Power transmission line fault classification method and system based on time sequence imaging, and storage medium

Abstract

The invention discloses a power transmission line fault classification method and system based on time sequence imaging and a storage medium. The method comprises the following steps: preprocessing historical fault transient current traveling wave signals of a power transmission line to obtain a GASF or GADF image data set; classifying historical fault types of the power transmission line to form a fault data set; inputting the GASF or GADF image data set into a neural network model, taking the fault data set as a label, training the neural network model, and obtaining a trained fault classification model; and inputting a GASF or GADF image obtained by preprocessing the real-time fault transient current traveling wave signal of the power transmission line into the trained fault classification model to obtain a corresponding fault type of the power transmission line, and classifying and identifying the fault of the power transmission line according to the fault type of the power transmission line. The power transmission line fault feature quantity can be effectively extracted, and power transmission line fault classification can be quickly and accurately realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power transmission line fault diagnosis in a power grid system, and in particular relates to a power transmission line fault classification method, system and storage medium based on time series imaging. Background Art

[0002] Transmission line failures occur frequently. On the one hand, large-capacity units and distributed power generation are continuously put into operation, the system scale is expanding, the structure is becoming more complex, and the accuracy of existing relay protection is reduced, making disturbances more likely to cause fault tripping. On the other hand, transmission lines, which are mostly overhead lines, are exposed to the wild for a long time and are frequently affected by environmental changes and human activities. This determines that transmission lines are the part of the power system that is most prone to failures.

[0003] For the operation and control level of the line, it is necessary not only to quickly eliminate the fault and restore power supply after the fault occurs, but also to improve the protection level of the line, reduce the failure rate, and take preventive measures. However, the protection strategies for different fault types are different, so it is necessary to count, analyze, and identify the fault types of the transmission line, and then propose targeted protection measures. Since the transmission line fault traveling wave contains a large amount of fault-related information, it is of great significance to identify the cause of the fault by using the transmission line fault traveling wave for adopting targeted protection strategies and improving the line protection level. Therefore, it is necessary to invent a transmission line fault classification method based on time series imaging. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the above-mentioned background technology and provide a method, system and storage medium for power transmission line fault classification based on time series imaging. The method has good accuracy and effectiveness, can effectively extract fault feature quantities and quickly and accurately classify faults.

[0005] The technical solution adopted by the present invention is: a transmission line fault classification method based on time series imaging, comprising:

[0006] Preprocess the transient current traveling wave signal of the historical fault of the transmission line to obtain the GASF or GADF image data set;

[0007] Classify the historical fault types of the transmission lines to form a fault data set;

[0008] Input the GASF or GADF image data set into the neural network model, use the fault data set as a label, train the neural network model, and obtain a trained fault classification model;

[0009] The GASF or GADF image obtained after preprocessing the real-time fault transient current traveling wave signal of the transmission line is input into the trained fault classification model to obtain the corresponding transmission line fault type, and the transmission line fault is classified and identified according to the transmission line fault type.

[0010] Further, the preprocessing includes:

[0011] By intercepting and interpolating the waveform of the transient current traveling wave signal of the transmission line fault, the number of sampling points and the sampling frequency are unified to form a fault standardized data set;

[0012] Perform GAF transformation on the fault standardized dataset to form a GASF or GADF image dataset.

[0013] Furthermore, the types of transmission line faults are divided into bypass failure, counterattack, ice damage, external damage, bird damage, and wind deviation.

[0014] Furthermore, the method for unifying the number of sampling points and the sampling frequency is:

[0015] The sampling frequency of the transient current traveling wave signal of the transmission line fault is unified as f, the sampling time is unified as a fixed value L within the set time, and the number of sampling points is f*L;

[0016] The traveling wave signals with sampling time greater than or equal to L are intercepted, and the traveling wave signals with sampling time less than L are discarded; the traveling wave signals with sampling time of L after interception but with sampling points less than f*L are interpolated and resampled until the sampling time is unified to L and the sampling points are unified to f*L.

