CN117746095A - Overhead transmission line fault and interference identification and classification system and method - Google Patents
Overhead transmission line fault and interference identification and classification system and method Download PDFInfo
- Publication number
- CN117746095A CN117746095A CN202311516531.1A CN202311516531A CN117746095A CN 117746095 A CN117746095 A CN 117746095A CN 202311516531 A CN202311516531 A CN 202311516531A CN 117746095 A CN117746095 A CN 117746095A
- Authority
- CN
- China
- Prior art keywords
- fault
- dimensional
- transmission line
- matrix
- classified
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005540 biological transmission Effects 0.000 title claims abstract description 74
- 238000000034 method Methods 0.000 title claims description 34
- 239000011159 matrix material Substances 0.000 claims abstract description 110
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 43
- 239000013598 vector Substances 0.000 claims abstract description 29
- 238000007637 random forest analysis Methods 0.000 claims abstract description 23
- 238000011156 evaluation Methods 0.000 claims abstract description 21
- 238000010606 normalization Methods 0.000 claims abstract description 20
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000006243 chemical reaction Methods 0.000 claims abstract description 11
- 238000003066 decision tree Methods 0.000 claims description 22
- 230000009466 transformation Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000012512 characterization method Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses an overhead transmission line fault and interference identification classification system, wherein a sample matrix acquisition module of the system is used for determining an initial sample matrix of transmission line fault traveling waves and interference clutter; the normalization and vector conversion module is used for carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix, converting a one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix, and forming a two-dimensional traveling wave gray level image; the convolutional neural network module is used for extracting the two-dimensional travelling wave gray level image through the convolutional neural network characteristics to obtain a fault characteristic sequence; the fault characteristic classification and evaluation module is used for classifying the power transmission line fault traveling wave and interference clutter data which need to be identified and classified by utilizing a random forest algorithm, and calculating an evaluation index of a classification result. The invention can effectively identify and classify the fault waveform and the interference clutter of the power transmission line.
Description
Technical Field
The invention relates to the technical field of overhead transmission line fault identification, in particular to an overhead transmission line fault and interference identification classification system and method.
Background
In recent decades, with the continued innovation of high and new technologies, traditional power systems are transitioning to intelligent power systems with advanced monitoring and control means. In addition, the demand of users for obtaining safe, sustainable and high-quality power is rising, eventually increasing the number of transmission lines, greatly increasing the complexity of the power system. The performance of the power transmission line plays an important role in continuous power supply, so the research on fault detection and classification technology of the power transmission line is particularly important, the current fault detection and classification technology is generally divided into two main types, namely a conventional technology and a machine learning method, the conventional fault type identification is realized by measuring the change condition of each conventional phase of electric quantity and then performing rapid protection and steady-state protection, but the conventional fault type identification is easily influenced by the current operation mode of the system, such as fault position points, different factors of transition resistance and the like, so that the deviation of fault identification results is caused; with the development and maturity of artificial intelligence, more intelligent technologies are applied to the field of power systems, and the current identification process of the fault type of the power transmission line mainly comprises two steps: the first step is to manually design and extract fault characteristics from the electrical quantity change when the line fails; the second step is to take the extracted feature as the input of the fault identification model, and to use the trained model to finish the classification of fault types, because the process of extracting the fault feature vector is complex, a strong research experience is needed, and the method for manually extracting the feature vector has certain limitation and subjectivity, the extracted feature vector is difficult to adapt to various easily-confused samples.
Disclosure of Invention
The invention aims to provide a method for identifying and classifying faults and interference of an overhead transmission line, which can be used for quickly and effectively identifying and classifying fault waveforms and interference clutter of the transmission line.
The invention provides an overhead transmission line fault and interference identification and classification system, which comprises a sample matrix acquisition module, a normalization and vector conversion module, a convolutional neural network module and a fault feature classification and evaluation module;
the sample matrix acquisition module is used for acquiring a historical time sequence of the power transmission line fault traveling wave and the interference clutter, obtaining a line mode component of an initial time domain waveform by using phase mode transformation, and determining an initial sample matrix of the power transmission line fault traveling wave and the interference clutter;
the normalization and vector conversion module is used for carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix, converting a one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix, and forming a two-dimensional traveling wave gray level image;
the convolutional neural network module is used for extracting the two-dimensional travelling wave gray level image through the convolutional neural network characteristics to obtain a fault characteristic sequence;
the fault characteristic classification and evaluation module is used for classifying the power transmission line fault traveling wave and interference clutter data which need to be identified and classified by utilizing a random forest algorithm, and calculating an evaluation index of a classification result.
