosis.
Among many deep learning algorithms, Convolutional Neural Networks (CNNs) are not only superior to traditional ANNs in planar perceptibility, but also can handle higher-dimensional data. The existing literature uses amplitude characteristics of voltage and current and frequency characteristics extracted by wavelet transformation as fault characteristics, and uses a parallel convolution neural network to realize fault location and fault type identification of a multi-terminal direct-current transmission line. Although the above method has good fault diagnosis capability, the wavelet transformation is affected by the Heisenberg inaccurate measurement principle, and the effect is not good when processing the mutation signal.
Disclosure of Invention
The neural network power distribution network fault diagnosis method and system based on variational modal decomposition can solve the technical problems of complex structure and difficult fault diagnosis of the power distribution network, and the technical problems of ANN fault diagnosis technology and wavelet transformation.
In order to achieve the purpose, the invention adopts the following technical scheme:
a neural network power distribution network fault diagnosis method based on variational modal decomposition comprises the following steps,
s100, extracting zero sequence current signals of each line in a power distribution network fault area, and obtaining IMF components of the zero sequence current of the lines in the power distribution network fault area through variational modal decomposition;
s200, screening IMF components with characteristic information reaching a specific amount;
s300, performing Hilbert-Huang transform processing on the screened IMF components, and extracting fault features;
s400, the extracted fault characteristics are used as input signals, and fault positioning and fault type judgment are carried out by using the trained CNN fault diagnosis model.
Further, the CNN fault diagnosis model in S400 includes:
the device consists of an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a convolution layer, a pooling layer and a full-connection layer;
where the convolutional layer is defined by a set of small filters through which the input data is convolved in the forward channel, the convolutional layer output is as follows:
wherein h isj(x) Is the spatial position x ═ x1,x2) The jth inactive output feature map of (g)ijIs hjAnd the ith input channel fiA core in between;
c represents the total input channel number, b represents the deviation, and the method has the capability of improving the nonlinear segmentation of the network;
the operator is a two-dimensional convolution defined as follows:
where m × n represents the kernel size;
the filter also comprises a pooling layer, wherein the pooling layer is directly inserted into the convolution layer, and then the output of the convolution layer is screened and filtered to remove non-important characteristic parameters, so that the calculation time can be effectively reduced;
the full-connection layer is a traditional multilayer perceptron which utilizes characteristics for classification;
the Softmax function acts as an activation function for the CNN fault diagnosis model.
Further, the step S100 of obtaining the IMF component of the zero-sequence current of the line in the fault area of the power distribution network through variational modal decomposition specifically includes:
this is achieved by solving the following constrained optimization problem:
wherein u iskIs the k-th modal component, wkIs the center frequency of the kth mode, f (t) is the input signal, (t) is the dirac function;
introducing a penalty factor alpha and a Lagrange multiplier lambda, and processing the formula (3) to obtain:
obtaining the optimal solution of (4) through multiple iterations, wherein modal components and center frequencies in the iteration process are respectively expressed as:
wherein the content of the first and second substances,
are respectively as
f(t)，uk(t), λ (t) results obtained by fourier transform.
Further, the step S200 of screening out an IMF component including feature information reaching a specific amount specifically includes:
performing amplitude characteristic calculation on each IMF component after VMD decomposition by using HHT, and screening the IMF components;
at time t, the nth IMF component of the zero sequence current of the mth line is represented as IMFmn(t) the reconstructed signal P (t) obtained by HHT is
Wherein the instantaneous amplitude a (t) and phase angle
Respectively as follows:
calculating the instantaneous frequency w (t) as
The amplitude characteristic of each IMF component is obtained by equation (8).
Further, the step S300 of performing hilbert-yellow transform processing on the screened IMF components, and extracting fault features specifically includes:
performing fault feature extraction on IMF components of which the screened fault feature information reaches a set value, and extracting frequency features of signals as fault features to perform fault positioning and fault type judgment;
the IMF2 component frequency characteristic is obtained by equation (10).
