Fault diagnosis method and device of power transmission line, storage medium and computer equipment

Abstract

The invention discloses a fault diagnosis method and device for a power transmission line, a storage medium and computer equipment, relates to the technical field of power fault analysis, and mainly aims to solve the problem of low fault diagnosis accuracy caused by unstable output during feature extraction of a power transmission line fault signal obtained in a high-noise environment. Comprising the following steps: acquiring a power transmission line transient current signal containing fault information; performing feature extraction processing on the transient current signal of the power transmission line by adopting a discrete wavelet transform method to obtain a target feature vector for fault analysis; and performing fault classification processing on the target feature vector by using a time convolutional neural network model to obtain a power transmission line fault type.

Description

Technical Field

[0001] The present invention relates to the technical field of power fault analysis, and in particular to a method and device for diagnosing faults of a power transmission line, a storage medium, and a computer device. Background Art

[0002] Fault diagnosis of ultra-high voltage overhead transmission lines plays a vital role in power systems, as it helps to identify potential fault points in advance, thereby taking preventive measures and reducing the risk of failure. However, in practical applications, transmission line fault prediction still faces some problems and defects.

[0003] Transmission line fault signals usually contain a lot of noise, which reduces the accuracy of fault detection and classification. In addition, fault signals are transient and contain multiple frequency components, which has a significant impact on accurate fault classification and often leads to unstable classification results. Although traditional research methods also perform denoising on fault signals, in high-noise environments, the output of existing feature extraction technologies for transmission line fault signals is unstable when extracting effective signals, which affects the fault diagnosis accuracy of the prediction model. Summary of the invention

[0004] In view of this, the present invention provides a transmission line fault diagnosis method and device, storage medium, and computer equipment, the main purpose of which is to solve the problem of low fault diagnosis accuracy caused by unstable output when extracting transmission line fault signal features obtained in a high noise environment.

[0005] According to one aspect of the present invention, a method for diagnosing a fault of a power transmission line is provided, comprising:

[0006] Acquire the transient current signal of the transmission line containing fault information;

[0007] Using discrete wavelet transform method to perform feature extraction processing on the transient current signal of the transmission line to obtain a target feature vector for fault analysis;

[0008] A temporal convolutional neural network model is used to perform fault classification processing on the target feature vector to obtain the transmission line fault type.

[0009] Furthermore, the discrete wavelet transform method is used to perform feature extraction processing on the transient current signal of the transmission line to obtain a target feature vector for fault analysis, including:

[0010] Obtaining a decomposition level of the discrete wavelet transform method;

[0011] Decomposing the transient current signal of the transmission line layer by layer according to the decomposition level to obtain a signal decomposition result corresponding to each decomposition level;

[0012] The signal decomposition result is selectively processed to obtain the target feature vector.

[0013] Furthermore, the decomposition processing of the transient current signal of the transmission line is performed layer by layer according to the decomposition level to obtain the signal decomposition result corresponding to each decomposition level, including:

[0014] Using a high-pass filter and a low-pass filter to perform a first-level decomposition process on the transient current signal of the transmission line to obtain a first detail component and a first approximate component;

[0015] Using the high-pass filter and the low-pass filter to perform a second-level decomposition process on the first approximate component to obtain a second detail component and a second approximate component;

[0016] The high-pass filter and the low-pass filter are used to perform layer-by-layer decomposition processing on the second approximate component until all decomposition levels are decomposed, thereby obtaining the signal decomposition result corresponding to each decomposition level.

[0017] Furthermore, when the signal decomposition result is selected and processed, one of an energy method, an entropy method and a statistical feature method is adopted.

[0018] Furthermore, before the temporal convolutional neural network model is used to perform fault classification processing on the target feature vector, the method further includes:

[0019] Acquire a signal sample set of a transient current signal of a transmission line, and use the discrete wavelet transform to perform feature extraction processing on each of the transient current signals in the signal sample set to obtain a feature vector sample set;

[0020] Using the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result;

[0021] Based on the model training results, the fault classification performance of the temporal convolutional neural network model is evaluated to obtain a temporal convolutional neural network model that meets the evaluation requirements.