The invention has the beneficial effects that:
the invention uses convolutional neural network as the feature extractor of traveling wave data, and uses random forest algorithm to classify faults and interferences. Firstly, extracting linear mode components of initial fault traveling wave and interference clutter time sequence as initial traveling wave data vectors, mapping one-dimensional traveling wave data into a two-dimensional matrix through a gram matrix, and representing the one-dimensional traveling wave data by using a gray image; secondly, constructing a two-dimensional convolutional neural model based on traveling wave data, realizing self-learning of traveling wave data characteristics, obtaining a waveform characteristic sequence of the traveling wave data, and then adopting a random forest algorithm to realize automatic identification and screening of fault traveling waves and interference clutter. The method can effectively identify and classify the fault waveform and the interference clutter of the power transmission line, and ensures the accuracy and the efficiency of the fault detection of the power transmission line.
Drawings
FIG. 1 is a schematic diagram of the structure of the present invention;
FIG. 2 is a flow chart of the present invention;
FIG. 3 is a diagram of a classification model according to the present invention;
Detailed Description
The invention is described in further detail below with reference to the attached drawings and specific examples:
as shown in fig. 1, the overhead transmission line fault and interference identification and classification system is characterized in that: the device comprises a sample matrix acquisition module, a normalization and vector conversion module, a convolutional neural network module and a fault feature classification and evaluation module;
the sample matrix acquisition module is used for acquiring transmission line fault traveling waves (the fault traveling waves are guided electromagnetic waves which are caused by faults and propagate along the transmission line) and interference clutter historical time sequence, obtaining line mode components of an initial time domain waveform by using phase mode transformation, and determining transmission line fault traveling waves and interference clutter initial sample matrices; when a general fault occurs, besides fault phases, the non-fault phases also have traveling wave waveforms due to factors such as mutual inductance and electromagnetic coupling among wires, and three-phase fault traveling waves contain rich fault information, and because of coupling of traveling wave waveforms of each phase in a three-phase system, three-phase phasors are converted into mutually independent moduli through phase-mode transformation in order to acquire more complete fault information characteristics; the modal component is divided into a linear mode component 1, a 2-mode component or an alpha-beta-mode component and a zero-mode component, the wave speed of the linear mode component is close to the light speed, the wave impedance is smaller than that of the zero-mode component, and the wave speed is not easily influenced by external factors in the propagation process, so that the linear mode component is generally selected to be extracted for analysis such as fault identification and positioning; the design is equivalent to the preprocessing of data, namely, the original sample containing fault information characteristics is obtained by processing each phase of fault traveling wave signals and clutter signals which are initially obtained from the traveling wave acquisition device, and the characteristics are extracted, identified and classified by a follow-up reuse algorithm;
the normalization and vector conversion module is used for carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix, converting a one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix, and forming a two-dimensional traveling wave gray level image; two-dimensional Convolutional Neural Networks (CNNs) have been widely used in image classification processing, but sample data are one-dimensional time sequence, and to extract features by using the two-dimensional convolutional neural networks, an initial sample must be subjected to two-dimensional processing, wherein a Gram Angle Field (GAF) mapping method is used to convert each one-dimensional sample data into a two-dimensional matrix, and then the two-dimensional matrix is input into the CNNs;
the convolutional neural network module is used for extracting the two-dimensional travelling wave gray level image through the convolutional neural network characteristics to obtain a fault characteristic sequence;
the fault characteristic classification and evaluation module is used for classifying the power transmission line fault traveling wave and interference clutter data which need to be identified and classified by utilizing a random forest algorithm, and calculating an evaluation index of a classification result.
In the technical scheme, the power transmission line faults mainly comprise single-phase, two-phase and three-phase grounding short-circuit faults and lightning flashover faults. The reason for the generation of the interference clutter is numerous, and many factors such as corona discharge of a high-voltage transmission line, power electronic equipment switching of a transformer substation, and actions of a protection device can cause the high-frequency traveling wave acquisition device to acquire a large amount of interference clutter.
