Intelligent monitoring method for grounding protection of feeder cables in electrified railways

By installing current sensors on the feeder cables of electrified railways and using deep learning technology for time-frequency analysis, the problems of insufficient fixed threshold and slow response speed in traditional methods have been solved, realizing the accuracy and reliability of cable grounding protection and ensuring the safe operation of electrified railways.

CN118759417BActive Publication Date: 2024-11-15SI CHUAN SHU DAO DIAN QI HUA JIAN SHE YOU XIAN GONG SI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411254535.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-09
Publication Date
2024-11-15
Estimated Expiration
2044-09-09

AI Technical Summary

Technical Problem

Traditional grounding protection methods for feeder cables in electrified railways lack flexibility due to fixed thresholds, making it difficult to identify complex or hidden faults. Furthermore, delays in monitoring and data processing result in slow response times and an inability to take timely measures.

Method used

By installing current sensors on the shielding layer of railway feeder cables, current monitoring signals are collected in real time and transmitted to the monitoring platform via a wireless communication module. Deep learning technology is used for time-frequency analysis and clustering to identify abnormal patterns in the current signals.

Benefits of technology

It has improved the accuracy and reliability of cable grounding protection, ensuring the safe operation of electrified railways and improving fault identification and response speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118759417B_ABST
    Figure CN118759417B_ABST
Patent Text Reader

Abstract

The present application relates to the field of intelligent monitoring, and provides an intelligent monitoring method for grounding protection of feeder cables of electrified railways, which collects current monitoring signals through current sensors installed on the grounding leads of the shielding layer of the railway feeder cables, and transmits the current monitoring signals to a monitoring platform through a wireless communication module, and uses deep learning-based signal processing and analysis technology in the monitoring platform to perform time-frequency analysis and correlation of the current monitoring signals, thereby automatically obtaining monitoring results of whether there are abnormalities based on the clustering representation features of the pixel granularity of the multiple current signals in the global time-frequency. In this way, the current monitoring signals can be collected and transmitted in real time, and the abnormal patterns in the signals can be accurately identified and analyzed using deep learning technology, thereby ensuring the accuracy of monitoring, thereby improving the reliability of cable grounding protection and ensuring the safe operation of electrified railways.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring, and more specifically, to an intelligent monitoring method for grounding protection of feeder cables in electrified railways. Background Art

[0002] The grounding protection of feeder cables in electrified railways is an important part of ensuring the safe operation of railway traction power supply systems. Specifically, in electrified railway systems, feeder cables are the key link in power transmission, and their grounding protection is crucial to the safe operation of cables. Therefore, grounding protection monitoring of feeder cables in electrified railways is an important technical means, which is mainly to reduce faults in cable operation, prevent excessive induced voltage in the cable sheath, thereby protecting the cable from damage and ensuring the reliability of traction power supply.

[0003] However, traditional methods often set fixed thresholds to determine the cable grounding protection status. However, since fixed thresholds are usually set based on ideal or general conditions, and the actual operating environment may vary due to factors such as season, weather, and temperature, such fixed thresholds lack flexibility and are difficult to identify complex or hidden fault modes. In addition, due to delays in monitoring and data processing, traditional methods have a slow response speed after discovering a problem, and cannot take timely measures, which leads to the expansion of faults.

[0004] Therefore, an intelligent monitoring method for grounding protection of feeder cables in electrified railways is desired. Summary of the invention

[0005] In view of the shortcomings in the prior art, the present application provides an intelligent monitoring method for grounding protection of feeder cables in electrified railways, which collects current monitoring signals through current sensors installed on the grounding leads of the shielding layer of the railway feeder cables, and transmits the current monitoring signals to a monitoring platform through a wireless communication module, in which signal processing and analysis techniques based on deep learning are used to perform time-frequency analysis and correlation of the current monitoring signals, thereby automatically obtaining monitoring results of whether there are abnormalities based on the clustering representation features of the pixel granularity of the multiple current signals in the global time-frequency. In this way, current monitoring signals can be collected and transmitted in real time, and deep learning techniques can be used to accurately identify and analyze abnormal patterns in the signals, thereby ensuring the accuracy of monitoring, thereby improving the reliability of cable grounding protection and ensuring the safe operation of electrified railways.