[0017] Furthermore, performing GAF transformation on the fault normalization data set includes:

[0018] The normalized transmission line fault transient current traveling wave time series X = {x1, x2, … x i ,…x n} is scaled to the range [-1,1] by the following formula

[0019]

[0020] In the formula, are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; max(X) is the maximum value of each data point in the time series X; min(X) is the minimum value of each data point in the time series X;

[0021] The scaled transmission line fault transient current traveling wave is expressed in polar coordinates. The polar coordinate transformation formula is as follows:

[0022]

[0023] Where: t i is the timestamp; N is the constant factor for regularizing the polar coordinate generation space; φ is the phase angle; r is the polar coordinate radius; is the standardized and scaled X; are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X;

[0024] Substituting the transient current traveling wave of the transmission line fault expressed in polar coordinates into the two unique inner product forms of the Gramian angular field, two different Gramian angular fields are obtained: GASF (Gramian Angular Summation Fields) and GADF (Gramian Angular Difference Fields). The calculation formulas are as follows:

[0025]

[0026]

[0027] In the formula, φ1, φ2...φ n For X~ i Each data point corresponds to a phase angle in the polar coordinate system.

[0028] Furthermore, the neural network model includes an input layer, a convolutional layer, a maximum pooling layer, 4 residual blocks, a global average pooling layer, a fully connected layer, a softmax layer and an output layer.

[0029] The present invention also discloses a transmission line fault classification system based on time series imaging, wherein the transmission line fault classification system based on time series imaging comprises a data acquisition module, a preprocessing module, a fault data module, a network training module and a fault classification module.

[0030] A data acquisition module, used to acquire historical fault transient current traveling wave signals and real-time fault transient current traveling wave signals of the transmission line;

[0031] A preprocessing module, used for preprocessing the historical fault transient current traveling wave signal and the real-time fault transient current traveling wave signal of the transmission line to obtain a GASF or GADF image data set;

[0032] A fault data module is used to classify the historical fault types of the transmission line to form a fault data set;

[0033] The network training module is used to input the GASF or GADF image data set into the neural network model, use the fault data set as a label, train the neural network model, and obtain a trained fault classification model;

[0034] The fault classification module is used to input the GASF or GADF image data set obtained after preprocessing the real-time fault transient current traveling wave signal of the transmission line into the trained fault classification model to obtain the corresponding transmission line fault type, and classify and identify the transmission line fault according to the transmission line fault type.

[0035] Further, the preprocessing includes:

[0036] By intercepting and interpolating the waveform of the transient current traveling wave signal of the transmission line fault, the number of sampling points and the sampling frequency are unified to form a fault standardized data set;

[0037] Perform GAF transformation on the fault standardized dataset to form a GASF or GADF image dataset.

[0038] Furthermore, the types of transmission line faults are divided into bypass failure, counterattack, ice damage, external damage, bird damage, and wind deviation.

[0039] Furthermore, the method for unifying the number of sampling points and the sampling frequency is:

[0040] The sampling frequency of the transient current traveling wave signal of the transmission line fault is unified as f, the sampling time is unified as a fixed value L within the set time, and the number of sampling points is f*L;

[0041] The traveling wave signals with sampling time greater than or equal to L are intercepted, and the traveling wave signals with sampling time less than L are discarded; the traveling wave signals with sampling time of L after interception but with sampling points less than f*L are interpolated and resampled until the sampling time is unified to L and the sampling points are unified to f*L.

[0042] Furthermore, performing GAF transformation on the fault normalization data set includes:

[0043] The normalized transmission line fault transient current traveling wave time series X={x1,x2,...x i ,...x n} is scaled to the range [-1,1] by the following formula

[0044]

[0045] In the formula, are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; max(X) is the maximum value of each data point in the time series X; min(X) is the minimum value of each data point in the time series X;

[0046] The scaled transmission line fault transient current traveling wave is expressed in polar coordinates. The polar coordinate transformation formula is as follows:

[0047]

[0048] Where: t i is the timestamp; N is the constant factor for regularizing the polar coordinate generation space; φ is the phase angle; r is the polar coordinate radius; is the standardized and scaled X; are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X;

[0049] Substituting the transient current traveling wave of the transmission line fault expressed in polar coordinates into the two unique inner product forms of the Gramian angular field, two different Gramian angular fields are obtained: GASF (Gramian Angular Summation Fields) and GADF (Gramian Angular Difference Fields). The calculation formulas are as follows:

[0050]

[0051]

[0052] In the formula, φ1, φ2...φ n for Each data point corresponds to a phase angle in the polar coordinate system.

[0053] Furthermore, the neural network model includes an input layer, a convolutional layer, a maximum pooling layer, 4 residual blocks, a global average pooling layer, a fully connected layer, a softmax layer and an output layer.

[0054] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the power transmission line fault classification method based on time series imaging are implemented.