The fault traveling wave is a transient traveling wave waveform generated at a fault point of the power transmission line and transmitted to two sides of the line, and has current and voltage traveling waves, and the current traveling wave signal is generally collected by a collecting device to perform fault analysis at present.
In the above technical solution, the specific method for determining the initial sample matrix of the transmission line fault traveling wave and the interference clutter by the sample matrix acquisition module is as follows:
extracting line-mode component characterization travelling wave information of a time domain waveform from one-dimensional time sequence signals of a power transmission line fault travelling wave and interference clutter historical time sequence through phase-mode transformation, wherein the phase-mode transformation calculation method comprises the following steps:
wherein X is 0 m 、X 1 m 、X 2 m The zero-mode component, the 1-mode component and the 2-mode component of the initial time domain waveform are respectively; x is X a 、X b 、X c The time sequences of fault traveling waves and interference signals of a phase, a b phase and a c phase of the power transmission line are respectively;
extracting 1-mode components of an initial time domain waveform, and constructing an initial sample matrix of the fault traveling wave and the interference clutter of the power transmission line with the size of KxM;
the initial sample matrix X of the transmission line fault traveling wave and the interference clutter is as follows:
m is the length of each time sequence sample extracted by phase-mode transformation of the power transmission line fault traveling wave and interference clutter history time sequence, and X is the length of each time sequence sample extracted by phase-mode transformation 1 The first time series sample, X, representing the initial sample matrix K The kth sequential sequence sample, x, representing the initial sample matrix 11 A first element, x, representing a first time series sample of the initial sample matrix KM Represents the mth element of the kth sequential sequence sample of the initial sample matrix.
The design preprocesses the initial traveling wave data, extracts a one-dimensional time sequence (namely, extracts linear mode components) containing fault information characteristics, and uses the one-dimensional time sequence as an initial sample matrix of a CNN algorithm.
In the above technical solution, the specific method for the normalization and vector conversion module to normalize the transmission line fault traveling wave and interference clutter initial sample matrix includes:
normalizing each one-dimensional time sequence of the transmission line fault traveling wave and interference clutter initial sample matrix, wherein the normalization method comprises the following steps:
wherein x 'is' ij Normalized value of j data for the i one-dimensional time series of the initial sample matrix, i=1, 2,..k; j=1, 2,. -%, M; x is x ij An initial value of the j-th data of the i-th one-dimensional time sequence of the initial sample matrix; max (X) i ) The maximum value of the ith one-dimensional time sequence of the initial sample matrix; min (X) i ) The minimum value of the ith one-dimensional time sequence of the initial sample matrix;
the normalized sample matrix X' is:
wherein X 'is' K Is X K Normalized value of x ', x' KM Is x KM Is included in the above formula (c).
In the above technical solution, the specific method for forming the two-dimensional traveling wave gray scale image by converting the one-dimensional vector of the normalized initial sample matrix into the two-dimensional matrix includes:
converting the one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix by using a gram matrix method, and obtaining a two-dimensional travelling wave gray image matrix of each group of sample data of the sample matrix;
and performing polar coordinate transformation on the normalized one-dimensional time sequence by using a gram matrix method, wherein the expression is as follows:
θ i,j =arccos x′ ij
i=1,2,...,K;j=1,2,...,M
in θ i.j The polar angle corresponding to the j-th data of the normalized i-th one-dimensional time sequence; x's' ij Normalized values of the j-th data of the i-th one-dimensional time sequence of the initial sample matrix; r is (r) j The corresponding polar diameter of the j-th data of the normalized i-th one-dimensional time sequence; k is the total number of sample data; m is the sample data length;
the two-dimensional traveling wave gray scale image matrix of each group of sample data of the sample matrix obtained by using the gram matrix method is as follows:
wherein G is i The two-dimensional gray image matrix G is the two-dimensional gray image matrix corresponding to the ith one-dimensional time sequence i There are K (K one-dimensional time sequence samples in the previous initial sample matrix, each of which is converted into a two-dimensional gray image matrix by normalized and reused glamer angle field (matrix) mapping), and the sample matrix is converted into K two-dimensional gray image matrices, which are used as inputs of the convolutional neural network. In order to perform feature extraction by using CNN, input data of the CNN algorithm must be two-dimensional data, so that all K one-dimensional time sequence samples in the initial sample matrix extracted previously are converted into K two-dimensional matrices for subsequent feature extraction of the CNN algorithm.