[0006] According to one aspect of the present application, there is provided an intelligent monitoring method for grounding protection of feeder cables of an electrified railway, which comprises:

[0007] Acquiring a current monitoring signal collected by a current sensor, wherein the current sensor is installed on a ground lead of a shielding layer of a railway feeder cable;

[0008] Transmitting the current monitoring signal to a monitoring platform via a wireless communication module;

[0009] On the monitoring platform, performing time-frequency conversion on the current monitoring signal to obtain a two-dimensional time-frequency graph of the current signal;

[0010] On the monitoring platform, extracting the time-frequency features of the current signal from the two-dimensional time-frequency graph of the current signal to obtain a time-frequency correlation feature graph of the current signal;

[0011] Decentralize the time-frequency correlation feature map of the current signal along the channel dimension to obtain a set of time-frequency correlation feature vectors of the current signal pixel granularity;

[0012] Inputting the set of the current signal pixel granularity time-frequency correlation feature vectors into a feature distribution reverse mask clustering analysis network to obtain a current signal global time-frequency clustering representation vector as a current signal global time-frequency clustering representation feature;

[0013] Based on the global time-frequency clustering representation characteristics of the current signal, a monitoring result is obtained, and the monitoring result is used to indicate whether there is an abnormality.

[0014] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables in electrified railways, the current monitoring signal is subjected to time-frequency conversion to obtain a two-dimensional time-frequency diagram of the current signal, including: performing a wavelet transform on the current monitoring signal to obtain a two-dimensional time-frequency diagram of the current signal.

[0015] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables in electrified railways, the two-dimensional time-frequency diagram of the current signal is subjected to time-frequency feature extraction of the current signal to obtain a time-frequency correlation feature diagram of the current signal, including: inputting the two-dimensional time-frequency diagram of the current signal into a current signal time-frequency feature extractor based on a void convolutional neural network model to obtain the time-frequency correlation feature diagram of the current signal.

[0016] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables in electrified railways, the set of the current signal pixel granularity time-frequency correlation feature vectors is input into the feature distribution reverse mask clustering analysis network to obtain the global time-frequency clustering representation vector of the current signal, including: clustering the set of the current signal pixel granularity time-frequency correlation feature vectors to obtain the current signal pixel granularity time-frequency correlation cluster center vector; calculating the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of the current signal pixel granularity time-frequency correlation feature vector and the current signal pixel granularity time-frequency correlation cluster center vector to obtain the set of current signal pixel granularity time-frequency correlation coefficients; taking the inverse of each current signal pixel granularity time-frequency correlation coefficient in the set of the current signal pixel granularity time-frequency correlation coefficient to obtain the set of current signal pixel granularity time-frequency anti-correlation coefficients; using the Sigmoid function to normalize the set of the current signal pixel granularity time-frequency anti-correlation coefficients Processing to obtain a set of current signal pixel granularity time-frequency anti-correlation weight coefficients; using each current signal pixel granularity time-frequency anti-correlation weight coefficient in the set of current signal pixel granularity time-frequency anti-correlation weight coefficients as a weight, weighting each current signal pixel granularity time-frequency association feature vector in the set of current signal pixel granularity time-frequency association feature vectors to obtain a set of suppressed current signal pixel granularity time-frequency association feature vectors; calculating the positional difference between each group of corresponding current signal pixel granularity time-frequency association feature vectors and the suppressed current signal pixel granularity time-frequency association feature vectors in the set of current signal pixel granularity time-frequency association feature vectors to obtain a set of enhanced current signal pixel granularity time-frequency association feature vectors; calculating the positional mean vector of the set of enhanced current signal pixel granularity time-frequency association feature vectors to obtain the current signal global time-frequency clustering representation vector.

[0017] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables of electrified railways, the set of time-frequency correlation feature vectors of the current signal pixel granularity is clustered to obtain the time-frequency correlation clustering center vector of the current signal pixel granularity, including: calculating the mean vector of the set of time-frequency correlation feature vectors of the current signal pixel granularity to obtain the time-frequency correlation clustering center vector of the current signal pixel granularity.

[0018] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables in electrified railways, the correlation coefficients between each current signal pixel granularity time-frequency correlation feature vector in the set of the current signal pixel granularity time-frequency correlation feature vectors and the current signal pixel granularity time-frequency correlation cluster center vector are calculated to obtain a set of current signal pixel granularity time-frequency correlation coefficients, including: vector multiplying the transposed vector of the current signal pixel granularity time-frequency correlation feature vector with the current signal pixel granularity time-frequency correlation cluster center vector to obtain the current signal pixel granularity time-frequency correlation value; respectively calculating the current signal pixel granularity time-frequency correlation feature vectors. The norm of the associated characteristic vector and the current signal pixel granularity time-frequency associated cluster center vector is calculated to obtain the current signal pixel granularity time-frequency associated norm value and the current signal pixel granularity time-frequency associated cluster norm value; the current signal pixel granularity time-frequency associated norm value and the current signal pixel granularity time-frequency associated cluster norm value are multiplied to obtain the current signal pixel granularity time-frequency interaction value; the current signal pixel granularity time-frequency associated value is divided by the current signal pixel granularity time-frequency interaction value to obtain the current signal pixel granularity time-frequency correlation coefficient corresponding to the current signal pixel granularity time-frequency associated characteristic vector.