On the other hand, the invention also discloses a convolutional neural network power distribution network fault diagnosis system based on variational modal decomposition, which comprises the following units:
the IMF component extraction unit is used for extracting zero-sequence current signals of all lines in the power distribution network fault region and obtaining IMF components of the zero-sequence currents of the lines in the power distribution network fault region through variational modal decomposition;
the IMF component screening unit is used for screening IMF components with characteristic information reaching a specific quantity;
the fault feature extraction unit is used for performing Hilbert-Huang transform processing on the screened IMF components and extracting fault features;
and the fault diagnosis unit is used for taking the extracted fault characteristics as input signals and utilizing the trained CNN fault diagnosis model to carry out fault positioning and fault type judgment.
According to the technical scheme, the neural network power distribution network fault diagnosis method and system based on the Variational modal Decomposition, which are disclosed by the invention, take the convolutional neural network as a main body, and firstly obtain the IMF component of the zero-sequence current of the power distribution network fault area line through Variational Modal Decomposition (VMD). And then, IMF components with rich fault characteristic information are screened out, and IMF component fault characteristic extraction is carried out by utilizing the characteristic that Hilbert-Huang transform (HHT) is not influenced by the Heisenberg inaccuracy measuring principle. And finally, the extracted fault characteristics are used as CNN input to carry out power distribution network fault diagnosis. The invention adopts an IEEE 33 node power system model for simulation. The example results show that the method provided by the invention not only can be used for fault positioning, but also can be used for judging the fault type, and has high fault diagnosis accuracy, and good generalization capability and robustness.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a fault diagnosis framework of the present invention;
FIG. 3 is a simple power system model;
FIG. 4 is the LI zero sequence current waveform and decomposition results of the present invention;
FIG. 5 is an IMF component amplitude characteristic and an IMF2 component frequency characteristic of the present invention;
FIG. 6 is a diagram of a VMD-CNN fault diagnosis architecture of the present invention;
FIG. 7 is an IEEE 33 node power system model;
FIG. 8 shows the decomposition results of the zero sequence current and VMD of the line S1 according to the embodiment of the present invention;
FIG. 9 is an IMF component amplitude characteristic and an IMF1 component frequency characteristic of an embodiment of the present invention;
FIG. 10 illustrates sample accuracy for testing embodiments of the present invention;
FIG. 11 is a comparison of fault location accuracy for different methods of embodiments of the invention;
FIG. 12 is a comparison of the failure type determination accuracy in different methods according to embodiments of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention.
As shown in fig. 1, the method for diagnosing a fault of a neural network power distribution network based on variational modal decomposition according to this embodiment includes the following steps,
s100, extracting zero sequence current signals of each line in a power distribution network fault area, and obtaining IMF components of the zero sequence current of the lines in the power distribution network fault area through variational modal decomposition;
s200, screening IMF components with characteristic information reaching a specific amount;
s300, performing Hilbert-Huang transform processing on the screened IMF components, and extracting fault features;
s400, the extracted fault characteristics are used as input signals, and fault positioning and fault type judgment are carried out by using the trained CNN fault diagnosis model.
The following is a detailed description:
1. convolutional neural network
A convolutional neural network is a supervised, feed-forward neural network (although unsupervised and recursive variations also exist). The multi-layer filter in a successfully trained convolutional network will respond to features of different complexity.
The structure of the CNN consists of an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a convolution layer, a pooling layer and a full-connection layer. Among them, convolutional layers are the most important components of CNN. The convolutional layer is mainly defined by a set of small filters (or kernels) through which the input data is convolved in the forward channel, the output of which is as follows:
wherein h isj(x) Is the spatial position x ═ x1,x2) The jth (inactive) output feature map of (g)ijIs hjAnd the ith input channel fiA core in between. C represents the total input channel number, b represents deviation, and the method has the capability of improving the nonlinear division of the network. The operator is a two-dimensional convolution defined as follows:
where m × n represents the kernel size.
Another commonly used layer is a pooling layer. Generally, after the convolution layer is directly inserted into the convolution layer, the output of the convolution layer is filtered to remove non-important characteristic parameters, and the calculation time can be effectively reduced.
The full link layer is a traditional multilayer sensor, and the function of the full link layer is to classify by using characteristics, which is commonly usedThe excitation functions include Rectified Linear Unit (ReLU), Sigmoid, Linear, Softmax, etc[20]. The Softmax function can map any real-valued vector to a value of (0,1), and has good classification capability, so the Softmax function is used as the activation function of the CNN in the invention.