In the above technical solution, the specific method for extracting the fault feature sequence from the two-dimensional traveling wave gray level image by the convolutional neural network module through the convolutional neural network feature comprises the following steps:
as shown in fig. 3, K two-dimensional gray-scale image matrices G i Input into the input layer of the convolutional neural network, the convolutional layer passes through the convolutional kernel and K two-dimensional gray image matrixes G i Each region is convolved to extract a feature matrix, and a pooling layer reduces a two-dimensional gray image matrix G in a maximum pooling mode i The dimension, the full connection layer converts the feature matrix output by the convolution and pooling alternating layer into a low-dimension feature vector through full connection operation;
the input K two-dimensional gray-scale image matrixes G i Feature extraction through convolutional neural networkK fault characteristic sequences with the length of L are obtained and expressed as:
wherein TT is a fault feature matrix extracted by a convolutional neural network (each two-dimensional gray image matrix finally obtains a feature sequence with the length of L through each layer structure of CNN, namely, the feature sequence is output through a convolutional layer-pooling layer-full-connection layer); TT (TT) pq Q-th fault signature, p=1, 2, for the p-th fault signature sequence; q=1, 2,..l.
In the technical scheme, the fault feature classification and evaluation module classifies the fault feature sequences by using a random forest algorithm, K feature sequences are obtained by data preprocessing and CNN algorithm, and are used as a training database of the random forest algorithm to train a random forest model, so that the random forest model can identify fault traveling waves and clutters; the trained random forest model is utilized to classify faults and clutter waveforms, and each evaluation index is calculated to evaluate classification results, and the specific method comprises the following steps:
taking a fault feature matrix TT extracted by a convolutional neural network as a training data set, randomly extracting K sample data from K sample data of the training data set TT in a put-back way, randomly selecting Q feature quantities from L feature quantities of each sample data as the candidate splitting conditions of the internal nodes of a sample subset decision tree, and constructing the sample subset decision tree; repeating the above operation to form N sample subsets TT 1 、TT 2 、…、TT i 、…、TT N Expressed as:
TT i =[r(TT)] K×Q
Q<<L
in TT i For the selected i-th sample subset; r (TT) is a submatrix formed by randomly extracting K fault feature sequences from a training data set TT with a place back and randomly selecting Q feature quantities in each fault feature sequence;
subset T of N samplesT 1 、TT 2 、…、TT i 、…、TT N Performing random forest training to generate N decision trees of a random forest;
carrying out phase-mode transformation on the power transmission line fault traveling wave and interference clutter data needing to be identified and classified to obtain power transmission line fault traveling wave and interference clutter sample matrixes needing to be identified and classified, carrying out normalization and vector conversion processing on the sample matrixes needing to be identified and classified to obtain two-dimensional traveling wave gray level images of the sample matrixes needing to be identified and classified, and carrying out convolutional neural network feature extraction on the two-dimensional traveling wave gray level images of the sample matrixes needing to be identified and classified to obtain fault feature sequences needing to be identified and classified; and putting the fault feature sequences needing to be identified and classified into N decision trees of a random forest, wherein each decision tree is used as a decision unit to vote on the fault feature sequences needing to be identified and classified, and the class with the largest vote number is the prediction class result of the fault feature sequences needing to be identified and classified.