[0019] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables in electrified railways, a Sigmoid function is used to normalize the set of time-frequency anti-correlation coefficients of the current signal pixel granularity to obtain a set of time-frequency anti-correlation weight coefficients of the current signal pixel granularity, including: using the negative number of each current signal pixel granularity time-frequency anti-correlation coefficient in the set of time-frequency anti-correlation coefficients of the current signal pixel granularity as the exponent of the natural constant to calculate the exponential function value with the natural constant as the base according to the position to obtain a set of time-frequency anti-correlation class support coefficients of the current signal pixel granularity; calculating the reciprocal of the sum of each current signal pixel granularity time-frequency anti-correlation class support coefficient in the set of time-frequency anti-correlation class support coefficients of the current signal pixel granularity and the constant one to obtain the set of time-frequency anti-correlation weight coefficients of the current signal pixel granularity.

[0020] In the above-mentioned intelligent monitoring method for grounding protection of feeder cables in electrified railways, a monitoring result is obtained based on the global time-frequency clustering representation characteristics of the current signal, including: inputting the global time-frequency clustering representation vector of the current signal into a monitoring result generation module based on a classifier to obtain the monitoring result.

[0021] This application has significant technical effects due to the adoption of the above technical solutions:

[0022] The present application provides an intelligent monitoring method for grounding protection of feeder cables in electrified railways, which collects current monitoring signals through a current sensor installed on the grounding lead of the shielding layer of the railway feeder cable, and transmits the current monitoring signals to a monitoring platform through a wireless communication module, in which a signal processing and analysis technology based on deep learning is used to perform time-frequency analysis and correlation of the current monitoring signals, thereby automatically obtaining monitoring results of whether there are abnormalities based on the clustering representation features of the pixel granularity of the multiple current signals in the global time-frequency. In this way, the current monitoring signals can be collected and transmitted in real time, and the abnormal patterns in the signals can be accurately identified and analyzed using deep learning technology, thereby ensuring the accuracy of monitoring, thereby improving the reliability of cable grounding protection and ensuring the safe operation of electrified railways. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0024] Figure 1 The present invention is a flowchart of an intelligent monitoring method for grounding protection of feeder cables in electrified railways according to an embodiment of the present application.

[0025] Figure 2 Schematic diagram of the architecture of an intelligent monitoring method for grounding protection of feeder cables in electrified railways according to an embodiment of the present application.

[0026] Figure 3 In the intelligent monitoring method for grounding protection of feeder cables of electrified railways according to an embodiment of the present application, the set of pixel-granularity time-frequency correlation feature vectors of the current signal is input into a feature distribution inverse mask clustering analysis network to obtain a flow chart of a global time-frequency clustering representation vector of the current signal.

[0027] Figure 4 In the intelligent monitoring method for grounding protection of feeder cables of electrified railways according to an embodiment of the present application, a flowchart is provided for calculating the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of current signal pixel granularity time-frequency correlation feature vectors and the current signal pixel granularity time-frequency correlation cluster center vector to obtain a set of current signal pixel granularity time-frequency correlation coefficients.

[0028] Figure 5The present invention is a flowchart of a method for intelligent monitoring of grounding protection of feeder cables of electrified railways according to an embodiment of the present application, in which a set of time-frequency anti-correlation coefficients of pixel granularity of current signals are normalized using a Sigmoid function to obtain a set of time-frequency anti-correlation weight coefficients of pixel granularity of current signals. DETAILED DESCRIPTION

[0029] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0030] The grounding protection mechanism of feeder cables in electrified railways is a key link in ensuring the stable operation of railway traction power supply systems. In electrified railway systems, feeder cables play a central role in power transmission, and their grounding protection strategy is of decisive significance for ensuring the safe operation of cables. Therefore, monitoring the grounding protection status of feeder cables in electrified railways is an important technical measure to maintain system safety. This monitoring aims to reduce the failure rate during cable operation, avoid abnormal increase in induced voltage on the cable sheath, protect the cable from damage, and ensure the continuity and reliability of traction power supply.

[0031] However, in existing traditional monitoring methods, fixed thresholds are usually used to determine the status of cable grounding protection. These thresholds are often set based on ideal or general conditions, while the actual operating environment may vary due to seasonal changes, climate conditions, temperature fluctuations and other factors. This results in the lack of flexibility of fixed thresholds and makes it difficult to accurately identify complex or hidden fault modes. In addition, traditional monitoring methods have delays in monitoring and data processing, which makes the response speed slow after the abnormal situation is discovered, and it is impossible to take effective measures quickly, which may aggravate the spread of faults.