[0022] Furthermore, the adopting of the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result includes:

[0023] Performing causal convolution processing on the feature vector samples as an input sequence to obtain a first output sequence; the first output sequence is offset from the input sequence by a step in time;

[0024] Performing dilated convolution processing on the input sequence to obtain a second output sequence; the second output sequence is suitable for a longer receiving field than the first output sequence;

[0025] Performing a residual connection based on the first output sequence and the second output sequence to obtain a residual connection result;

[0026] After stacking the residual connection results, a fully connected layer is used to perform fault classification to obtain the model training result.

[0027] Furthermore, the evaluation process of the fault classification performance of the temporal convolutional neural network model based on the model training result includes:

[0028] Based on the model training result, at least one of the confusion matrix, precision, recall rate and F1 score is used to evaluate the fault classification performance of the temporal convolutional neural network model.

[0029] According to another aspect of the present invention, there is provided a fault diagnosis device for a power transmission line, comprising:

[0030] An acquisition module, used for acquiring a transient current signal of a power transmission line containing fault information;

[0031] A feature extraction module, used to perform feature extraction processing on the transient current signal of the transmission line using a discrete wavelet transform method to obtain a target feature vector for fault analysis;

[0032] The fault diagnosis module is used to perform fault classification processing on the target feature vector using a temporal convolutional neural network model to obtain a transmission line fault type.

[0033] Furthermore, the feature extraction module includes a hierarchical acquisition unit, a decomposition unit, and a selection unit;

[0034] The level acquisition unit is used to acquire the decomposition level of the discrete wavelet transform method;

[0035] The decomposition unit is used to decompose the transient current signal of the transmission line layer by layer according to the decomposition level to obtain a signal decomposition result corresponding to each decomposition level;

[0036] The selection unit is used to perform selection processing on the signal decomposition result to obtain the target feature vector.

[0037] Furthermore, the decomposition unit is also used for:

[0038] Using a high-pass filter and a low-pass filter to perform a first-level decomposition process on the transient current signal of the transmission line to obtain a first detail component and a first approximate component;

[0039] Using the high-pass filter and the low-pass filter to perform a second-level decomposition process on the first approximate component to obtain a second detail component and a second approximate component;

[0040] The high-pass filter and the low-pass filter are used to perform layer-by-layer decomposition processing on the second approximate component until all decomposition levels are decomposed, thereby obtaining the signal decomposition result corresponding to each decomposition level.

[0041] Furthermore, the selection unit is also used for:

[0042] When selecting and processing the signal decomposition result, one of the energy method, entropy method and statistical feature method is adopted.

[0043] Furthermore, the device also includes a model training module, and the model training module is used to:

[0044] Acquire a signal sample set of a transient current signal of a transmission line, and use the discrete wavelet transform to perform feature extraction processing on each of the transient current signals in the signal sample set to obtain a feature vector sample set;

[0045] Using the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result;

[0046] Based on the model training results, the fault classification performance of the temporal convolutional neural network model is evaluated to obtain a temporal convolutional neural network model that meets the evaluation requirements.

[0047] Furthermore, the model training module is also used for:

[0048] Performing causal convolution processing on the feature vector samples as an input sequence to obtain a first output sequence; the first output sequence is offset from the input sequence by a step in time;

[0049] Performing dilated convolution processing on the input sequence to obtain a second output sequence; the second output sequence is suitable for a longer receiving field than the first output sequence;

[0050] Performing a residual connection based on the first output sequence and the second output sequence to obtain a residual connection result;

[0051] After stacking the residual connection results, a fully connected layer is used to perform fault classification to obtain the model training result.

[0052] Furthermore, the model training module is also used for:

[0053] Based on the model training result, at least one of the confusion matrix, precision, recall rate and F1 score is used to evaluate the fault classification performance of the temporal convolutional neural network model.

[0054] According to another aspect of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, wherein the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for diagnosing faults of power transmission lines.

[0055] According to another aspect of the present invention, there is provided a computer device, comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0056] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned power transmission line fault diagnosis method.