In the technical scheme, the specific method for generating N decision trees comprises the following steps: and respectively calculating the coefficient gains of Q characteristics of the decision tree nodes, determining the characteristic corresponding to the maximum coefficient gain as the optimal splitting condition of the decision tree nodes, and splitting the nodes to form N decision trees. The random forest training is implemented by extracting training samples from a training database obtained by CNN for multiple times in a random and put-back mode, so that a plurality of training sample subsets are formed, each decision tree is trained by utilizing each sample subset, each decision tree represents a classifier, the subsequent input waveform is classified through each decision tree, and the most classification result of all decision trees is used as the classification result of a final algorithm. To judge whether a waveform is a fault traveling wave or clutter, whether the waveform is a fault is judged by the characteristics of a characteristic sequence (each element of the characteristic sequence represents a characteristic), the invention determines which characteristic is selected to judge classification by calculating the coefficient gain of the characteristic, and continues splitting downwards and then selects another characteristic by calculating the coefficient of the characteristic until the downwards selected characteristic is repeated before, so that a characteristic classification tree is obtained, and the input waveform is judged and classified by the characteristic classification tree.
In the above technical solution, the specific method for calculating the evaluation index of the classification result is:
the classification result evaluation indexes comprise an accuracy index AC, an accuracy index PC, a recall index RC and a specificity index SC;
the accuracy index AC is the proportion of the correct sample number of the classification result to the total sample number, and the calculation formula is as follows:
in the formula, TP is the number of samples which are classified as fault traveling waves and are correctly classified; FP is the number of samples classified as traveling fault but misclassified; TN is the number of samples classified as interference clutter and correctly classified; FN is the number of samples classified as interference clutter but misclassified;
the accuracy index PC is the proportion of the number of correctly classified samples in the number of samples with the classification result of fault traveling waves, and the calculation formula is as follows:
in the formula, TP is the number of samples which are classified as fault traveling waves and are correctly classified; FP is the number of samples classified as traveling fault but misclassified;
the recall index RC is the proportion of the number of samples classified as fault traveling waves and with correct results to the number of all fault samples, and the calculation formula is as follows:
in the formula, TP is the number of samples which are classified as fault traveling waves and are correctly classified; FN is the number of samples classified as interference clutter but misclassified;
the specificity index SC is the proportion of the number of samples classified as interference clutter and having correct results to the number of samples of all interference clutter, and the calculation formula is as follows:
in the formula, TN is the number of samples which are classified as interference clutter and are correctly classified; FP is the number of samples classified as traveling fault but misclassified.
The evaluation and classification result is more visual and reliable, and is helpful for accurate identification and classification of faults and interferences of the power transmission line.
An overhead transmission line fault and interference identification and classification method, as shown in figure 2, comprises the following steps:
step 1: the obtained historical time sequence of the power transmission line fault traveling wave and the interference clutter is transformed by using a phase mode to obtain a line mode component of an initial time domain waveform, and an initial sample matrix of the power transmission line fault traveling wave and the interference clutter is determined;
step 2: carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix, and converting a one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix to form a two-dimensional traveling wave gray level image;
step 3: extracting the two-dimensional traveling wave gray level image through a convolutional neural network characteristic to obtain a fault characteristic sequence;
step 4: and classifying the power transmission line fault traveling wave and interference clutter data which need to be identified and classified by using a random forest algorithm, and calculating an evaluation index of a classification result.
A computer-readable storage medium storing a computer program, characterized in that: the computer program, when being executed by a processor, implements the steps of the method as described above.
According to the invention, the convolutional neural network (convolutional neural network, CNN) is used as a feature extractor of traveling wave data, features can be extracted from the data autonomously through multi-stage alternating convolutional layer and pooling layer operations, the generalization performance of a model can be reduced by considering partial redundant features in feature vectors extracted by the convolutional neural network, the feature vectors output by the full-connection layer are transmitted as new training data sets to a random forest algorithm for learning and classifying, the training speed of the random forest algorithm is high, the classification accuracy is high, the features do not need to be selected manually, the generalization capability is better, and the recognition precision and efficiency can be effectively improved.
What is not described in detail in this specification is prior art known to those skilled in the art.
Claims (10)
1. The utility model provides an overhead transmission line trouble and interference discernment classification system which characterized in that: the device comprises a sample matrix acquisition module, a normalization and vector conversion module, a convolutional neural network module and a fault feature classification and evaluation module;
the sample matrix acquisition module is used for acquiring a historical time sequence of the power transmission line fault traveling wave and the interference clutter, obtaining a line mode component of an initial time domain waveform by using phase mode transformation, and determining an initial sample matrix of the power transmission line fault traveling wave and the interference clutter;
the normalization and vector conversion module is used for carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix, converting a one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix, and forming a two-dimensional traveling wave gray level image;
the convolutional neural network module is used for extracting the two-dimensional travelling wave gray level image through the convolutional neural network characteristics to obtain a fault characteristic sequence;
the fault characteristic classification and evaluation module is used for classifying the power transmission line fault traveling wave and interference clutter data which need to be identified and classified by utilizing a random forest algorithm, and calculating an evaluation index of a classification result.