[0032] Therefore, in response to the above technical problems, the technical solution of the present application is to collect current monitoring signals by a current sensor installed on the grounding lead of the shielding layer of the railway feeder cable, and transmit the current monitoring signals to a monitoring platform through a wireless communication module, and use deep learning-based signal processing and analysis technology in the monitoring platform to perform time-frequency analysis and correlation of the current monitoring signals, so as to automatically obtain the monitoring results of whether there are abnormalities based on the clustering representation characteristics of the pixel granularity of the multiple current signals in the global time-frequency. In this way, the current monitoring signals can be collected and transmitted in real time, and the deep learning technology can be used to accurately identify and analyze the abnormal patterns in the signals, thereby ensuring the accuracy of the monitoring, thereby improving the reliability of the cable grounding protection and ensuring the safe operation of the electrified railway.

[0033] Figure 1The present invention is a flowchart of an intelligent monitoring method for grounding protection of feeder cables in electrified railways according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of an intelligent monitoring method for grounding protection of feeder cables in an electrified railway according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the intelligent monitoring method for grounding protection of an electrified railway feeder cable includes: S110, obtaining a current monitoring signal collected by a current sensor, wherein the current sensor is installed on the grounding lead of the shielding layer of the railway feeder cable; S120, transmitting the current monitoring signal to a monitoring platform through a wireless communication module; S130, on the monitoring platform, performing time-frequency conversion on the current monitoring signal to obtain a two-dimensional time-frequency diagram of the current signal; S140, on the monitoring platform, performing time-frequency feature conversion on the two-dimensional time-frequency diagram of the current signal Extract to obtain a time-frequency correlation feature map of the current signal; S150, feature-disperse the time-frequency correlation feature map of the current signal along the channel dimension to obtain a set of time-frequency correlation feature vectors at the pixel granularity of the current signal; S160, input the set of time-frequency correlation feature vectors at the pixel granularity of the current signal into a feature distribution inverse mask clustering analysis network to obtain a global time-frequency clustering representation vector of the current signal as a global time-frequency clustering representation feature of the current signal; S170, based on the global time-frequency clustering representation feature of the current signal, obtain a monitoring result, and the monitoring result is used to indicate whether there is an abnormality.

[0034] In step S110 and step S120, a current monitoring signal collected by a current sensor is obtained, wherein the current sensor is installed on the grounding lead of the shielding layer of the railway feeder cable, and the current monitoring signal is transmitted to the monitoring platform through a wireless communication module. It should be understood that the abnormal change of the grounding current is often a direct manifestation of the abnormal state of the cable grounding protection. By accurately monitoring the dynamic changes of the current signal, it is possible to timely identify and judge whether the cable grounding state is in a normal state. In order to achieve this monitoring goal and ensure the real-time transmission and remote access of the monitoring data, a wireless communication module is introduced in the technical solution of the present application, and the main function of the module is to wirelessly transmit the current monitoring signal from the sensor endpoint to the monitoring platform. Specifically, the wireless communication module is an advanced device that uses wireless technology means such as radio waves and infrared rays to achieve efficient transmission of data between different devices. In this way, monitoring personnel can access the real-time monitoring data of the cable anytime and anywhere, so as to promptly discover any abnormal situation and quickly take necessary countermeasures.

[0035] In step S130, on the monitoring platform, the current monitoring signal is subjected to time-frequency conversion to obtain a two-dimensional time-frequency diagram of the current signal. Specifically, in an embodiment of the present application, the current monitoring signal is subjected to time-frequency conversion to obtain a two-dimensional time-frequency diagram of the current signal, including: the current monitoring signal is subjected to wavelet transformation to obtain a two-dimensional time-frequency diagram of the current signal. It should be understood that, considering that the current monitoring signal contains features of different time scales and frequencies, that is, the local feature information existing in the time domain and frequency domain in the current monitoring signal, and the wavelet transform is a powerful time-frequency analysis tool that can simultaneously provide local information of the time and frequency of the signal, which is very useful for identifying and understanding instantaneous changes and local events in the signal. Based on this, in the technical solution of the present application, on the monitoring platform, the current monitoring signal is subjected to wavelet transformation to obtain a two-dimensional time-frequency diagram of the current signal, so that abnormal or atypical behaviors can be accurately identified, thereby improving the accuracy of abnormality detection.