[0055] Transient traveling wave data of transmission lines belongs to one-dimensional time series data. Compared with traditional machine learning methods, deep learning can automatically learn from the low-level attributes of the original data and extract higher-level meaningful features. Therefore, deep learning is used for time series classification research, such as convolutional neural networks, recurrent neural networks, attention mechanisms, etc. In order to better extract the time-frequency characteristics of transient current traveling wave signals and classify them with deep learning models,

[0056] Gramian Angular Field (GAF) is a time series imaging method that can convert one-dimensional time series signals into two-dimensional images and can retain the characteristics of the original signal to the greatest extent. The basic principle of Gramian Angular Field is to first convert one-dimensional data from the Cartesian coordinate system to the polar coordinate system, and then obtain information characterizing time correlation through trigonometric functions and / or differences based on its unique inner product definition form, and then tile it into the image from the upper left to the lower right, generating two types of images: Gramian Angular Summation Fields (GASF) and Gramian Angular Difference Fields (GADF).

[0057] The neural network model of the present invention is ResNeXt-18, which is an optimized CNN model with the advantages of simple structure, high precision and easy model transplantation. It effectively solves the shortcomings of weak representation ability and easy overfitting of traditional ResNet deep learning. The method has higher classification accuracy and robustness. The present invention converts the one-dimensional time signal collected by the distributed fault monitoring equipment into two-dimensional image information through GAF transformation, retaining the complete information and time dependence of the signal; then, the ResNeXt-18 neural network model is introduced to realize the accurate diagnosis of transmission line faults.

[0058] The present invention proposes to adopt GAF transformation for transient current traveling wave signals, converting time-frequency domain features into image features, thereby retaining the complete information and time dependence of the signal. At the same time, it can also make full use of the advantages of deep learning networks in image classification and recognition to perform transmission line fault classification processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is the overall flow chart of the transmission line fault classification method based on time series imaging;

[0060] Figure 2 Generate a 2D image flow chart for GAF transformation;

[0061] Figure 3 Schematic diagram of converting current traveling wave signals of different fault types into GASF images;

[0062] Figure 4 It is a system diagram of the module connection of the present invention. DETAILED DESCRIPTION

[0063] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments to facilitate a clear understanding of the present invention, but they do not constitute a limitation on the present invention.

[0064] Embodiment 1:

[0065] like Figure 1 As shown, the present invention provides a method for classifying power transmission line faults based on time series imaging, comprising the following steps:

[0066] Acquire the historical fault transient current traveling wave signal of the transmission line, and preprocess the historical fault transient current traveling wave signal of the transmission line to obtain a GASF or GADF image data set;

[0067] Classify the historical fault types of the transmission lines to form a fault data set;

[0068] Input the GASF or GADF image dataset into the ResNeXt-18 neural network model, use the fault dataset as a label, train the ResNeXt-18 neural network model, and obtain a trained fault classification model;

[0069] The GASF or GADF image obtained after preprocessing the real-time fault transient current traveling wave signal of the transmission line is input into the trained fault classification model to obtain the corresponding transmission line fault type, and the transmission line fault is classified and identified according to the transmission line fault type.

[0070] In the above technical solution, the preprocessing includes:

[0071] By intercepting and interpolating the waveform of the transient current traveling wave signal of the transmission line fault, the number of sampling points and the sampling frequency are unified to form a fault standardized data set;

[0072] Perform GAF transformation on the fault standardized dataset to form a GASF or GADF image dataset.

[0073] In the above technical solution, the types of transmission line faults are divided into bypass failure, counterattack, ice damage, external damage, bird damage, and wind deviation.

[0074] In the above technical solution, the method for unifying the number of sampling points and the sampling frequency is:

[0075] The sampling frequency of the transient current traveling wave signal of the transmission line fault is unified as f, the sampling time is unified as a fixed value L within 1.2ms~1.8ms, and the number of sampling points is f*L;

[0076] The traveling wave signals with sampling time greater than L are truncated, and the traveling wave signals with sampling time less than L are discarded; the traveling wave signals with sampling time of L after truncation but with sampling points less than f*L are interpolated and resampled until the sampling time is unified to L and the sampling points are unified to f*L.