2. Variational modal decomposition
The VMD processes the original signal through a non-recursive and variational modal solving mode, and has better anti-noise performance and non-stationary performance signal processing effect. VMD is implemented by solving the following constrained optimization problem:
wherein u iskIs the k-th modal component, wkIs the center frequency of the kth mode, f (t) is the input signal, and (t) is the dirac function. Introducing a penalty factor alpha and a Lagrange multiplier lambda, and processing the formula (3) to obtain:
obtaining the optimal solution of (4) through multiple iterations, wherein modal components and center frequencies in the iteration process are respectively expressed as:
wherein the content of the first and second substances,
are respectively as
f(t)，uk(t), λ (t) by Fourier transformAnd exchanging the obtained result.
3. VMD-CNN fault diagnosis strategy
The specific technical framework of the VMD-CNN power distribution network fault diagnosis method provided by the invention is shown in FIG. 2. Firstly, VMD decomposition is carried out on a zero sequence current signal of a fault area line. And then, screening IMF components containing more characteristic information. And performing Hilbert-Huang transform processing on the screened IMF components to extract fault characteristics. And finally, the extracted fault characteristics are used as input signals of the CNN to carry out fault positioning and fault type judgment.
3.1 VMD-based Fault feature extraction
Taking a simple power system simulation model as shown in fig. 3 as an example, the model has three lines L1-L3, a single-phase ground fault occurs in the line L1, and the zero-sequence current of the fault line L1 is shown in fig. 4 (a).
VMD Decomposition and Empirical Mode Decomposition (EMD) are performed on the zero-sequence current of the fault line L1, and the Decomposition results are shown in fig. 4 (b) and (c), respectively. As can be seen from the decomposition results of the graph (b) and the graph (c), the IMF components of the EMD decomposition method are as many as 5, while the IMF components decomposed by the VMD are only 3, so that the calculation scale is reduced, and the calculation time is reduced; the high-frequency component amplitude value decomposed by the VMD has less occupation ratio, and is beneficial to reducing classification errors; the regularity of the VMD decomposed components of the middle and low frequency parts is obviously stronger than that of the EMD decomposed components. From the comparison results of the two decomposition results, it can be seen that: the VMD method is more beneficial to subsequent modeling and fault diagnosis.
And performing amplitude characteristic calculation on each IMF component after VMD decomposition by using HHT, and screening the IMF components. At time t, the nth IMF component of the zero sequence current of the mth line is represented as IMFmn(t) the reconstructed signal P (t) obtained by HHT is
Wherein the instantaneous amplitude a (t) and phase angle
Respectively as follows:
calculating the instantaneous frequency w (t) as
The amplitude characteristic of each IMF component is obtained through the formula (8), as shown in fig. 5 (a), it can be seen that the IMF2 component contains more fault characteristic information of the original zero-sequence current signal, and is more suitable for subsequent modeling and fault location compared with the IMF1 and the IMF3 components. In addition, the IMF components with rich characteristic information are used for fault diagnosis, and compared with the method that all IMF components are used, the calculation scale and difficulty are reduced.
And extracting fault characteristics of the screened IMF components with more fault characteristic information, and extracting frequency characteristics of signals as fault characteristics to perform fault positioning and fault type judgment. The frequency characteristic of the IMF2 component obtained by equation (10) is shown in fig. 5 (b).
3.2VMD-CNN Fault diagnosis
The VMD-CNN-based fault diagnosis structure is shown in fig. 6, and the specific steps are as follows:
1. extracting a zero sequence current signal I of each line in a power distribution network fault area, wherein the form is as follows:
I＝{I1,I2,…,Im}
wherein ImA zero sequence current signal representing the mth line, m being 1,2,3, …;
VMD decomposition. For the zero sequence current signal I of the mth linemPerforming VMD decomposition to obtain IMFmThe form is as follows:
IMFm＝{imfm1,imfm2,…,imfmn}
wherein, imfmnAn nth IMF component obtained by VMD decomposition of a zero-sequence current signal of an mth line, wherein n is 1,2,3, …;
3. the IMF component is screened. For IMFmAmplitude characteristic calculation is carried out, components containing a large amount of fault information are screened out, and a vector X is constructedmThe form is as follows:
Xm＝{xm1,xm2,…,xmk}
wherein x ismkRepresentation of IMFmThe k-th IMF component, k is 1,2,3, …;
4. and extracting fault characteristics. For xmkHHT transform is performed to obtain xmkFrequency characteristic fmkConstructing a fault feature vector FmThe form is as follows:
Fm＝{fm1,fm2,…,fmk}
5. and (5) training the CNN and carrying out fault diagnosis on the power distribution network. Using fault feature vectors F selected from a number of samplesmAnd (5) training the CNN, and performing fault positioning and fault type judgment.