[0057] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0058] The present invention provides a method and device for diagnosing faults of a power transmission line, a storage medium, and a computer device. Compared with the prior art, the present invention uses a discrete wavelet transform method to perform feature extraction processing on a transient current signal of a power transmission line containing fault information, which can not only filter out high-frequency noise in the transient current signal of the power transmission line, but also analyze and process the transient current signal of the power transmission line according to the frequency to obtain the most representative feature vector, thereby improving the model's ability to identify power transmission line faults. The present invention also uses a temporal convolutional neural network model to perform fault classification processing on the target feature vector, which not only improves the accuracy of transmission line fault classification, but also has lower memory requirements due to the use of extended convolution. In addition, the temporal convolutional neural network model uses a parallel processing method to improve computational efficiency, thereby improving the real-time performance of power transmission line fault diagnosis.

[0059] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0061] Figure 1 A schematic flow chart of a method for diagnosing a fault of a power transmission line provided by an embodiment of the present invention is shown;

[0062] Figure 2 A schematic flow chart of another method for diagnosing a fault of a power transmission line provided by an embodiment of the present invention is shown;

[0063] Figure 3 A schematic flow chart of another method for diagnosing a fault of a power transmission line provided by an embodiment of the present invention is shown;

[0064] Figure 4 A schematic flow chart of another method for diagnosing a fault of a power transmission line provided in an embodiment of the present invention is shown;

[0065] Figure 5 A schematic structural diagram of a fault diagnosis device for a power transmission line provided by an embodiment of the present invention is shown;

[0066] Figure 6 A schematic structural diagram of a computer device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0067] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0068] The embodiment of the present invention provides a method for diagnosing a fault in a power transmission line, such as Figure 1 As shown, the method includes:

[0069] 101. Obtain a transient current signal of a transmission line containing fault information;

[0070] In the embodiment of the present invention, since the power transmission line fault is transient, the current execution end obtains a power transmission line transient current signal containing fault information.

[0071] 102. Perform feature extraction processing on the transient current signal of the transmission line using a discrete wavelet transform method to obtain a target feature vector for fault analysis;

[0072] In the embodiment of the present invention, since the transient current signal obtained in step 101 contains multiple frequency components and a large amount of noise, the accuracy of the transmission line fault diagnosis is seriously affected. The current execution end uses the discrete wavelet transform method to perform feature extraction processing on the transient current signal of the transmission line obtained in step 101 to obtain a target feature vector for fault analysis. Among them, the discrete wavelet transform method is a signal processing technology that decomposes the signal containing the fault feature into different frequency components through multi-resolution analysis, thereby extracting features that are helpful for fault detection and classification, that is, the target feature vector in the embodiment of the present invention.

[0073] It should be noted that the discrete wavelet transform in the embodiment of the present invention can be used not only for feature extraction but also for signal denoising. By removing the high-frequency noise components in the transient current signal of the transmission line, the purity of the fault signal can be improved, thereby improving the accuracy of fault detection and classification.

[0074] 103. Use a temporal convolutional neural network model to perform fault classification processing on the target feature vector to obtain a transmission line fault type.

[0075] In an embodiment of the present invention, since the collected transient current signal of the transmission line is a sequence signal arranged in chronological order, the current execution end uses a time convolutional neural network model to perform fault classification processing on the target feature vector obtained in step 102 to obtain the type of transmission line fault. Among them, the time convolutional neural network model has multiple unique advantages in processing time series data. The time convolutional neural network model uses causal convolution to ensure that the model only depends on past inputs and not on future inputs, which enables the model to process input sequences of arbitrary length. The time convolutional neural network model can learn from a larger history through dilated convolutions, thereby capturing longer dependencies. In addition, the residual connection of the time convolutional neural network model allows the layer to learn the identity mapping modification instead of the entire transformation, which helps to accelerate the convergence of the model and allows training at a deeper network level. The high efficiency and parallelism of the time convolutional neural network model enable it to perform well in processing large-scale time series data.

[0076] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to obtain the most representative features from the transient current signal of the transmission line, another fault diagnosis method for the transmission line is provided, such as Figure 2 As shown, the step uses discrete wavelet transform method to perform feature extraction processing on the transient current signal of the transmission line to obtain a target feature vector for fault analysis, including:

[0077] 201. Obtaining a decomposition level of the discrete wavelet transform method;

[0078] In the embodiment of the present invention, the current execution end obtains a preset decomposition level of the discrete wavelet transform method. The decomposition level can be set based on experience or repeated experiments, and the embodiment of the present invention does not specifically limit it.