[0077] In the above technical solution, if Figure 2 As shown in FIG. 1 , performing GAF transformation on the fault normalization data set includes:

[0078] The normalized transmission line fault transient current traveling wave time series X={x1,x2,...x i ,...x n} is scaled to the range [-1,1] by the following formula

[0079]

[0080] In the formula, are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; max(X) is the maximum value of each data point in the time series X; min(X) is the minimum value of each data point in the time series X.

[0081] The scaled transmission line fault transient current traveling wave is expressed in polar coordinates. The polar coordinate transformation formula is as follows:

[0082]

[0083] Where: t i is the timestamp; N is the constant factor for regularizing the polar coordinate generation space; φ is the phase angle; r is the polar coordinate radius; is the standardized and scaled X; are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X.

[0084] Substituting the transient current traveling wave of the transmission line fault expressed in polar coordinates into the two unique inner product forms of the Gramian angular field, two different Gramian angular fields are obtained: GASF (Gramian Angular Summation Fields) and GADF (Gramian Angular Difference Fields). The calculation formulas are as follows:

[0085]

[0086]

[0087] In the formula, φ1, φ2...φ n for Each data point corresponds to a phase angle in the polar coordinate system.

[0088] In the above technical solution, the ResNeXt-18 neural network model includes an input layer, a convolutional layer, a maximum pooling layer, 4 residual blocks, a global average pooling layer, a fully connected layer, a softmax layer and an output layer.

[0089] In the above technical solution, the training process of the ResNeXt-18 neural network model is:

[0090] The GASF or GADF image dataset is divided into a training set and a test set in proportion. The input layer uses the training set as input to the ResNeXt-18 neural network model for training. The image size in the GASF or GADF image dataset is 224*224*3.

[0091] The input layer is used to receive image input. The convolution layer completes the preliminary feature extraction of the input image through the two characteristics of local receptive field and weight sharing. The convolution kernel size K is set to 7*7, the number is 64, the step size S is 2, and the padding P is set to 3.

[0092] The formula for calculating the output image size is:

[0093]

[0094] Where N out is the output image size; N in is the input image size; P is the padding size; K is the convolution kernel size; S is the step size.

[0095] The input layer image size is 224*224*3, that is, 3 channels, each channel size is 224*224. After processing by the convolution layer, the convolution layer output image size is 112*112*64.

[0096] The main work of the pooling layer is to reduce the dimension of the convolutional data, so as to further extract the features. After the convolutional layer processing, a maximum pooling layer is used to extract the maximum value of the local feature vector. The convolution kernel size of the pooling layer is 3*3, the step size is 2, and the padding is 1. After the pooling layer processing, the output neuron size is 56*56*64, that is, the maximum pooling layer does not change the number of channels of the data, but will halve the size of the data.

[0097] The residual block consists of two convolutional layers and a skip connection. The skip connection adds the input directly to the output of the second convolutional layer to obtain the final output of the residual block. The residual block passes the input directly to the output through the skip connection and does not perform any processing during the transmission process, thereby avoiding the gradient vanishing and gradient exploding problems and making the network easier to train. This design enables the network to better learn the residual between the input and output, thereby improving the performance of the model.

[0098] The convolution kernel size of the two convolutional layers in residual block 1 is 3*3, the stride is 1, and the padding is 1. After processing, the output neuron size is 56*56*64, that is, residual block 1 does not change the size and number of channels of the data.

[0099] Residual block 2 first passes through a 1*1 convolution layer and then undergoes a downsampling. The final output data is 28*28*128. That is, the output channels are doubled and the output data size is halved. After being processed by residual block 2, the output neuron size is 28*28*128.

[0100] Residual block 3 also performs 1*1 convolution and downsampling. The final output is 14*14*256. That is, the output channels are doubled and the output data size is halved.

[0101] The structure of residual block 4 is the same as that of residual block 3, and the final output is 7*7*512.

[0102] The convolutional data is further reduced in dimension through the global average pooling layer to extract the average value of the local feature vector; the average pooling layer parameters are set to 7*7*512, and the output neuron size is 1*1*512.

[0103] After the global average pooling layer, ResNet-18 uses a fully connected layer to map the feature vector to a vector with the same number of categories.

[0104] The last layer converts this vector into a category probability distribution through the softmax activation function.

[0105] The transmission line fault type corresponding to the highest probability value of the final ResNet-18 output result is the model identification result.