4 example analysis
The IEEE 33 node power system shown in fig. 7 was used as a simulation analysis model. The lines S1-S4 are respectively set to have faults at different fault types, and zero sequence current signals of all the lines in a fault area (such as a dotted line part in figure 6) are recorded.
4.1 Fault feature extraction
The invention sets 4 fault positions for an IEEE 33 node power system model, each fault position is simulated under the condition of different fault types, and the change condition of the zero sequence current of each line is recorded. Among them, the fault lines provided are S1, S2, S3 and S4, and the fault types are single-phase fault, two-phase fault and three-phase fault.
Taking the single-phase ground fault occurring in the line S1 as an example, the zero-sequence current waveform and VMD-decomposed IMF component of the line S1 are shown in fig. 8. As can be seen from the amplitude characteristic curve of the IMF component in fig. 9 (a), the IMF1 component has more fault feature information than other IMF components, so that fault feature extraction is performed on the IMF1 component. The frequency characteristics of IMF1 obtained by HHT are shown in fig. 9 (b). And taking the frequency characteristics of the IMF1 as fault characteristics, and extracting the fault characteristics of other lines in the same way to construct a fault characteristic vector.
4.2 CNN network architecture selection
And performing fault feature extraction on the screened IMF components by using HHT, constructing a fault feature vector to train CNN, and performing fault positioning and fault type judgment by using the trained CNN. The configuration of the fault status label is shown in table 1.
Table 1 fault status tag configuration table
Table 1 Fault status label configuration table
The setting of the parameters of the CNN model directly influences the accuracy and efficiency of the whole fault diagnosis. Therefore, five different CNN topological structures are designed, different parameters are selected, and the structural parameters are shown in Table 2.
TABLE 2 structural parameters of different CNN models
Table 2 Structural parameters of different CNN models
The CNN sets the maximum number of iterations to 20 and the learning rate to 0.1. Each topological structure is trained under the same training sample, the same test sample is used for testing, each structure is subjected to 50 times of repeated experimental simulation, and the average fault location accuracy is shown in table 3. It can be seen that although the model serial No. 1 has the shortest average time of only 46.419s, the fault location accuracy is very low, namely only 42.65%. Compared with the model with the sequence number 1, the model with the sequence number 5 realizes the improvement of the fault positioning accuracy rate by adding the convolution layer and the pooling layer, but the overall average training time is also obviously improved to 131.62 s. As can be seen from the model with the sequence number of 2-4, the average training time can be adjusted along with the adjustment of the size of the convolution kernel. By combining the comparison of the experimental results in the table 3, the model with the sequence number 3 is selected to ensure that the fault positioning accuracy and the training time reach relatively optimal results.
TABLE 3 Fault location accuracy for different CNN models
Table 3 Fault location accuracy of different CNN models
4.3 example simulation results
The invention selects the zero sequence current signal which is not subjected to any pretreatment to carry out CNN fault diagnosis (CNN) and the zero sequence current signal which is decomposed by EMD to carry out CNN fault diagnosis (EMD-CNN), and compares the two methods with the method (VMD-CNN) provided by the invention to verify the advantages of the method provided by the invention.
The CNN training is performed under different training sample numbers, and the accuracy of fault location and fault type determination of the same test sample is shown in fig. 10. In terms of fault location, after training of 200 groups of samples, the fault location accuracy of the VMD-CNN is higher than that of the CNN and the EMD-CNN. In addition, after the VMD-CNN is trained by 400 groups of training samples and 500 groups of training samples, the fault location accuracy can reach 100%, while the fault location accuracy of the CNN and the EMD-CNN is 83.33% and 79.83% respectively when the CNN and the EMD-CNN are trained under 500 groups of training samples, and is slightly insufficient compared with the VMD-CNN. In the aspect of fault type judgment, the VMD-CNN is trained under 500 groups of training samples, the test accuracy can also reach 100%, and the VMD-CNN is also better than the test accuracy results of the CNN and the EMD-CNN when trained under 200 groups of training samples and 400 groups of training samples.