2. The overhead transmission line fault and interference identification classification system of claim 1, wherein: the specific method for determining the initial sample matrix of the power transmission line fault traveling wave and the interference clutter by the sample matrix acquisition module comprises the following steps:
extracting line-mode component characterization travelling wave information of a time domain waveform from one-dimensional time sequence signals of a power transmission line fault travelling wave and interference clutter historical time sequence through phase-mode transformation, wherein the phase-mode transformation calculation method comprises the following steps:
wherein X is 0 m 、X 1 m 、X 2 m The zero-mode component, the 1-mode component and the 2-mode component of the initial time domain waveform are respectively; x is X a 、X b 、X c The time sequences of fault traveling waves and interference signals of a phase, a b phase and a c phase of the power transmission line are respectively;
extracting 1-mode components of an initial time domain waveform, and constructing an initial sample matrix of the fault traveling wave and the interference clutter of the power transmission line with the size of KxM;
the initial sample matrix X of the transmission line fault traveling wave and the interference clutter is as follows:
m is the length of each time sequence sample extracted by phase-mode transformation of the power transmission line fault traveling wave and interference clutter history time sequence, and X is the length of each time sequence sample extracted by phase-mode transformation 1 The first time series sample, X, representing the initial sample matrix K The kth sequential sequence sample, x, representing the initial sample matrix 11 A first element, x, representing a first time series sample of the initial sample matrix KM Represents the mth element of the kth sequential sequence sample of the initial sample matrix.
3. The overhead transmission line fault and interference identification classification system of claim 2, wherein: the specific method for carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix by the normalization and vector conversion module comprises the following steps:
normalizing each one-dimensional time sequence of the transmission line fault traveling wave and interference clutter initial sample matrix, wherein the normalization method comprises the following steps:
wherein x 'is' ij Normalized value of j data for the i one-dimensional time series of the initial sample matrix, i=1, 2,..k; j=1, 2,. -%, M; x is x ij An initial value of the j-th data of the i-th one-dimensional time sequence of the initial sample matrix; max (X) i ) The maximum value of the ith one-dimensional time sequence of the initial sample matrix; min (X) i ) The minimum value of the ith one-dimensional time sequence of the initial sample matrix;
the normalized sample matrix X' is:
wherein X 'is' K Is X K Normalized value of x ', x' KM Is x KM Is included in the above formula (c).
4. The overhead transmission line fault and interference identification classification system of claim 3, wherein: the specific method for forming the two-dimensional traveling wave gray level image by converting the one-dimensional vector of the normalized initial sample matrix into the two-dimensional matrix comprises the following steps:
converting the one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix by using a gram matrix method, and obtaining a two-dimensional travelling wave gray image matrix of each group of sample data of the sample matrix;
and performing polar coordinate transformation on the normalized one-dimensional time sequence by using a gram matrix method, wherein the expression is as follows:
θ i,j =arccos x′ ij
i=1,2,...,K;j=1,2,...,M
in θ i.j The polar angle corresponding to the j-th data of the normalized i-th one-dimensional time sequence; x's' ij Normalized values of the j-th data of the i-th one-dimensional time sequence of the initial sample matrix; r is (r) j The corresponding polar diameter of the j-th data of the normalized i-th one-dimensional time sequence; k is the total number of sample data; m is the sample data length;
the two-dimensional traveling wave gray scale image matrix of each group of sample data of the sample matrix obtained by using the gram matrix method is as follows:
wherein G is i The two-dimensional gray image matrix G is the two-dimensional gray image matrix corresponding to the ith one-dimensional time sequence i There are K.