[0106] The overall architecture and parameters of the ResNet-18 model are shown in the following table:

[0107]

[0108]

[0109] The experimental data set of this embodiment is the standardized transient current traveling wave data, including lightning strike and non-lightning strike fault data, where lightning strike fault data can be divided into shielding failure and counterattack, and non-lightning strike fault data includes external damage, ice damage, bird damage and wind deflection. Among them, 600 lightning strike fault data and 400 non-lightning strike fault data are selected, and the number of data points is 2400.

[0110] First, the experimental data set is converted into Gram angular field images, and then the image resolution is compressed to 224*224, and finally input into the ResNeXt-18 classification model for training. Figure 3 shown.

[0111] The experiment of this embodiment uses matlab 2023b software for simulation and verification, and the experimental platform is the Windows 10 system of GeForce RTX1650. The image preprocessing is performed in the CPU environment, and the model training is performed in the GPU environment, thereby improving the experimental efficiency. The experiment first selects 200 lightning strike and non-lightning strike fault samples for binary classification simulation and verification, and then selects 150 bypass strike, counterattack, and non-lightning strike fault samples for three classifications; the classification model is ResNeXt-18.

[0112] For the classification of transmission line fault causes, the experiment in this embodiment uses confusion matrix (Confusion matrix), sensitivity (Sensitivity, SEN), specificity (Specificity, SPE), precision (Precision, PRE), F1 value, accuracy (Accuracy, ACC) and loss (Loss) to comprehensively evaluate the classification effect of the model.

[0113] Each row of the confusion matrix represents the actual class of the data (Actual class). The total number of each row is the total number of data instances of that class. Each column represents the predicted class (Predict class). The total number of each column is the total number of data predicted to be of that class. Based on the confusion matrix, the classification effect of each class (Positive and Negative) can be intuitively analyzed. The confusion matrix is ​​shown in the table below.

[0114]

[0115] In order to further verify the superiority of the improved network model in this embodiment, this embodiment also introduces four popular classification models (CNN, GoogLeNet, ResNet-18) to perform binary classification of lightning strike and non-lightning strike fault samples, and compares them according to the selected evaluation indicators to prove the superiority of ResNeXt-18 used in this article. The specific results are shown in the following table.

[0116]

[0117]

[0118] The SEN, PRE, F1, and ACC of the ResNeXt-18 model are 2.26%, 0.00%, 1.18%, and 1.25% higher than those of the ResNet-18 model in lightning fault classification, and 0.07%, 2.50%, 1.33%, and 1.25% higher in non-lightning fault classification; the SEN, PRE, F1, and ACC of the ResNeXt-18 model are -2.25%, 5%, 1.43%, and 1.2 higher than those of the GoogLeNet model in lightning fault classification. 5%, and 4.58%, -2.50%, 1.08%, and 1.25% higher in non-lightning fault classification; the ResNeXt-18 model is 4.64%, 2.50%, 3.62%, and 3.75% higher than the CNN model in lightning fault classification, and 2.66%, 5.00%, 3.89%, and 3.75% higher in non-lightning fault classification. This verifies that ResNeXt-18 shows better classification effect than other models.

[0119] Embodiment 2:

[0120] like Figure 4 As shown, a transmission line fault classification system based on time series imaging, the transmission line fault classification system based on time series imaging includes a data acquisition module, a preprocessing module, a fault data module, a network training module and a fault classification module.

[0121] A data acquisition module, used to acquire historical fault transient current traveling wave signals and real-time fault transient current traveling wave signals of the transmission line;

[0122] A preprocessing module, used for preprocessing the historical fault transient current traveling wave signal and the real-time fault transient current traveling wave signal of the transmission line to obtain a GASF or GADF image data set;

[0123] A fault data module is used to classify the historical fault types of the transmission line to form a fault data set;

[0124] The network training module is used to input the GASF or GADF image data set into the ResNeXt-18 neural network model, and use the fault data set as a label to train the ResNeXt-18 neural network model to obtain a trained fault classification model;

[0125] The fault classification module is used to input the GASF or GADF image data set obtained after preprocessing the real-time fault transient current traveling wave signal of the transmission line into the trained fault classification model to obtain the corresponding transmission line fault type, and classify and identify the transmission line fault according to the transmission line fault type.

[0126] In the above technical solution, the preprocessing includes:

[0127] By intercepting and interpolating the waveform of the transient current traveling wave signal of the transmission line fault, the number of sampling points and the sampling frequency are unified to form a fault standardized data set;

[0128] Perform GAF transformation on the fault standardized dataset to form a GASF or GADF image dataset.