[0079] 202. Decompose the transient current signal of the transmission line layer by layer according to the decomposition levels to obtain a signal decomposition result corresponding to each decomposition level;

[0080] In the embodiment of the present invention, the current execution end decomposes the transient current signal of the power transmission line obtained in step 101 according to the decomposition level obtained in step 201. The specific decomposition steps are as follows:

[0081] S1. Use a high-pass filter and a low-pass filter to perform a first-level decomposition process on the transient current signal of the transmission line obtained in step 101 to obtain a first detail component and a first approximate component; wherein the first detail component is a high-frequency component of the original transient current signal of the transmission line, the first approximate component is a relatively low-frequency component of the original transient current signal of the transmission line, and the data volume of the first detail component and the first approximate component are both half of the data volume in the original transient current signal of the transmission line.

[0082] S2. Perform a second-level decomposition process on the first approximate component obtained by the high-pass filter and the low-pass filter S1 to obtain a second detail component and a second approximate component; wherein the second detail component is the high-frequency component in the first approximate component obtained in step S1, and the second approximate component is the relatively low-frequency component in the first approximate component obtained in step S1, that is, during the second-level decomposition process, only the relatively low-frequency component in S1 is decomposed, and the data amounts of the second detail component and the second approximate component obtained are both half of the data amount of the first approximate component obtained in S1.

[0083] S3. Using the same method as step S2, the second approximate component obtained in step S2 is decomposed layer by layer through a high-pass filter and a low-pass filter, that is, only the relatively low-frequency components obtained by the previous level of decomposition are decomposed, and the decomposition steps are repeated until all decomposition levels are decomposed, and the signal decomposition results corresponding to each decomposition level are obtained. The signal decomposition results contain information about the signal in different frequency ranges, which is helpful to distinguish different types of faults. For example, when the set decomposition level is 3, after decomposing the second approximate component in S2, the third detail component and the third approximate component are obtained, and the decomposition is terminated; when the set decomposition level is 4, the third approximate component needs to be decomposed to obtain the fourth detail component and the fourth approximate component before the decomposition is terminated, etc. The embodiment of the present invention does not make specific limitations. With each additional layer of decomposition, the frequency resolution and spatial resolution of the signal will change.

[0084] It should be noted that selecting a suitable wavelet basis function is the key to discrete wavelet transform. The embodiment of the present invention adopts Db4 wavelet because it is suitable for processing transient data.

[0085] 203. Selectively process the signal decomposition result to obtain the target feature vector.

[0086] In the embodiment of the present invention, the current execution end selects and processes the signal decomposition result obtained in step 202 to obtain the target feature vector. Specifically, when the signal decomposition result is selected and processed, one of the energy method, entropy method and statistical feature method can be used, which is not specifically limited in the embodiment of the present invention.

[0087] The energy method is used to calculate the energy of the signal. For time domain signals, energy is usually the sum of the squares of the signal amplitudes, that is, E = ∑|x[n]|^ 2 , where x[n] is a sample of the signal and n is a sample index. For frequency domain signals, the energy may be the sum of the squares of the frequency domain components. In an embodiment of the present invention, the current execution end selects a portion with higher energy as a feature: in the time domain or frequency domain, a portion with energy significantly higher than that in the normal operating state is selected as a fault indication feature, which is not specifically limited in the embodiment of the present invention.

[0088] Among them, the entropy method is used to calculate the entropy of the signal. Entropy is a measure of signal uncertainty and can be calculated using a probability distribution function. For discrete signals, entropy H(X) = -∑p(x)*log 2 (p(x)), where p(x) is the probability of the signal taking a certain value. In the embodiment of the present invention, the current execution end selects the part with higher entropy value as the feature: if the fault signal is more random or complex than the normal signal, then its entropy value may be higher. The part with entropy value significantly higher than the normal operation state can be selected as the fault indication feature, which is not specifically limited in the embodiment of the present invention.

