A method and system for detecting faults in distribution network lines based on traveling waves

Abstract

This invention provides a method and system for detecting faults in distribution network lines based on traveling waves. The method includes: triggering a traveling wave test with a preset step size starting from a traveling wave detection device; the traveling wave detection device acquiring and recording the received traveling wave signal; performing a Fast Fourier Transform (FFT) on the traveling wave signal to obtain a traveling wave frequency domain signal; clustering the traveling wave frequency domain signal; inputting the clustered signal into a first neural network for training; inputting the clustered signal of each category into a second neural network for training; when a fault occurs, the detection device acquires the traveling wave signal, performs a FFT on the traveling wave signal and clusters it to obtain a fault clustering signal; inputting the fault clustering signal into a line segment classification model to obtain a fault line segment classification; obtaining a corresponding distance calculation model based on the fault line segment classification; and inputting the fault clustering signal into the distance calculation model to obtain the fault distance. This solution solves the technical problem of difficult traveling wave fault location in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of power, and more specifically to a method and system for detecting faults in distribution network lines based on traveling waves. Background Technology

[0002] In power transmission networks, when a line fault occurs, a traveling wave propagates from the fault point along the line to all parts of the system. The traveling wave method utilizes instrument transformers to extract the traveling wave signal for fault location. Existing technologies for locating single-type power grid traveling wave faults include... Figure 2 As shown, when a fault occurs at point F, a traveling wave will propagate from point F to points a and b. The arrival time of the traveling wave can be measured at point a. After the traveling wave reaches point b, it reflects back to point a, where the arrival time of the second traveling wave can be measured. The fault point can be located by using the two time differences and the propagation speed of the traveling wave.

[0003] In the past, power distribution networks were mostly constructed using single overhead lines, primarily for transmitting electricity within and outside the city. These lines were characterized by low construction costs, short construction periods, and ease of expansion. However, in recent years, with the development of power grid technology and the in-depth implementation of power grid renovation projects, underground cables have been widely used in cities to beautify the urban landscape and save land resources. Compared to overhead lines built outside the city, cable lines are generally not in direct contact with the outside world, being buried deep underground or laid indoors. This not only beautifies the urban environment and reduces safety hazards but also provides stable transmission performance unaffected by external weather and climate. For power distribution lines outside the city, considering both economic efficiency and practicality, overhead lines suitable for long-distance power transmission and easy expansion are adopted. This has led to the gradual development of power distribution networks from single overhead lines to a hybrid overhead line-cable system.

[0004] For overhead line-cable hybrid lines, traveling waves at the junction of the overhead line and the cable will exhibit phenomena such as refraction and reflection. For example... Figure 3 As shown, when a fault occurs at point F, the traveling wave will originate from point F and undergo refraction and reflection at points c and d. The superposition of multiple refracted and reflected signals results in a chaotic traveling wave signal received at point a. Existing technologies typically use complex signal decomposition algorithms to determine the propagation characteristics of the traveling wave signal and then locate the fault. Existing technologies for traveling wave fault detection in overhead line-cable hybrid lines have the problems of complex processes and high precision requirements for detection equipment. Summary of the Invention

[0005] To address the problems mentioned above, this invention provides a method and system for detecting faults in power distribution lines based on traveling waves. The method employs a neural network to analyze the frequency domain characteristics of traveling waves, thereby reducing the difficulty of fault detection using traveling waves.

[0006] In one aspect of the present invention, a method for detecting faults in a distribution network line based on traveling waves is provided, comprising: setting a traveling wave detection device in the distribution network line to be detected; triggering a test traveling wave excitation from the traveling wave detection device with a preset step size; the traveling wave detection device acquiring and recording the received traveling wave signal and the corresponding test traveling wave excitation trigger position; performing a fast Fourier transform on the traveling wave signal to obtain a traveling wave frequency domain signal; clustering the traveling wave frequency domain signal to obtain a cluster signal, wherein the cluster signal is a frequency sequence composed of cluster centers; adding line segment labels and distance labels to the cluster signal according to the test traveling wave excitation trigger position; and storing the cluster signal and... Line segment labels are input into a first neural network for training to obtain a line segment classification model. The clustered signals are then classified by line segment, and the clustered signals of each category, along with distance labels, are input into a second neural network for training to obtain a distance calculation model corresponding to each line segment category. When a fault occurs, a wave detection device collects traveling wave signals, performs Fast Fourier Transform and clustering on the traveling wave signals to obtain fault clustering signals. The fault clustering signals are input into the line segment classification model to obtain fault line segment categories. Based on the fault line segment categories, corresponding distance calculation models are obtained. Finally, the fault clustering signals are input into the corresponding distance calculation models to obtain the fault distance.