[0129] In the above technical solution, the types of transmission line faults are divided into bypass failure, counterattack, ice damage, external damage, bird damage, and wind deviation.

[0130] In the above technical solution, the method for unifying the number of sampling points and the sampling frequency is:

[0131] The sampling frequency of the transient current traveling wave signal of the transmission line fault is unified as f, the sampling time is unified as a fixed value L within 1.2ms~1.8ms, and the number of sampling points is f*L;

[0132] The traveling wave signals with sampling time greater than L are truncated, and the traveling wave signals with sampling time less than L are discarded; the traveling wave signals with sampling time of L after truncation but with sampling points less than f*L are interpolated and resampled until the sampling time is unified to L and the sampling points are unified to f*L.

[0133] In the above technical solution, if Figure 2 As shown in FIG. 1 , performing GAF transformation on the fault normalization data set includes:

[0134] The normalized transmission line fault transient current traveling wave time series X={x1,x2,...x i ,...x n} is scaled to the range [-1,1] by the following formula

[0135]

[0136] In the formula, are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; max(X) is the maximum value of each data point in the time series X; min(X) is the minimum value of each data point in the time series X.

[0137] The scaled transmission line fault transient current traveling wave is expressed in polar coordinates. The polar coordinate transformation formula is as follows:

[0138]

[0139] Where: t iis the timestamp; N is the constant factor for regularizing the polar coordinate generation space; φ is the phase angle; r is the polar coordinate radius; is the standardized and scaled X; are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X.

[0140] Substituting the transient current traveling wave of the transmission line fault expressed in polar coordinates into the two unique inner product forms of the Gramian angular field, two different Gramian angular fields are obtained: GASF (Gramian Angular Summation Fields) and GADF (Gramian Angular Difference Fields). The calculation formulas are as follows:

[0141]

[0142]

[0143] In the formula, φ1, φ2...φ n for Each data point corresponds to a phase angle in the polar coordinate system.

[0144] In the above technical solution, the ResNeXt-18 neural network model includes an input layer, a convolutional layer, a maximum pooling layer, 4 residual blocks, a global average pooling layer, a fully connected layer, a softmax layer and an output layer.

[0145] Embodiment 3:

[0146] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the power transmission line fault classification method based on time series imaging are implemented.

[0147] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0148] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0150] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.

Claims

1. A method for classifying transmission line faults based on time series imaging, characterized in that: Preprocess the transient current traveling wave signal of the historical fault of the transmission line to obtain the GASF or GADF image data set; Classify the historical fault types of the transmission lines to form a fault data set; Input the GASF or GADF image data set into the neural network model, use the fault data set as a label, train the neural network model, and obtain a trained fault classification model; The GASF or GADF image obtained after preprocessing the real-time fault transient current traveling wave signal of the transmission line is input into the trained fault classification model to obtain the corresponding transmission line fault type, and the transmission line fault is classified and identified according to the transmission line fault type.

2. The method for classifying power transmission line faults based on time series imaging according to claim 1, characterized in that: Preprocessing includes: By intercepting and interpolating the waveform of the transient current traveling wave signal of the transmission line fault, the number of sampling points and the sampling frequency are unified to form a fault standardized data set; Perform GAF transformation on the fault standardized dataset to form a GASF or GADF image dataset.

3. The method for classifying power transmission line faults based on time series imaging according to claim 1, characterized in that: Transmission line fault types are divided into bypass failure, counterattack, ice damage, external damage, bird damage, and wind deviation.

4. The method for classifying power transmission line faults based on time series imaging according to claim 2, characterized in that: The method to unify the number of sampling points and sampling frequency is: The sampling frequency of the transient current traveling wave signal of the transmission line fault is unified to f, the sampling time is unified to a fixed value L within the set time, and the number of sampling points is f*L; The traveling wave signals with sampling time greater than or equal to L are intercepted, and the traveling wave signals with sampling time less than L are discarded; the traveling wave signals with sampling time of L after interception but with sampling points less than f*L are interpolated and resampled until the sampling time is unified to L and the sampling points are unified to f*L.