A comparison of the fault location accuracy for different methods at different sample numbers is shown in fig. 11. In graph (a), after training under 200 training samples, the VMD-CNN has better fault location results than CNN and EMD-CNN, although the test accuracy of the method in FIG. 3 is good. As can be seen from the graph (b) in FIG. 11, the test accuracy of the VMD-CNN can reach 100% in 400 training samples. As can be seen from the graph (c) in FIG. 11, after 16 iterations, the VMD-CNN test accuracy reaches 100%, which is higher than the CNN and EMD-CNN fault location accuracy.
After 500 training samples are trained, the comparison results of the judgment accuracy rates of the fault types of different methods are shown in fig. 12. It can be seen that after 17 iterations of the VMD-CNN, the test accuracy can reach 100%, and the fault type judgment accuracy is higher than that of the other two methods. After training of 200 groups and 400 groups of training samples, the comparison results of the judgment accuracy rates of the fault types of different methods are shown in appendix A.
As can be seen from fig. 11 and 12, compared with the CNN and EMD-CNN methods, the VMD-CNN method proposed by the present invention has good performance in fault location and high accuracy in fault type judgment. In addition, the VMD-CNN has good fitting effect in fault location and fault type judgment, and the VMD-CNN is suitable for fault location and fault type judgment and has good robustness and good generalization capability.
Therefore, the embodiment of the invention provides a neural network power distribution network fault diagnosis method based on variational modal decomposition. Firstly, VMD decomposition is carried out on zero sequence current to obtain IMF components, then IMF components with abundant fault characteristic information are selected from the obtained IMF components, HHT is carried out to extract fault characteristics, and finally fault diagnosis is carried out by utilizing the fault characteristics and the trained CNN. The following conclusions are obtained through experimental simulation:
(1) the method provided by the invention not only can realize the fault location of the power distribution network, but also can realize the fault type judgment, and has higher diagnosis precision compared with other methods.
(2) Through the selection of the CNN model and the adjustment of model parameters, the fault diagnosis precision can be obviously improved, and the time consumption of fault diagnosis can be reduced.
(3) Compared with other methods, the method can effectively improve the fault precision and has good generalization capability.
Therefore, the neural network power distribution network fault diagnosis method based on variational modal decomposition can provide an effective means for realizing accurate and fast fault diagnosis of the power distribution network.
On the other hand, the invention also discloses a convolutional neural network power distribution network fault diagnosis system based on variational modal decomposition, which comprises the following units:
the IMF component extraction unit is used for extracting zero-sequence current signals of all lines in the power distribution network fault region and obtaining IMF components of the zero-sequence currents of the lines in the power distribution network fault region through variational modal decomposition;
the IMF component screening unit is used for screening IMF components with characteristic information reaching a specific quantity;
the fault feature extraction unit is used for performing Hilbert-Huang transform processing on the screened IMF components and extracting fault features;
and the fault diagnosis unit is used for taking the extracted fault characteristics as input signals and utilizing the trained CNN fault diagnosis model to carry out fault positioning and fault type judgment.
It is understood that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, the example and the beneficial effects of the related contents can refer to the corresponding parts in the method.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (6)
1. A neural network power distribution network fault diagnosis method based on variational modal decomposition is characterized in that: comprises the following steps of (a) carrying out,
s100, extracting zero sequence current signals of each line in a power distribution network fault area, and obtaining IMF components of the zero sequence current of the lines in the power distribution network fault area through variational modal decomposition;
s200, screening IMF components with characteristic information reaching a specific amount;
s300, performing Hilbert-Huang transform processing on the screened IMF components, and extracting fault features;
s400, the extracted fault characteristics are used as input signals, and fault positioning and fault type judgment are carried out by using the trained CNN fault diagnosis model.