[0036] In step S140, on the monitoring platform, the two-dimensional time-frequency graph of the current signal is subjected to current signal time-frequency feature extraction to obtain a current signal time-frequency correlation feature graph. Specifically, in an embodiment of the present application, the two-dimensional time-frequency graph of the current signal is subjected to current signal time-frequency feature extraction to obtain a current signal time-frequency correlation feature graph, including: inputting the two-dimensional time-frequency graph of the current signal into a current signal time-frequency feature extractor based on a hole convolutional neural network model to obtain the current signal time-frequency correlation feature graph. It should be understood that there is a mutual correlation relationship between different local times in the two-dimensional time-frequency graph of the current signal with respect to frequency, and the hole convolutional neural network model expands the receptive field by introducing intervals between the convolution kernels, and at the same time, by adjusting the hole rate of the hole convolution, the size of the receptive field of the model can be flexibly controlled, so as to better perceive and capture the mutual correlation relationship between the input features at different time scales. Therefore, in the technical solution of the present application, the two-dimensional time-frequency diagram of the current signal is input into a current signal time-frequency feature extractor based on a void convolutional neural network model to extract and mine the mutual correlation and dependency between the frequencies of current signals at different time scales, thereby obtaining a current signal time-frequency correlation feature diagram.

[0037] In step S150, the time-frequency correlation feature map of the current signal is feature-dispersed along the channel dimension to obtain a set of time-frequency correlation feature vectors of the current signal pixel granularity. It should be understood that the fine-grained features of the current signal frequency expressed in different channels in the time-frequency correlation feature map of the current signal are different, and these different fine-grained features have an important influence and effect on judging whether the current signal is abnormal. Based on this, in order to enhance the model's ability to represent the time-frequency details of the signal in the time-frequency correlation feature map of the current signal, so that the features of each granularity can capture more detailed time-frequency information, in the technical solution of the present application, the time-frequency correlation feature map of the current signal is feature-dispersed along the channel dimension to obtain a set of time-frequency correlation feature vectors of the current signal pixel granularity, so that each current signal pixel granularity time-frequency correlation feature vector can focus on representing a specific aspect of the current signal, thereby providing rich input for subsequent processing and analysis of the model.

[0038] In step S160, the set of the current signal pixel-granularity time-frequency correlation feature vectors is input into the feature distribution reverse mask clustering analysis network to obtain the current signal global time-frequency clustering representation vector as the current signal global time-frequency clustering representation feature. Accordingly, considering that each current signal pixel-granularity time-frequency correlation feature vector in the set of the current signal pixel-granularity time-frequency correlation feature vectors expresses the correlation feature information of the frequency within a local time period, therefore, in order to further strengthen the mutual correlation dependency between the frequencies within each local time period and form a global time-frequency representation to better characterize and display the macroscopic characteristics of the entire signal, in the technical solution of the present application, the set of the current signal pixel-granularity time-frequency correlation feature vectors is input into the feature distribution reverse mask clustering analysis network to obtain the current signal global time-frequency clustering representation vector.

[0039] In detail, the current signal pixel granularity time-frequency correlation feature vector is first clustered to reveal the potential internal structure and pattern in the current signal pixel granularity time-frequency correlation feature, and the current signal pixel granularity time-frequency correlation cluster center vector with a specific behavior pattern is obtained. Next, the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector and the current signal pixel granularity time-frequency correlation cluster center vector is calculated to quantify the similarity between each feature vector and the cluster center to obtain the correlation coefficient. Then the reciprocal of each correlation coefficient is calculated to obtain the anti-correlation coefficient, that is, a low correlation coefficient may indicate that the feature vector is very different from the cluster center. Calculating the reciprocals of these coefficients can enhance the visibility of these differences and help identify outliers. Afterwards, the anti-correlation coefficient is normalized using the Sigmoid function to map the anti-correlation coefficient to a specific range and retain its original feature information, thereby providing accurate input for subsequent feature weighting. Then, the anti-correlation weight coefficient obtained after normalization is used as the weight to weight each current signal pixel granularity time-frequency correlation feature vector accordingly to reversely filter out noise and irrelevant features to obtain a sequence of suppressed feature vectors. Next, the difference between each current signal pixel granularity time-frequency correlation feature vector and each suppressed vector is calculated to filter out noise or irrelevant feature information to obtain a set of enhanced feature vectors. Finally, the mean vector of the set of enhanced feature vectors is calculated to obtain the global time-frequency clustering representation vector of the current signal with the most key features and rich features.