5. The overhead transmission line fault and interference identification classification system of claim 4, wherein: the specific method for extracting the fault characteristic sequence from the two-dimensional traveling wave gray level image by the convolutional neural network module through the convolutional neural network characteristics comprises the following steps:
matrix G of K two-dimensional gray-scale images i Input into the input layer of the convolutional neural network, the convolutional layer passes through the convolutional kernel and K two-dimensional gray image matrixes G i Each region is convolved to extract a feature matrix, and a pooling layer reduces a two-dimensional gray image matrix G in a maximum pooling mode i The dimension, the full connection layer converts the feature matrix output by the convolution and pooling alternating layer into a low-dimension feature vector through full connection operation;
the input K two-dimensional gray-scale image matrixes G i K fault feature sequences with the length of L are obtained through convolutional neural network feature extraction, and are expressed as follows:
wherein TT is a fault feature matrix extracted by a convolutional neural network; TT (TT) pq Q-th fault signature, p=1, 2, for the p-th fault signature sequence; q=1, 2,..l.
6. The overhead transmission line fault and interference identification classification system of claim 5, wherein: the specific method for classifying the fault characteristic sequence by the fault characteristic classification and evaluation module through the random forest algorithm comprises the following steps:
taking a fault feature matrix TT extracted by a convolutional neural network as a training data set, randomly extracting K sample data from K sample data of the training data set TT in a put-back way, randomly selecting Q feature quantities from L feature quantities of each sample data as the candidate splitting conditions of the internal nodes of a sample subset decision tree, and constructing the sample subset decision tree; repeating the above operation to form N sample subsets TT 1 、TT 2 、…、TT i 、…、TT N Expressed as:
TT i =[r(TT)] K×Q
Q<<L
in TT i For the selected i-th sample subset; r (TT) is a submatrix formed by randomly extracting K fault feature sequences from a training data set TT with a place back and randomly selecting Q feature quantities in each fault feature sequence;
subset TT of N samples 1 、TT 2 、…、TT i 、…、TT N Performing random forest training to generate N decision trees of a random forest;
carrying out phase-mode transformation on the power transmission line fault traveling wave and interference clutter data needing to be identified and classified to obtain power transmission line fault traveling wave and interference clutter sample matrixes needing to be identified and classified, carrying out normalization and vector conversion processing on the sample matrixes needing to be identified and classified to obtain two-dimensional traveling wave gray level images of the sample matrixes needing to be identified and classified, and carrying out convolutional neural network feature extraction on the two-dimensional traveling wave gray level images of the sample matrixes needing to be identified and classified to obtain fault feature sequences needing to be identified and classified; and putting the fault feature sequences needing to be identified and classified into N decision trees of a random forest, wherein each decision tree is used as a decision unit to vote on the fault feature sequences needing to be identified and classified, and the class with the largest vote number is the prediction class result of the fault feature sequences needing to be identified and classified.
7. The overhead transmission line fault and interference identification classification system of claim 6, wherein: the specific method for generating N decision trees comprises the following steps: and respectively calculating the coefficient gains of Q characteristics of the decision tree nodes, determining the characteristic corresponding to the maximum coefficient gain as the optimal splitting condition of the decision tree nodes, and splitting the nodes to form N decision trees.
8. The overhead transmission line fault and interference identification classification system of claim 6, wherein: the specific method for calculating the evaluation index of the classification result comprises the following steps:
the classification result evaluation indexes comprise an accuracy index AC, an accuracy index PC, a recall index RC and a specificity index SC;
the accuracy index AC is the proportion of the correct sample number of the classification result to the total sample number, and the calculation formula is as follows:
in the formula, TP is the number of samples which are classified as fault traveling waves and are correctly classified; FP is the number of samples classified as traveling fault but misclassified; TN is the number of samples classified as interference clutter and correctly classified; FN is the number of samples classified as interference clutter but misclassified;
the accuracy index PC is the proportion of the number of correctly classified samples in the number of samples with the classification result of fault traveling waves, and the calculation formula is as follows:
in the formula, TP is the number of samples which are classified as fault traveling waves and are correctly classified; FP is the number of samples classified as traveling fault but misclassified;
the recall index RC is the proportion of the number of samples classified as fault traveling waves and with correct results to the number of all fault samples, and the calculation formula is as follows:
in the formula, TP is the number of samples which are classified as fault traveling waves and are correctly classified; FN is the number of samples classified as interference clutter but misclassified;
the specificity index SC is the proportion of the number of samples classified as interference clutter and having correct results to the number of samples of all interference clutter, and the calculation formula is as follows:
in the formula, TN is the number of samples which are classified as interference clutter and are correctly classified; FP is the number of samples classified as traveling fault but misclassified.