[0007] Furthermore, the line is a hybrid overhead line-cable line.

[0008] Furthermore, during clustering, signal clusters with amplitudes less than a second preset value and / or signal clusters with frequency domain widths less than a third preset value are removed.

[0009] Furthermore, the first neural network is a DNN network.

[0010] Furthermore, the second neural network is a backpropagation (BP) neural network.

[0011] Another method of the present invention also provides a fault detection system for distribution network lines based on traveling waves. The distribution network line includes a traveling wave detection device. The system comprises: an excitation module for triggering a test traveling wave excitation with a preset step size starting from the traveling wave detection device; an acquisition module for the traveling wave detection device to acquire and record the received traveling wave signal and the corresponding test traveling wave excitation trigger position; a first transformation module for performing a fast Fourier transform on the traveling wave signal to obtain a traveling wave frequency domain signal; a second transformation module for clustering the traveling wave frequency domain signal to obtain a cluster signal, wherein the cluster signal is a frequency sequence composed of cluster centers; a third transformation module for adding line segment labels and distance labels to the cluster signal according to the test traveling wave excitation trigger position; and a first training module. The system includes a first neural network for training, which inputs the clustering signals and line segment labels to obtain a line segment classification model; a second training module for classifying the clustering signals by line segments, inputting the clustering signals and distance labels of each category into the second neural network for training, and obtaining a distance calculation model corresponding to each line segment category; a fault signal detection module for collecting traveling wave signals by a wave detection device when a fault occurs, performing fast Fourier transform and clustering on the traveling wave signals to obtain fault clustering signals; a classification module for inputting the fault clustering signals into the line segment classification model to obtain fault line segment classifications; and a calculation module for obtaining the corresponding distance calculation model based on the fault line segment classifications, inputting the fault clustering signals into the corresponding distance calculation model to obtain the fault distance.

[0012] Furthermore, the line is a hybrid overhead line-cable line.

[0013] Furthermore, during clustering, signal clusters with amplitudes less than a second preset value and / or signal clusters with frequency domain widths less than a third preset value are removed.

[0014] Furthermore, the first neural network is a DNN network.

[0015] Furthermore, the second neural network is a backpropagation (BP) neural network.

[0016] The present invention uses the above-mentioned technical solution to analyze the frequency domain characteristics of traveling waves through neural networks, thereby reducing the difficulty of fault detection through traveling waves. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 Schematic diagram of the method of this invention;

[0019] Figure 2 Schematic diagram of traveling wave;

[0020] Figure 3 Schematic diagram of traveling wave in a hybrid circuit;

[0021] Figure 4 Schematic diagram of traveling wave frequency domain decomposition;

[0022] Figure 5 Schematic diagram of interference wave. Detailed Implementation

[0023] The invention will now be described in preferred form with reference to the accompanying drawings and specific embodiments.

[0024] This embodiment solves the above problem through the following steps:

[0025] In one embodiment, reference Figure 1 This invention provides a method for detecting faults in distribution network lines based on traveling waves, comprising the following steps:

[0026] A traveling wave detection device is installed in the power distribution network line to be tested.

[0027] The distribution network to be tested can be a section of power line requiring maintenance, a local exchange power line, or a power line connecting several transformers. The traveling wave detection device can detect traveling waves in the power line; conventional traveling wave detection devices can be used to obtain basic traveling wave parameters such as frequency, amplitude, and duration. Furthermore, the traveling wave detection device can be installed at one end of the test line, such as... Figure 2 As shown, the traveling wave detection device can be set at point a.

[0028] The traveling wave excitation is triggered by the traveling wave detection device with a preset step size.

[0029] In subsequent steps, this invention requires a neural network to process the traveling wave data, thus necessitating the acquisition of a large amount of training data. To obtain this training data, a test traveling wave excitation is triggered beforehand from the traveling wave detection device with a preset step size; exemplarily, such as... Figure 2 As shown, if the traveling wave detection device is set at point a, a test traveling wave will be triggered at intervals from point a. The specific length can be set according to the line length and accuracy requirements. For example, a test traveling wave excitation can be triggered every 1 km. The test traveling wave excitation can be triggered by any existing technology, such as using a shock wave device or artificially creating a line fault to generate a traveling wave.