[0040] Specifically, Figure 3 In the intelligent monitoring method for grounding protection of feeder cables of electrified railways according to an embodiment of the present application, the set of pixel-granularity time-frequency associated feature vectors of the current signal is input into a feature distribution reverse mask clustering analysis network to obtain a flow chart of a global time-frequency clustering representation vector of the current signal. Figure 3As shown, the set of the current signal pixel granularity time-frequency correlation feature vectors is input into the feature distribution reverse mask clustering analysis network to obtain the current signal global time-frequency clustering representation vector, including: S161, clustering the set of the current signal pixel granularity time-frequency correlation feature vectors to obtain the current signal pixel granularity time-frequency correlation cluster center vector; S162, calculating the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of the current signal pixel granularity time-frequency correlation feature vector and the current signal pixel granularity time-frequency correlation cluster center vector to obtain the set of current signal pixel granularity time-frequency correlation coefficients; S163, taking the inverse of each current signal pixel granularity time-frequency correlation coefficient in the set of the current signal pixel granularity time-frequency correlation coefficient to obtain the set of current signal pixel granularity time-frequency anti-correlation coefficients; S164, using the Sigmoid function to normalize the set of the current signal pixel granularity time-frequency anti-correlation coefficients to obtain the current signal S165, using each current signal pixel granularity time-frequency anti-correlation weight coefficient in the set of current signal pixel granularity time-frequency anti-correlation weight coefficients as a weight, weighting each current signal pixel granularity time-frequency association feature vector in the set of current signal pixel granularity time-frequency association feature vectors to obtain a set of suppressed current signal pixel granularity time-frequency association feature vectors; S166, calculating the positional difference between the set of current signal pixel granularity time-frequency association feature vectors and each corresponding group of the current signal pixel granularity time-frequency association feature vectors in the set of suppressed current signal pixel granularity time-frequency association feature vectors and the suppressed current signal pixel granularity time-frequency association feature vectors to obtain a set of enhanced current signal pixel granularity time-frequency association feature vectors; S167, calculating the positional mean vector of the set of enhanced current signal pixel granularity time-frequency association feature vectors to obtain the current signal global time-frequency clustering representation vector.

[0041] More specifically, in an embodiment of the present application, the set of the current signal pixel granularity time-frequency correlation feature vectors is clustered to obtain the current signal pixel granularity time-frequency correlation cluster center vector, including: calculating the mean vector of the set of the current signal pixel granularity time-frequency correlation feature vectors to obtain the current signal pixel granularity time-frequency correlation cluster center vector.

[0042] More specifically, Figure 4 The present invention is a flowchart of calculating the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of current signal pixel granularity time-frequency correlation feature vectors and the current signal pixel granularity time-frequency correlation cluster center vector to obtain a set of current signal pixel granularity time-frequency correlation coefficients in the intelligent monitoring method for grounding protection of feeder cables of electrified railways according to an embodiment of the present application. Figure 4As shown, the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of the current signal pixel granularity time-frequency correlation feature vector and the current signal pixel granularity time-frequency correlation cluster center vector is calculated to obtain a set of current signal pixel granularity time-frequency correlation coefficients, including: S1621, vector multiplication of the transposed vector of the current signal pixel granularity time-frequency correlation feature vector and the current signal pixel granularity time-frequency correlation cluster center vector to obtain a current signal pixel granularity time-frequency correlation value; S1622, respectively calculating the current signal pixel granularity time-frequency correlation feature vector and the current S1623, multiplying the current signal pixel granularity time-frequency association norm value and the current signal pixel granularity time-frequency association cluster norm value to obtain the current signal pixel granularity time-frequency interaction value; S1624, dividing the current signal pixel granularity time-frequency association value by the current signal pixel granularity time-frequency interaction value to obtain the current signal pixel granularity time-frequency correlation coefficient corresponding to the current signal pixel granularity time-frequency association feature vector.

[0043] More specifically, Figure 5 The present invention is a flow chart of using a Sigmoid function to normalize the set of time-frequency anti-correlation coefficients of the pixel granularity of the current signal to obtain a set of time-frequency anti-correlation weight coefficients of the pixel granularity of the current signal in the intelligent monitoring method for grounding protection of feeder cables of electrified railways according to an embodiment of the present application. Figure 5 As shown, the set of the current signal pixel granularity time-frequency anti-correlation coefficients is normalized using a Sigmoid function to obtain a set of current signal pixel granularity time-frequency anti-correlation weight coefficients, including: S1641, using the negative number of each current signal pixel granularity time-frequency anti-correlation coefficient in the set of the current signal pixel granularity time-frequency anti-correlation coefficients as the exponent of a natural constant to calculate an exponential function value based on the position with the natural constant as the base to obtain a set of current signal pixel granularity time-frequency anti-correlation class support coefficients; S1642, calculating the reciprocal of the sum of each current signal pixel granularity time-frequency anti-correlation class support coefficient in the set of the current signal pixel granularity time-frequency anti-correlation class support coefficient and a constant one to obtain a set of current signal pixel granularity time-frequency anti-correlation weight coefficients.