[0089] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to further improve the accuracy of high power transmission line fault diagnosis and improve the generalization ability of the model used for fault diagnosis, another transmission line fault diagnosis method is provided, such as Figure 3 As shown, before the step of using the time convolutional neural network model to perform fault classification processing on the target feature vector, the method also includes:

[0090] 301. Acquire a signal sample set of a transient current signal of a transmission line, and use the discrete wavelet transform to perform feature extraction processing on each of the transient current signals in the signal sample set to obtain a feature vector sample set;

[0091] In an embodiment of the present invention, the current execution end obtains a signal sample set of a transient current signal of a transmission line containing fault information. It should be noted that the transient current signals in the signal sample set all correspond to fault type identifiers. According to the fault diagnosis requirements, the fault type identifier can be divided into two categories: faulty and non-faulty; the fault type identifier can also be subdivided into multiple categories such as short circuit faults, lightning faults, external force faults, ground faults, etc., which are not specifically limited in the embodiment of the present invention. After obtaining the signal sample set, the current execution end uses the same discrete wavelet transform method as steps 201 to 203 to perform feature extraction processing on each transient current signal in the signal sample set to obtain a feature vector sample set. It should be noted that the feature vector samples in the feature vector sample set obtained in the embodiment of the present invention have the same fault type identifier as the transient current signal in the original signal sample set, which is not specifically limited in the embodiment of the present invention.

[0092] 302. Perform model training processing on the temporal convolutional neural network model using the feature vector sample set to obtain a model training result;

[0093] In an embodiment of the present invention, the current execution end uses a feature vector sample set to train and process the temporal convolutional neural network model. In the temporal convolutional neural network model architecture, the input layer receives the fault current signal from the transmission line. The convolution layer uses a one-dimensional convolution layer to perform a convolution operation on the input data to extract the features in the time series. The key parameters of the convolution layer, the kernel size (k) and the dilation factor (d), determine the size of the receptive field of the model. The input data is directly added to the output of the convolution layer through a residual connection to help the model learn the identity mapping, accelerate convergence and allow deeper network training. The ReLU activation function is used to increase the nonlinear characteristics of the network. The pooling layer is used to reduce the dimension of the data and reduce the amount of calculation. A fully connected layer is added after the convolution layer for the final classification decision.

[0094] In the embodiment of the present invention, the current execution end uses the cross entropy loss function to measure the accuracy of the prediction of the temporal convolutional neural network model. And the Adam optimizer is used because of its fast convergence characteristics. The learning rate of the model is set to 0.001; the appropriate batch size is selected, and 64 is used; the number of training cycles is set to 30 cycles. Based on the above steps, the model training results are obtained.

[0095] 303. Evaluate the fault classification performance of the temporal convolutional neural network model based on the model training result to obtain a temporal convolutional neural network model that meets the evaluation requirements.

[0096] In the embodiment of the present invention, the current execution end evaluates the fault classification performance of the temporal convolutional neural network model based on the model training result. In the embodiment of the present invention, at least one of the confusion matrix, precision-recall curve, precision, recall and F1 score can be selected to evaluate the fault classification performance of the temporal convolutional neural network model, and the embodiment of the present invention does not make specific restrictions. Among them, the confusion matrix is ​​an error matrix, which is often used in machine learning to visually evaluate the performance of supervised learning algorithms. Among them, the F1 score (F1 Score) is an indicator used in statistics to measure the accuracy of a binary classification model. It takes into account both the precision and recall of the classification model, and can be regarded as a harmonic average of the precision and recall. The maximum value of the F1 score is 1, indicating perfect precision and recall, and the minimum value is 0, indicating complete inaccuracy or no recall. The specific evaluation method can be reasonably selected according to the needs, and the embodiment of the present invention does not make specific restrictions.

[0097] It should be noted that in the embodiment of the present invention, the convergence and stability of the model can also be evaluated by drawing a trend graph of the loss and accuracy changing with the training cycle. After the above evaluation of the fault classification performance of the model, if the fault classification performance of the model does not meet the evaluation requirements, it is necessary to reasonably adjust the model parameters and then train the model until a temporal convolutional neural network model that meets the evaluation requirements is obtained.