2. The neural network power distribution network fault diagnosis method based on variational modal decomposition according to claim 1, characterized in that: the CNN fault diagnosis model in S400 includes:
the device consists of an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a convolution layer, a pooling layer and a full-connection layer;
where the convolutional layer is defined by a set of small filters through which the input data is convolved in the forward channel, the convolutional layer output is as follows:
wherein h isj(x) Is the spatial position x ═ x1,x2) The jth inactive output feature map of (g)ijIs hjAnd the ith input channel fiA core in between;
c represents the total input channel number, b represents the deviation, and the method has the capability of improving the nonlinear segmentation of the network;
the operator is a two-dimensional convolution defined as follows:
where m × n represents the kernel size;
the filter also comprises a pooling layer, wherein the pooling layer is directly inserted into the convolution layer, and then the output of the convolution layer is screened and filtered to remove non-important characteristic parameters, so that the calculation time can be effectively reduced;
the full-connection layer is a traditional multilayer perceptron which utilizes characteristics for classification;
the Softmax function acts as an activation function for the CNN fault diagnosis model.
3. The neural network power distribution network fault diagnosis method based on variational modal decomposition according to claim 1, characterized in that: the S100, obtaining the IMF component of the zero sequence current of the power distribution network fault area line through variational modal decomposition specifically comprises:
this is achieved by solving the following constrained optimization problem:
wherein u iskIs the k-th modal component, wkIs the center frequency of the kth mode, f (t) is the input signal, (t) is the dirac function;
introducing a penalty factor alpha and a Lagrange multiplier lambda, and processing the formula (3) to obtain:
obtaining the optimal solution of (4) through multiple iterations, wherein modal components and center frequencies in the iteration process are respectively expressed as:
wherein the content of the first and second substances,
are respectively as
f(t)，uk(t), λ (t) results obtained by fourier transform.
4. The neural network power distribution network fault diagnosis method based on variational modal decomposition according to claim 1, characterized in that: the step S200 of screening out an IMF component including feature information up to a specific amount specifically includes:
performing amplitude characteristic calculation on each IMF component after VMD decomposition by using HHT, and screening the IMF components;
at time t, the nth IMF component of the zero sequence current of the mth line is represented as IMFmn(t) the reconstructed signal P (t) obtained by HHT is
Wherein the instantaneous amplitude a (t) and phase angle
Respectively as follows:
calculating the instantaneous frequency w (t) as
The amplitude characteristic of each IMF component is obtained by equation (8).
5. The neural network power distribution network fault diagnosis method based on variational modal decomposition according to claim 1, characterized in that: the step S300 of performing hilbert-yellow transform processing on the screened IMF components, and extracting fault features specifically includes:
performing fault feature extraction on IMF components of which the screened fault feature information reaches a set value, and extracting frequency features of signals as fault features to perform fault positioning and fault type judgment;
the IMF2 component frequency characteristic is obtained by equation (10).
6. The utility model provides a convolutional neural network distribution network fault diagnosis system based on variational modal decomposition which characterized in that includes the following unit:
the IMF component extraction unit is used for extracting zero-sequence current signals of all lines in the power distribution network fault region and obtaining IMF components of the zero-sequence currents of the lines in the power distribution network fault region through variational modal decomposition;
the IMF component screening unit is used for screening IMF components with characteristic information reaching a specific quantity;
the fault feature extraction unit is used for performing Hilbert-Huang transform processing on the screened IMF components and extracting fault features;