[0044] In the embodiment of the present application, specifically, the set of the current signal pixel granularity time-frequency correlation feature vectors is input into the feature distribution reverse mask clustering analysis network to obtain the current signal global time-frequency clustering representation vector, including: inputting the set of the current signal pixel granularity time-frequency correlation feature vectors into the feature distribution reverse mask clustering analysis network, and processing them with the following clustering formula to obtain the current signal global time-frequency clustering representation vector; wherein the clustering formula is:

[0045]

[0046]

[0047] in, is the set of time-frequency correlation feature vectors of the current signal pixel granularity, is the number of feature vectors in the set of the current signal pixel granularity time-frequency correlation feature vectors, is the first in the set of pixel-granularity time-frequency correlation feature vectors of the current signal The pixel-level time-frequency correlation feature vector of the current signal, is the current signal pixel granularity time-frequency correlation cluster center vector, It is The transposed vector of the pixel-level time-frequency correlation feature vector of the current signal, is the one-norm of the eigenvector, is the first in the set of the current signal pixel granularity time-frequency correlation coefficients The time-frequency correlation coefficient of the pixel granularity of the current signal, is the first in the set of the current signal pixel granularity time-frequency anti-correlation coefficients The time-frequency anti-correlation coefficient of the pixel granularity of the current signal, is the first in the set of the current signal pixel granularity time-frequency anti-correlation weight coefficients The pixel-level time-frequency anti-correlation weight coefficient of the current signal, is the set of pixel-granularity time-frequency correlation feature vectors of the suppression current signal, is the first in the set of pixel-granularity time-frequency correlation feature vectors of the suppression current signal The pixel-level time-frequency correlation feature vector of the suppression current signal, is the set of time-frequency correlation feature vectors of the enhanced current signal pixel granularity, is the first in the set of pixel-granularity time-frequency correlation feature vectors of the enhanced current signal Enhanced current signal pixel-level time-frequency correlation feature vector, is the global time-frequency clustering representation vector of the current signal.

[0048] In step S170, a monitoring result is obtained based on the global time-frequency clustering representation feature of the current signal, and the monitoring result is used to indicate whether an abnormality exists. Specifically, in an embodiment of the present application, a monitoring result is obtained based on the global time-frequency clustering representation feature of the current signal, including: inputting the global time-frequency clustering representation vector of the current signal into a monitoring result generation module based on a classifier to obtain the monitoring result. That is, the global time-frequency clustering representation feature of the current signal obtained by global time-frequency clustering of the set of pixel-granularity time-frequency correlation feature vectors of the current signal is used for classification processing, so as to automatically obtain a monitoring result of whether an abnormality exists. In this way, the current monitoring signal can be collected and transmitted in real time, and the abnormal pattern in the signal can be accurately identified and analyzed using deep learning technology, thereby ensuring the accuracy of monitoring, thereby improving the reliability of cable grounding protection and ensuring the safe operation of electrified railways.

[0049] In particular, in the technical solution of the present application, the current signal time-frequency correlation feature map represents the image semantic correlation features of the two-dimensional time-frequency map of the current signal obtained by wavelet transforming the current monitoring signal. When the current signal time-frequency correlation feature map is feature-dispersed along the channel dimension, and the set of the current signal pixel-granularity time-frequency correlation feature vectors is input into the feature distribution inverse mask clustering analysis network for clustering analysis, the signal undergoes a transformation from the time domain to the frequency domain and then undergoes feature structure dispersion, which will result in the global time-frequency clustering representation vector of the current signal being obtained having a priori-posteriori probabilistic causal association missing relative to the signal modal expression of the current monitoring signal in the time domain, thereby affecting the accuracy of the monitoring results.