[0030] The traveling wave detection device collects and records the received traveling wave signal and the corresponding test traveling wave excitation trigger position.

[0031] After the traveling wave excitation is triggered, the traveling wave detection device can detect the corresponding traveling wave signal. In order to correspond the traveling wave signal with the trigger position, it is also necessary to record the corresponding position. The corresponding position can be recorded in any form such as coordinates. Preferredly, the trigger position of the traveling wave excitation is recorded by the relative distance to the traveling wave detection device. For example, Table 1 shows several traveling wave signal records.

[0032] Table 1

[0033] traveling wave signal Location Traveling wave signal 1 1km Traveling wave signal 2 2km Traveling wave signal 3 8km

[0034] The traveling wave signal is subjected to a fast Fourier transform to obtain the traveling wave frequency domain signal.

[0035] In hybrid networks, traveling wave signals are composed of a mixture of various clutter signals, such as... Figure 4 As shown, from the frequency domain perspective, the signal distribution has relatively obvious discrete characteristics. Since the traveling waves generated at different locations have different degrees of reflection and refraction through different connection points, the generated clutter characteristics are also different, and the distribution pattern from the frequency domain perspective is also different. Based on this principle, the present invention performs subsequent detection steps. Therefore, after obtaining the traveling wave signal, the obtained signal is subjected to a fast Fourier transform to obtain the traveling wave frequency domain signal.

[0036] Clustering is performed on the traveling wave frequency domain signal to obtain a cluster signal, which is a frequency sequence composed of cluster centers;

[0037] The frequency domain of a traveling wave consists of clusters of signals at different frequencies. To extract cluster features, each cluster is clustered, and cluster centers are obtained. For example, Figure 3 The signal shown consists of four signal clusters. After clustering, the cluster centers obtained are 55, 110, 230, and 310, respectively. The frequency sequence of this signal is (55, 110, 230, 310), which is the cluster signal of the traveling wave frequency domain signal.

[0038] Furthermore, to reduce interference, signal clusters with amplitudes smaller than a second preset value are removed. Additionally, signal clusters with frequency bandwidths smaller than a third preset value can also be removed. These two methods for removing interference signals can be used in any combination. For example... Figure 5 As shown, the signal near frequency 150 (represented by the dashed line) has too small an amplitude and can be deleted; the signal near frequency 250 (represented by the dashed line) has too small a frequency domain width and can be deleted.

[0039] Furthermore, the longest length of all frequency sequences is obtained, and frequency sequences shorter than the longest length are padded with 0s to reach the longest length; for example, if the longest length of all frequency sequences is 6, then (55, 110, 230, 310) is padded to (55, 110, 230, 310, 0, 0).

[0040] Add line segment labels and distance labels to the clustered signals based on the test traveling wave excitation trigger position;

[0041] As verified by the inventors, if distance and clustering signals are used directly for training, the training load is very large and it is very difficult to converge. In order to solve this problem, this invention performs segmented training on the clustering signals. Therefore, it is necessary to first obtain the line segment label and distance label of each clustering signal.

[0042] For example, such as Figure 2 As shown, it includes three line segments, denoted as ac, cd, and db respectively. Since each cluster signal corresponds to a test excitation position, it corresponds to one of the line segments and a distance value from the traveling wave detection device. Table 2 shows some exemplary data.

[0043] Table 2

[0044] Clustering signal Line section distance 55，110，230，310，0，0 ac 1km 50，100，210，250，265，0 ac 2km 54，103，200，230，235，0 cd 8km 30，40，159，180，231，267 db 18km

[0045] The clustering signals and line segment labels are input into the first neural network for training to obtain a line segment classification model.

[0046] Adding line segment labels to the clustered signal transforms it into a regular data classification sample. The clustered signal consists of a frequency sequence, i.e., one-dimensional data, which can be classified using a conventional DNN network. Of course, any other classification network can also be used; this invention does not impose any particular limitations. After obtaining the line segment classification model, inputting the clustered signal into the model will yield the line segment classification. For example, after inputting a set of data, the output line segment is ac.