and the fault diagnosis unit is used for taking the extracted fault characteristics as input signals and utilizing the trained CNN fault diagnosis model to carry out fault positioning and fault type judgment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010783864.0A CN112051480A (en) | 2020-08-06 | 2020-08-06 | Neural network power distribution network fault diagnosis method and system based on variational modal decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010783864.0A CN112051480A (en) | 2020-08-06 | 2020-08-06 | Neural network power distribution network fault diagnosis method and system based on variational modal decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112051480A true CN112051480A (en) | 2020-12-08 |
Family
ID=73602168
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010783864.0A Pending CN112051480A (en) | 2020-08-06 | 2020-08-06 | Neural network power distribution network fault diagnosis method and system based on variational modal decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112051480A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112903294A (en) * | 2021-01-07 | 2021-06-04 | 泰华宏业(天津)智能科技有限责任公司 | Rolling bearing fault diagnosis method based on VMD and deep convolution neural network |
| CN113049914A (en) * | 2021-03-15 | 2021-06-29 | 广东电网有限责任公司清远供电局 | Power transmission line fault diagnosis method and system, electronic equipment and storage medium |
| CN113138322A (en) * | 2021-05-11 | 2021-07-20 | 南方电网科学研究院有限责任公司 | Method, device, equipment and storage medium for detecting high-resistance earth fault of power distribution network |
| CN113325267A (en) * | 2021-05-20 | 2021-08-31 | 国网辽宁省电力有限公司 | Power distribution network fault diagnosis system and method and computer program product |
| CN113466624A (en) * | 2021-06-30 | 2021-10-01 | 云南电网有限责任公司电力科学研究院 | Method and system for detecting fault area of multi-terminal hybrid direct-current transmission line |
| CN113625164A (en) * | 2021-08-02 | 2021-11-09 | 南京航空航天大学 | Aviation generator fault feature extraction method, system, medium and computing device |
| CN114742182A (en) * | 2022-06-15 | 2022-07-12 | 深圳市明珞锋科技有限责任公司 | Intelligent equipment output data information processing method and operation evaluation method |
| CN115221769A (en) * | 2021-04-15 | 2022-10-21 | 广州中国科学院先进技术研究所 | Fault prediction method, system, electronic equipment and storage medium |
| CN115935244A (en) * | 2023-03-09 | 2023-04-07 | 西南交通大学 | Single-phase rectifier fault diagnosis method based on data driving |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103048137A (en) * | 2012-12-20 | 2013-04-17 | 北京航空航天大学 | Fault diagnosis method of rolling bearing under variable working conditions |
| CN107315892A (en) * | 2017-08-10 | 2017-11-03 | 北京交通大学 | A kind of Method for Bearing Fault Diagnosis based on extreme learning machine |
| CN110069886A (en) * | 2019-05-05 | 2019-07-30 | 四川大学 | The identification of cable initial failure and classification method based on VMD and CNN |
| CN110146281A (en) * | 2019-06-06 | 2019-08-20 | 南京航空航天大学 | A kind of epicyclic gearbox method for diagnosing faults based on VMD-SDAE |
| CN110222887A (en) * | 2019-05-27 | 2019-09-10 | 国网安徽省电力有限公司 | Prediction technique based on VMD and DNN and the application in short-term load forecasting |
| CN110907177A (en) * | 2019-12-10 | 2020-03-24 | 北京交通大学 | Bearing fault diagnosis method based on layered extreme learning machine |
| CN111222458A (en) * | 2020-01-06 | 2020-06-02 | 浙江工业大学 | Rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and convolutional neural network |
| CN111474442A (en) * | 2020-03-30 | 2020-07-31 | 国网山东省电力公司德州供电公司 | Power distribution network single-phase earth fault positioning method and system |
-
2020
- 2020-08-06 CN CN202010783864.0A patent/CN112051480A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103048137A (en) * | 2012-12-20 | 2013-04-17 | 北京航空航天大学 | Fault diagnosis method of rolling bearing under variable working conditions |
| CN107315892A (en) * | 2017-08-10 | 2017-11-03 | 北京交通大学 | A kind of Method for Bearing Fault Diagnosis based on extreme learning machine |
| CN110069886A (en) * | 2019-05-05 | 2019-07-30 | 四川大学 | The identification of cable initial failure and classification method based on VMD and CNN |
| CN110222887A (en) * | 2019-05-27 | 2019-09-10 | 国网安徽省电力有限公司 | Prediction technique based on VMD and DNN and the application in short-term load forecasting |
| CN110146281A (en) * | 2019-06-06 | 2019-08-20 | 南京航空航天大学 | A kind of epicyclic gearbox method for diagnosing faults based on VMD-SDAE |
| CN110907177A (en) * | 2019-12-10 | 2020-03-24 | 北京交通大学 | Bearing fault diagnosis method based on layered extreme learning machine |
| CN111222458A (en) * | 2020-01-06 | 2020-06-02 | 浙江工业大学 | Rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and convolutional neural network |