9. The overhead transmission line fault and interference identification and classification method is characterized by comprising the following steps of:
step 1: the obtained historical time sequence of the power transmission line fault traveling wave and the interference clutter is transformed by using a phase mode to obtain a line mode component of an initial time domain waveform, and an initial sample matrix of the power transmission line fault traveling wave and the interference clutter is determined;
step 2: carrying out normalization processing on the transmission line fault traveling wave and interference clutter initial sample matrix, and converting a one-dimensional vector of the normalized initial sample matrix into a two-dimensional matrix to form a two-dimensional traveling wave gray level image;
step 3: extracting the two-dimensional traveling wave gray level image through a convolutional neural network characteristic to obtain a fault characteristic sequence;
step 4: and classifying the power transmission line fault traveling wave and interference clutter data which need to be identified and classified by using a random forest algorithm, and calculating an evaluation index of a classification result.
10. A computer-readable storage medium storing a computer program, characterized in that: which computer program, when being executed by a processor, carries out the steps of the method according to claim 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311516531.1A CN117746095A (en) | 2023-11-14 | 2023-11-14 | Overhead transmission line fault and interference identification and classification system and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311516531.1A CN117746095A (en) | 2023-11-14 | 2023-11-14 | Overhead transmission line fault and interference identification and classification system and method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117746095A true CN117746095A (en) | 2024-03-22 |
Family
ID=90255225
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311516531.1A Pending CN117746095A (en) | 2023-11-14 | 2023-11-14 | Overhead transmission line fault and interference identification and classification system and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117746095A (en) |
-
2023
- 2023-11-14 CN CN202311516531.1A patent/CN117746095A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
CN111598166B (en) | Single-phase earth fault classification method and system based on principal component analysis and Softmax function | |
CN110726898B (en) | Power distribution network fault type identification method | |
Luo et al. | Stacked auto-encoder based fault location in VSC-HVDC | |
CN110702966B (en) | Fault arc detection method, device and system based on probabilistic neural network | |
CN110808580B (en) | Quick voltage sag source identification method based on wavelet transformation and extreme learning machine | |
CN110247420B (en) | Intelligent fault identification method for HVDC transmission line | |
CN112557826A (en) | Ship electric power system fault diagnosis method | |
CN111723684A (en) | Method for identifying transient overvoltage type in offshore wind farm | |
CN114239703A (en) | Active power distribution system fault diagnosis method, system, equipment and storage medium | |
CN113466634A (en) | Ground fault waveform identification method based on fault indicator | |
CN114252739B (en) | Power distribution network single-phase earth fault discrimination method, system, equipment and storage medium | |
Yang et al. | Fa-Mb-ResNet for grounding fault identification and line selection in the distribution networks | |
CN114218988A (en) | Method for identifying unidirectional ground fault feeder line under unbalanced samples | |
CN113702767A (en) | Island direct-current microgrid fault diagnosis method based on wavelet sliding window energy | |
CN117113180A (en) | Direct current distribution network high-resistance fault identification method based on VMD and convolutional neural network | |
Guo et al. | High impedance fault semi-supervised detection of distribution networks based on tri-training and support vector machine | |
CN117746095A (en) | Overhead transmission line fault and interference identification and classification system and method | |
CN113591792B (en) | Transformer fault identification method based on self-organizing competitive neural network algorithm | |
CN114943254A (en) | A-CNN-based optical fiber differential protection algorithm for output line of photovoltaic power station | |
CN110968073B (en) | Double-layer tracing identification method for commutation failure reasons of HVDC system | |
CN114755529A (en) | Multi-feature fusion single-phase earth fault type identification method based on deep learning | |
CN114186590A (en) | Power distribution network single-phase earth fault identification method based on wavelet and deep learning | |
CN114062832A (en) | Method and system for identifying short-circuit fault type of power distribution network | |
Patel et al. | Wavelet and Machine learning based approach for Fault classification in AC Micro-grid system |
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 |