[0098] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to improve the accuracy of transmission line fault diagnosis and improve the generalization ability of the model, another transmission line fault diagnosis method is provided, such as Figure 4 As shown, the step uses the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result, including:

[0099] 401. Perform causal convolution processing on the feature vector samples as an input sequence to obtain a first output sequence; the first output sequence is offset from the input sequence by a step in time;

[0100] In the embodiment of the present invention, the current execution end uses the feature vector sample as an input sequence to perform causal convolution processing to obtain a first output sequence. For example, the transient current x of the transmission line containing a series of time step data points is used as an input sequence, denoted as {x 0 ,x 1 …,x T}, T is the length of the sequence. After causal convolution processing, a relevant output is generated at each time step, denoted as {y 0 ,y 1 …,y T}. The mapping relationship is expressed by the model function f∶XT+1 →Y T+1 Completed, the function f maps the x signal vector with (T+1) elements as input to the elements of the output vector Y as follows:

[0101] y t =σ(W*x)+b

[0102] Among them, y t is the output at time t, x is the input sequence, W is the convolution kernel, * represents the convolution operation, σ is the activation function, and b is the bias term.

[0103] It should be noted that there is a step-size offset in time between the first output sequence and the input sequence, that is, the output of the causal convolution at time t only depends on the elements at time t and before, rather than future elements.

[0104] 402. Perform dilated convolution processing on the input sequence to obtain a second output sequence; the second output sequence is suitable for a longer receiving field than the first output sequence;

[0105] In the embodiment of the present invention, the current execution end also performs dilated convolution processing on the input sequence to obtain a second output sequence. The temporal convolutional neural network model can learn from a larger history through dilated convolution, thereby capturing longer dependencies. Among them, the dilated convolution achieves a wider input coverage by adding a constant interval between each filter element. The specific formula is as follows:

[0106] x t-(k-1)d , …x t-2d , x t-d , x t

[0107] Wherein, d is the dilation factor and k is the filter size. A larger d value allows the temporal convolutional neural network to have a larger receptive field, that is, the second output sequence is suitable for a longer receptive field than the first output sequence obtained in step 401, which is very effective for processing transient current signals.

[0108] 403. Perform a residual connection based on the first output sequence and the second output sequence to obtain a residual connection result;

[0109] 404. After stacking the residual connection results, a fully connected layer is used to perform fault classification to obtain the model training result.

[0110] In the embodiment of the present invention, the current execution end performs residual connection based on the first output sequence obtained in step 401 and the second output sequence obtained in step 402 to obtain a residual connection result. Among them, the main advantage of residual connection is that it allows the gradient to propagate directly through the identity path, thereby alleviating the gradient vanishing problem, allowing the network to be deeper, thereby improving the performance of the model. The specific formula of residual connection is as follows:

[0111] y=F(x)+x

[0112] Among them, x is the input, F(x) is the nonlinear transformation part of the network, and y is the output. The current execution end stacks the residual connection results and uses the fully connected layer to perform fault classification to obtain the model training result.

[0113] The embodiment of the present invention provides a method for diagnosing faults of a power transmission line. Compared with the prior art, the present invention uses a discrete wavelet transform method to perform feature extraction processing on a transient current signal of a power transmission line containing fault information, which can not only filter out high-frequency noise in the transient current signal of the power transmission line, but also analyze and process the transient current signal of the power transmission line according to the frequency to obtain the most representative feature vector, thereby improving the model's ability to identify power transmission line faults. The present invention also uses a temporal convolutional neural network model to perform fault classification processing on the target feature vector, which not only improves the accuracy of transmission line fault classification, but also has lower memory requirements due to the use of extended convolution. In addition, the temporal convolutional neural network model uses a parallel processing method to improve computational efficiency, thereby improving the real-time performance of power transmission line fault diagnosis.

[0114] As the above Figure 1 The embodiment of the present invention provides a fault diagnosis device for a power transmission line, such as Figure 5 As shown, the device comprises:

[0115] An acquisition module 51 is used to acquire a transient current signal of a power transmission line containing fault information;

[0116] A feature extraction module 52 is used to perform feature extraction processing on the transient current signal of the transmission line using a discrete wavelet transform method to obtain a target feature vector for fault analysis;

[0117] The fault diagnosis module 53 is used to perform fault classification processing on the target feature vector using a temporal convolutional neural network model to obtain a transmission line fault type.