| CN111474442A (en) * | 2020-03-30 | 2020-07-31 | 国网山东省电力公司德州供电公司 | Power distribution network single-phase earth fault positioning method and system |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112903294A (en) * | 2021-01-07 | 2021-06-04 | 泰华宏业(天津)智能科技有限责任公司 | Rolling bearing fault diagnosis method based on VMD and deep convolution neural network |
| CN113049914A (en) * | 2021-03-15 | 2021-06-29 | 广东电网有限责任公司清远供电局 | Power transmission line fault diagnosis method and system, electronic equipment and storage medium |
| CN115221769A (en) * | 2021-04-15 | 2022-10-21 | 广州中国科学院先进技术研究所 | Fault prediction method, system, electronic equipment and storage medium |
| CN115221769B (en) * | 2021-04-15 | 2023-10-10 | 广州中国科学院先进技术研究所 | Fault prediction method, system, electronic equipment and storage medium |
| CN113138322A (en) * | 2021-05-11 | 2021-07-20 | 南方电网科学研究院有限责任公司 | Method, device, equipment and storage medium for detecting high-resistance earth fault of power distribution network |
| CN113325267A (en) * | 2021-05-20 | 2021-08-31 | 国网辽宁省电力有限公司 | Power distribution network fault diagnosis system and method and computer program product |
| CN113466624A (en) * | 2021-06-30 | 2021-10-01 | 云南电网有限责任公司电力科学研究院 | Method and system for detecting fault area of multi-terminal hybrid direct-current transmission line |
| CN113625164A (en) * | 2021-08-02 | 2021-11-09 | 南京航空航天大学 | Aviation generator fault feature extraction method, system, medium and computing device |
| CN114742182A (en) * | 2022-06-15 | 2022-07-12 | 深圳市明珞锋科技有限责任公司 | Intelligent equipment output data information processing method and operation evaluation method |
| CN115935244A (en) * | 2023-03-09 | 2023-04-07 | 西南交通大学 | Single-phase rectifier fault diagnosis method based on data driving |
| CN115935244B (en) * | 2023-03-09 | 2023-05-09 | 西南交通大学 | Single-phase rectifier fault diagnosis method based on data driving |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112051480A (en) | Neural network power distribution network fault diagnosis method and system based on variational modal decomposition | |
| CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
| CN109614981B (en) | Power system intelligent fault detection method and system based on spearman level-dependent convolutional neural network | |
| Pan et al. | An approach for HVCB mechanical fault diagnosis based on a deep belief network and a transfer learning strategy | |
| CN113177357B (en) | Transient stability assessment method for power system | |
| CN109886464B (en) | Low-information-loss short-term wind speed prediction method based on optimized singular value decomposition generated feature set | |
| Kurbatskii et al. | On the neural network approach for forecasting of nonstationary time series on the basis of the Hilbert-Huang transform | |
| CN115081316A (en) | DC/DC converter fault diagnosis method and system based on improved sparrow search algorithm | |
| CN110672905A (en) | CNN-based self-supervision voltage sag source identification method | |
| CN115238980A (en) | Micro-grid transient power angle trajectory prediction method based on attention mechanism and space-time diagram convolutional neural network | |
| Baumann et al. | Impulse test fault diagnosis on power transformers using Kohonen's self-organizing neural network | |
| CN114355240A (en) | Power distribution network ground fault diagnosis method and device | |
| CN114966312A (en) | Power distribution network fault detection and positioning method and system based on migration convolutional neural network | |
| CN112085111A (en) | Load identification method and device | |
| CN115586444A (en) | Lithium battery residual life prediction method based on VMD and BP neural network | |
| CN109903181B (en) | Line loss prediction method under missing data set based on compressed sensing | |
| Wang et al. | An Automatic Identification Framework for Complex Power Quality Disturbances Based on Ensemble CNN | |
| CN116400168A (en) | Power grid fault diagnosis method and system based on depth feature clustering | |
| Li et al. | Artificial intelligence for real-time topology identification in power distribution systems | |
| CN114298413A (en) | Hydroelectric generating set runout trend prediction method | |
| CN114062832A (en) | Method and system for identifying short-circuit fault type of power distribution network | |
| CN114254828A (en) | Power load prediction method based on hybrid convolution feature extractor and GRU | |
| CN113627539A (en) | Rolling bearing composite fault diagnosis method and system, storage medium and computing equipment | |
| CN112308208A (en) | Transformer fault diagnosis method based on deep learning model | |
| Long et al. | Recognition of Internal Overvoltage in Distribution Network Based on Convolutional Neural Network. |
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: 20201208 |
|
| RJ01 | Rejection of invention patent application after publication |