[0047] The clustering signals are classified by line segment. The clustering signals of each category and the distance label are input into the second neural network for training to obtain the distance calculation model corresponding to each line segment category.

[0048] Because the influencing factors of traveling waves generated when a fault occurs on each line segment are different, and the factors affecting distance are superimposed with the factors affecting nodes of different line segments, it is difficult to fit a reasonable function to represent the global situation, and the training cannot converge effectively. Therefore, this invention classifies the clustered signals by line segment. Figure 2Taking the line segment as an example, all clustering signals are divided into several categories: ac, cd, and db, and each sample has a distance label. The purpose of training the model is to output distance values ​​by inputting the clustering signals, which is essentially fitting a function. Therefore, the second neural network in this invention uses a backpropagation (BP) neural network. Since it is a one-dimensional data fitting, a conventional BP neural network is sufficient. A BP neural network model is trained for each line segment classification (ac, cd, db). The distance value is output after inputting the clustering signal of the corresponding line segment classification into the corresponding model.

[0049] When a fault occurs, the wave detection device collects traveling wave signals, performs fast Fourier transform and clustering on the traveling wave signals to obtain fault clustering signals.

[0050] The aforementioned steps prepare the complete model for computation. When a fault occurs, the wave detection device detects the traveling wave signal online and performs Fast Fourier Transform and clustering using the same method as in the training phase to obtain the fault cluster signal. Both the fault cluster signal and the aforementioned cluster signal are frequency sequences composed of cluster centers. The specific transformation process is the same as in the training phase and will not be described in detail here.

[0051] The fault clustering signal is input into the line segment classification model to obtain the fault line segment classification.

[0052] The line segment classification model can output line segment classifications. When a fault occurs, the fault signal is processed into the same format as during the training phase and input into the line segment classification model to obtain the faulty line segment classification. Figure 2 Taking the line as an example, the output is one of the following types: ac, cd, db.

[0053] Based on the classification of the faulty line segments, a corresponding distance calculation model is obtained;

[0054] During the model training phase, a distance calculation model corresponding to each line segment classification was obtained. Therefore, once the obstacle line segment classification is obtained, the corresponding distance calculation model can be obtained. Figure 2 Taking the line as an example, we can obtain the distance calculation model corresponding to one of the line categories ac, cd, and db.

[0055] The fault clustering signal is input into the corresponding distance calculation model to obtain the fault distance.

[0056] The output of the distance calculation model is the fault distance. In the aforementioned steps, the detected fault signal is processed into the same data format as in the training phase. Therefore, the processed fault clustering signal can be directly input into the corresponding distance calculation model to obtain the fault distance, thereby detecting the specific location of the fault.

[0057] Another aspect of the present invention discloses a fault detection system for distribution network lines based on traveling waves, wherein the distribution network line includes a traveling wave detection device, characterized in that the system comprises:

[0058] The excitation module is used to trigger the test traveling wave excitation from the traveling wave detection device with a preset step size;

[0059] The acquisition module is used by the traveling wave detection device to acquire and record the received traveling wave signal and the corresponding test traveling wave excitation trigger position;

[0060] The first transformation module is used to perform a fast Fourier transform on the traveling wave signal to obtain a traveling wave frequency domain signal;

[0061] The second transformation module is used to cluster the traveling wave frequency domain signal to obtain a cluster signal, wherein the cluster signal is a frequency sequence composed of cluster centers;

[0062] The third transformation module is used to add line segment labels and distance labels to the clustered signal according to the test traveling wave excitation trigger position;

[0063] The first training module is used to input the clustering signal and the line segment label into the first neural network for training to obtain the line segment classification model.

[0064] The second training module is used to classify the clustered signals by line segment, and input the clustered signals of each category and the distance label into the second neural network for training to obtain the distance calculation model corresponding to each line segment category.

[0065] The fault signal detection module is used to collect traveling wave signals when a fault occurs, perform fast Fourier transform and clustering on the traveling wave signals to obtain fault clustering signals.

[0066] The classification module is used to input the fault clustering signal into the line segment classification model to obtain the fault line segment classification;

[0067] The calculation module is used to obtain the corresponding distance calculation model according to the classification of the faulty line segment, input the fault clustering signal into the corresponding distance calculation model, and obtain the fault distance.