[0118] Furthermore, the feature extraction module 52 includes a hierarchical acquisition unit, a decomposition unit, and a selection unit;

[0119] The level acquisition unit is used to acquire the decomposition level of the discrete wavelet transform method;

[0120] The decomposition unit is used to decompose the transient current signal of the transmission line layer by layer according to the decomposition level to obtain a signal decomposition result corresponding to each decomposition level;

[0121] The selection unit is used to perform selection processing on the signal decomposition result to obtain the target feature vector.

[0122] Furthermore, the decomposition unit is also used for:

[0123] Using a high-pass filter and a low-pass filter to perform a first-level decomposition process on the transient current signal of the transmission line to obtain a first detail component and a first approximate component;

[0124] Using the high-pass filter and the low-pass filter to perform a second-level decomposition process on the first approximate component to obtain a second detail component and a second approximate component;

[0125] The high-pass filter and the low-pass filter are used to perform layer-by-layer decomposition processing on the second approximate component until all decomposition levels are decomposed, thereby obtaining the signal decomposition result corresponding to each decomposition level.

[0126] Furthermore, the selection unit is also used for:

[0127] When selecting and processing the signal decomposition result, one of the energy method, entropy method and statistical feature method is adopted.

[0128] Furthermore, the device also includes a model training module, and the model training module is used to:

[0129] Acquire a signal sample set of a transient current signal of a transmission line, and use the discrete wavelet transform to perform feature extraction processing on each of the transient current signals in the signal sample set to obtain a feature vector sample set;

[0130] Using the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result;

[0131] Based on the model training results, the fault classification performance of the temporal convolutional neural network model is evaluated to obtain a temporal convolutional neural network model that meets the evaluation requirements.

[0132] Furthermore, the model training module is also used for:

[0133] Performing causal convolution processing on the feature vector samples as an input sequence to obtain a first output sequence; the first output sequence is offset from the input sequence by a step in time;

[0134] Performing dilated convolution processing on the input sequence to obtain a second output sequence; the second output sequence is suitable for a longer receiving field than the first output sequence;

[0135] Performing a residual connection based on the first output sequence and the second output sequence to obtain a residual connection result;

[0136] After stacking the residual connection results, a fully connected layer is used to perform fault classification to obtain the model training result.

[0137] Furthermore, the model training module is also used for:

[0138] Based on the model training result, at least one of the confusion matrix, precision, recall rate and F1 score is used to evaluate the fault classification performance of the temporal convolutional neural network model.

[0139] The embodiment of the present invention provides a fault diagnosis device for a power transmission line. Compared with the prior art, the present invention uses a discrete wavelet transform method to perform feature extraction processing on the transient current signal of the power transmission line containing fault information, which can not only filter out the high-frequency noise in the transient current signal of the power transmission line, but also analyze and process the transient current signal of the power transmission line according to the frequency to obtain the most representative feature vector, thereby improving the model's recognition ability for power transmission line faults. The present invention also uses a time convolutional neural network model to perform fault classification processing on the target feature vector, which not only improves the accuracy of transmission line fault classification, but also has lower memory requirements due to the use of extended convolution. In addition, the time convolutional neural network model uses a parallel processing method to improve the calculation efficiency, thereby improving the real-time performance of power transmission line fault diagnosis.

[0140] According to an embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the power transmission line fault diagnosis method in any of the above method embodiments.

[0141] Figure 6 A schematic diagram of the structure of a computer device provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computer device.

[0142] like Figure 6 As shown, the computer device may include: a processor (processor) 602 , a communication interface (Communications Interface) 604 , a memory (memory) 606 , and a communication bus 608 .

[0143] The processor 602 , the communication interface 604 , and the memory 606 communicate with each other via a communication bus 608 .

[0144] The communication interface 604 is used to communicate with other devices such as clients or other servers.

[0145] The processor 602 is used to execute the program 610, and specifically can execute the relevant steps of the above-mentioned power transmission line fault diagnosis method.

[0146] Specifically, the program 610 may include program codes, which include computer operation instructions.