5. The method for classifying power transmission line faults based on time series imaging according to claim 2, characterized in that: GAF transformation of fault normalization data set includes: The normalized transmission line fault transient current traveling wave time series X={x1,x2,...x i ,...x n } is scaled to the range [-1,1] by the following formula In the formula, are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; max(X) is the maximum value of each data point in the time series X; min(X) is the minimum value of each data point in the time series X; The scaled transmission line fault transient current traveling wave is expressed in polar coordinates. The polar coordinate transformation formula is as follows: Where: t i is the timestamp; N is the constant factor of the regularized polar coordinate generation space; φ is the phase angle; r is the polar coordinate radius; X~ is the standardized and scaled X; are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; Substituting the transient current traveling wave of the transmission line fault represented in polar coordinates into the two unique inner product forms of the Gram angle field, two different Gram angle fields are obtained: GASF and GADF. The calculation formulas are as follows: In the formula, φ1, φ2...φ n for Each data point corresponds to a phase angle in the polar coordinate system.

6. The method for classifying power transmission line faults based on time series imaging according to claim 1, characterized in that: The neural network model includes an input layer, a convolutional layer, a maximum pooling layer, 4 residual blocks, a global average pooling layer, a fully connected layer, a softmax layer and an output layer.

7. A transmission line fault classification system based on time series imaging, characterized in that: The transmission line fault classification system based on time series imaging includes a data acquisition module, a preprocessing module, a fault data module, a network training module and a fault classification module. A data acquisition module, used to acquire historical fault transient current traveling wave signals and real-time fault transient current traveling wave signals of the transmission line; A preprocessing module, used for preprocessing the historical fault transient current traveling wave signal and the real-time fault transient current traveling wave signal of the transmission line to obtain a GASF or GADF image data set; A fault data module is used to classify the historical fault types of the transmission line to form a fault data set; The network training module is used to input the GASF or GADF image data set into the neural network model, use the fault data set as a label, train the neural network model, and obtain a trained fault classification model; The fault classification module is used to input the GASF or GADF image data set obtained after preprocessing the real-time fault transient current traveling wave signal of the transmission line into the trained fault classification model to obtain the corresponding transmission line fault type, and classify and identify the transmission line fault according to the transmission line fault type.

8. The power transmission line fault classification system based on time series imaging according to claim 7, characterized in that: Preprocessing includes: By intercepting and interpolating the waveform of the transient current traveling wave signal of the transmission line fault, the number of sampling points and the sampling frequency are unified to form a fault standardized data set; Perform GAF transformation on the fault standardized dataset to form a GASF or GADF image dataset.

9. The power transmission line fault classification system based on time series imaging according to claim 7, characterized in that: Transmission line fault types are divided into bypass failure, counterattack, ice damage, external damage, bird damage, and wind deviation.

10. The power transmission line fault classification system based on time series imaging according to claim 7, characterized in that: The method to unify the number of sampling points and sampling frequency is: The sampling frequency of the transient current traveling wave signal of the transmission line fault is unified to f, the sampling time is unified to a fixed value L within the set time, and the number of sampling points is f*L; The traveling wave signals with sampling time greater than or equal to L are intercepted, and the traveling wave signals with sampling time less than L are discarded; the traveling wave signals with sampling time of L after interception but with sampling points less than f*L are interpolated and resampled until the sampling time is unified to L and the sampling points are unified to f*L.

11. The power transmission line fault classification system based on time series imaging according to claim 7, characterized in that: GAF transformation of fault normalization data set includes: The normalized transmission line fault transient current traveling wave time series X = {x1, x2, …x i ,...x n } is scaled to the range [-1,1] by the following formula In the formula, are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; max(X) is the maximum value of each data point in the time series X; min(X) is the minimum value of each data point in the time series X; The scaled transmission line fault transient current traveling wave is expressed in polar coordinates. The polar coordinate transformation formula is as follows: Where: t i is the timestamp; N is the constant factor for regularizing the polar coordinate generation space; φ is the phase angle; r is the polar coordinate radius; is the standardized and scaled X; are the data points of the normalized and scaled transmission line fault transient current traveling wave time series X; Substituting the transient current traveling wave of the transmission line fault represented in polar coordinates into the two unique inner product forms of the Gram angle field, two different Gram angle fields are obtained: GASF and GADF. The calculation formulas are as follows: In the formula, φ1, φ2...φ n for Each data point corresponds to a phase angle in the polar coordinate system.

12. The power transmission line fault classification system based on time series imaging according to claim 7, characterized in that: The neural network model includes an input layer, a convolutional layer, a maximum pooling layer, 4 residual blocks, a global average pooling layer, a fully connected layer, a softmax layer and an output layer.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the power transmission line fault classification method based on time series imaging as described in any one of claims 1 to 6 are implemented.

Images