[0068] It should be noted that the detailed implementation principle and further improvement measures of the above-mentioned traveling wave-based distribution network line fault detection system are the same as those of the aforementioned traveling wave-based distribution network line fault detection method. They will not be described in detail in this embodiment. Those skilled in the art can implement the traveling wave-based distribution network line fault detection method in the traveling wave-based distribution network line fault detection system based on the existing technology.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0070] For any module structures not specifically defined in this invention, the existing technical specifications shall prevail. The existing technical specifications mentioned in the foregoing background and specific embodiments sections are considered part of this invention and are used to understand the meaning of certain technical features or parameters. The scope of protection of this invention is determined by the actual contents of the claims.

Claims

1. A method of line fault detection for a power distribution network based on travelling waves, characterized by The method comprises: setting a traveling wave detection device in a power distribution network line to be detected; triggering a test traveling wave excitation at a preset step length from the traveling wave detection device; the traveling wave detection device collecting and recording the received traveling wave signal and the corresponding test traveling wave excitation trigger position; performing fast Fourier transform on the traveling wave signal to obtain a traveling wave frequency domain signal; performing clustering on the traveling wave frequency domain signal to obtain a clustered signal, the clustered signal being a frequency sequence composed of clustering centers; adding a line segment label and a distance label to the clustered signal according to the test traveling wave excitation trigger position; inputting the clustered signal and the line segment label into a first neural network for training to obtain a line segment classification model; classifying the clustered signal by line segment, inputting the clustered signal and the distance label of each classification into a second neural network for training to obtain a distance calculation model corresponding to each line segment classification; when a fault occurs, the traveling wave detection device collects a traveling wave signal, performs fast Fourier transform and clustering on the traveling wave signal to obtain a fault clustered signal; inputting the fault clustered signal into the line segment classification model to obtain a fault line segment classification; obtaining a corresponding distance calculation model according to the fault line segment classification; inputting the fault clustered signal into the corresponding distance calculation model to obtain a fault distance.

2. The method of claim 1, wherein: The line is an overhead line-cable hybrid line.

3. The method of claim 1, wherein: When clustering, signal clusters with amplitudes less than a second preset value and / or signal clusters with frequency domain widths less than a third preset value are removed.

4. The method of claim 1, wherein: The first neural network is a DNN network.

5. The method of claim 1, wherein: The second neural network is a BP neural network.

6. A power distribution network line fault detection system based on traveling wave, said power distribution network line including a traveling wave detection device therein, characterized by The system comprises: an excitation module for triggering a test traveling wave excitation at a preset step length from the traveling wave detection device; an acquisition module for the traveling wave detection device to collect and record the received traveling wave signal and the corresponding test traveling wave excitation trigger position; a first transform module for performing fast Fourier transform on the traveling wave signal to obtain a traveling wave frequency domain signal; a second transform module for performing clustering on the traveling wave frequency domain signal to obtain a clustered signal, the clustered signal being a frequency sequence composed of clustering centers; a third transform module for adding a line segment label and a distance label to the clustered signal according to the test traveling wave excitation trigger position; a first training module for inputting the clustered signal and the line segment label into a first neural network for training to obtain a line segment classification model; a second training module for classifying the clustered signal by line segment, inputting the clustered signal and the distance label of each classification into a second neural network for training to obtain a distance calculation model corresponding to each line segment classification; a fault signal detection module for, when a fault occurs, the traveling wave detection device collecting a traveling wave signal, performing fast Fourier transform and clustering on the traveling wave signal to obtain a fault clustered signal; a classification module for inputting the fault clustered signal into the line segment classification model to obtain a fault line segment classification; a calculation module for obtaining a corresponding distance calculation model according to the fault line segment classification, inputting the fault clustered signal into the corresponding distance calculation model to obtain a fault distance.

7. The traveling wave based power distribution network line fault detection system of claim 6, wherein: The line is an overhead line-cable hybrid line.

8. The traveling wave based power distribution network line fault detection system of claim 6, wherein: The signal clusters with amplitudes less than a second preset value and / or the signal clusters with frequency domain widths less than a third preset value are removed when clustering.

9. The traveling wave based power distribution network line fault detection system of claim 6, wherein: The first neural network is a DNN network.

10. The traveling wave based power distribution network line fault detection system of claim 6, wherein: The second neural network is a BP neural network.

Images