[0147] The processor 602 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computer device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0148] The memory 606 is used to store the program 610. The memory 606 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0149] The program 610 may be specifically configured to enable the processor 602 to perform the following operations:

[0150] Acquire the transient current signal of the transmission line containing fault information;

[0151] Using discrete wavelet transform method to perform feature extraction processing on the transient current signal of the transmission line to obtain a target feature vector for fault analysis;

[0152] A temporal convolutional neural network model is used to perform fault classification processing on the target feature vector to obtain the transmission line fault type.

[0153] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for diagnosing a fault in a power transmission line, characterized in that: include: Acquire the transient current signal of the transmission line containing fault information; Using discrete wavelet transform method to perform feature extraction processing on the transient current signal of the transmission line to obtain a target feature vector for fault analysis; A temporal convolutional neural network model is used to perform fault classification processing on the target feature vector to obtain the transmission line fault type.

2. The method according to claim 1, characterized in that: The method of using discrete wavelet transform to perform feature extraction processing on the transient current signal of the transmission line to obtain a target feature vector for fault analysis includes: Obtaining a decomposition level of the discrete wavelet transform method; Decomposing the transient current signal of the transmission line layer by layer according to the decomposition level to obtain a signal decomposition result corresponding to each decomposition level; The signal decomposition result is selectively processed to obtain the target feature vector.

3. The method according to claim 2, characterized in that The step of decomposing the transient current signal of the transmission line layer by layer according to the decomposition level to obtain a signal decomposition result corresponding to each decomposition level includes: Using a high-pass filter and a low-pass filter to perform a first-level decomposition process on the transient current signal of the transmission line to obtain a first detail component and a first approximate component; Using the high-pass filter and the low-pass filter to perform a second-level decomposition process on the first approximate component to obtain a second detail component and a second approximate component; The high-pass filter and the low-pass filter are used to perform layer-by-layer decomposition processing on the second approximate component until all decomposition levels are decomposed, thereby obtaining the signal decomposition result corresponding to each decomposition level.

4. The method according to claim 2, characterized in that: When the signal decomposition result is selected and processed, one of an energy method, an entropy method and a statistical feature method is adopted.

5. The method according to claim 1, characterized in that Before the time convolution neural network model is used to perform fault classification processing on the target feature vector, the method further includes: Acquire a signal sample set of a transient current signal of a transmission line, and use the discrete wavelet transform to perform feature extraction processing on each of the transient current signals in the signal sample set to obtain a feature vector sample set; Using the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result; Based on the model training results, the fault classification performance of the temporal convolutional neural network model is evaluated to obtain a temporal convolutional neural network model that meets the evaluation requirements.

6. The method according to claim 5, characterized in that The method of using the feature vector sample set to perform model training processing on the temporal convolutional neural network model to obtain a model training result includes: Performing causal convolution processing on the feature vector samples as an input sequence to obtain a first output sequence; the first output sequence is offset from the input sequence by a step in time; Performing dilated convolution processing on the input sequence to obtain a second output sequence; the second output sequence is suitable for a longer receiving field than the first output sequence; Performing a residual connection based on the first output sequence and the second output sequence to obtain a residual connection result; After stacking the residual connection results, a fully connected layer is used to perform fault classification to obtain the model training result.

7. The method according to claim 5, characterized in that The evaluating and processing of the fault classification performance of the temporal convolutional neural network model based on the model training result includes: Based on the model training result, at least one of the confusion matrix, precision, recall rate and F1 score is used to evaluate the fault classification performance of the temporal convolutional neural network model.

8. A fault diagnosis device for a power transmission line, characterized in that: include: An acquisition module, used for acquiring a transient current signal of a power transmission line containing fault information; A feature extraction module, used to perform feature extraction processing on the transient current signal of the transmission line using a discrete wavelet transform method to obtain a target feature vector for fault analysis; The fault diagnosis module is used to perform fault classification processing on the target feature vector using a temporal convolutional neural network model to obtain a transmission line fault type.

9. A storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction executes an operation corresponding to the fault diagnosis method for a power transmission line according to any one of claims 1 to 7.

10. A computer device, comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the fault diagnosis method for a power transmission line according to any one of claims 1-7.

Images