[0050] Based on this, in a preferred embodiment, the global time-frequency clustering representation vector of the current signal is input into a monitoring result generation module based on a classifier to obtain a monitoring result, including: obtaining an abnormality probability value corresponding to the presence of an abnormality obtained by inputting the global time-frequency clustering representation vector of the current signal into the monitoring result generation module based on the classifier; probabilizing each eigenvalue of the global time-frequency clustering representation vector of the current signal based on a probability activation function, such as a sigmoid function, to obtain a probabilistic global time-frequency clustering representation vector of the current signal; determining a global time-frequency clustering encoding cognitive symbol value of the current signal based on a comparison between each eigenvalue of the probabilistic global time-frequency clustering representation vector of the current signal and the abnormality probability value, wherein the global time-frequency clustering encoding cognitive symbol value of the current signal is equal to one, zero and negative one respectively in response to the eigenvalue of the probabilistic global time-frequency clustering representation vector of the current signal being greater than, equal to and less than the abnormality probability value; calculating all eigenvalues ​​of the probabilistic global time-frequency clustering representation vector of the current signal The mean is used to obtain the overall phase shift value of the global time-frequency clustering encoding of the current signal; each eigenvalue of the probabilistic global time-frequency clustering representation vector of the current signal is multiplied by the cognitive symbol value of the global time-frequency clustering encoding of the current signal and the overall phase shift value of the global time-frequency clustering encoding of the current signal to obtain the probabilistic global time-frequency clustering encoding cognitive feature vector of the current signal and the probabilistic global time-frequency clustering encoding phase shift feature vector of the current signal; the probabilistic global time-frequency clustering encoding cognitive feature vector of the current signal and the probabilistic global time-frequency clustering encoding phase shift feature vector are multiplied by the first weight and the second weight as hyperparameters respectively, and then the difference is calculated to obtain the probabilistic global time-frequency clustering encoding differential feature vector of the current signal; the absolute value of each eigenvalue of the probabilistic global time-frequency clustering encoding differential feature vector of the current signal is taken to obtain the optimized global time-frequency clustering representation vector of the current signal; the optimized global time-frequency clustering representation vector of the current signal is input into the monitoring result generation module based on the classifier to obtain the monitoring result.

[0051] Specifically, in this preferred embodiment, the probabilistic global time-frequency clustering representation vector of the current signal is optimized to obtain an optimized global time-frequency clustering representation vector of the current signal. The process formula is expressed as follows:

[0052]

[0053] in, is the eigenvalue of each position in the probabilistic global time-frequency clustering representation vector of the current signal, is the abnormal probability value, is a sign function, and is the weight hyperparameter, is the overall phase shift value of the global time-frequency clustering encoding of the current signal, is the eigenvalue of each position in the optimized global time-frequency clustering representation vector of the current signal.

[0054] Accordingly, in the above preferred embodiment, the class cognitive phase transformation response of the global time-frequency clustering representation vector of the current signal is obtained by comparing the probabilistic amplitude of the eigenvalue of the global time-frequency clustering representation vector of the current signal with the conforming probability representation, and the characteristic distribution sequence invariance transformation based on the class probability differential distribution expansion of the eigenvalue of the global time-frequency clustering representation vector of the current signal relative to the class probability representation of the feature as a whole is performed to achieve the causal constraint of the posterior class probability of the global time-frequency clustering representation vector of the current signal on its prior feature distribution representation, so as to improve the accuracy of the monitoring results obtained by inputting the global time-frequency clustering representation vector of the current signal into the monitoring result generation module based on the classifier. In this way, the current monitoring signal can be collected and transmitted in real time, and the abnormal pattern in the signal can be accurately identified and analyzed by using deep learning technology, thereby ensuring the accuracy of monitoring, thereby improving the reliability of cable grounding protection and ensuring the safe operation of electrified railways.

[0055] In summary, the intelligent monitoring method for grounding protection of feeder cables of electrified railways based on the embodiment of the present application is explained, which collects current monitoring signals through current sensors installed on the grounding leads of the shielding layer of the railway feeder cables, and transmits the current monitoring signals to the monitoring platform through a wireless communication module, and uses deep learning-based signal processing and analysis technology in the monitoring platform to perform time-frequency analysis and correlation of the current monitoring signals, so as to automatically obtain the monitoring results of whether there are abnormalities based on the clustering representation characteristics of the pixel granularity of the multiple current signals in the global time-frequency. In this way, the current monitoring signals can be collected and transmitted in real time, and the abnormal patterns in the signals can be accurately identified and analyzed using deep learning technology, thereby ensuring the accuracy of monitoring, thereby improving the reliability of cable grounding protection and ensuring the safe operation of electrified railways.

[0056] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

[0057] In this application, words such as "including", "comprising", "having", etc. are open words, meaning "including but not limited to", and can be used interchangeably therewith. The words "or" and "and" used herein mean the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein means the phrase "such as but not limited to", and can be used interchangeably therewith.

[0058] It should also be noted that in the method of the present application, each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present application.

[0059] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features of the present application.

Claims

1. An intelligent monitoring method for grounding protection of feeder cables in electrified railways, characterized in that: include: Acquiring a current monitoring signal collected by a current sensor, wherein the current sensor is installed on a ground lead of a shielding layer of a railway feeder cable; Transmitting the current monitoring signal to a monitoring platform via a wireless communication module; On the monitoring platform, performing time-frequency conversion on the current monitoring signal to obtain a two-dimensional time-frequency graph of the current signal; On the monitoring platform, extracting the time-frequency features of the current signal from the two-dimensional time-frequency graph of the current signal to obtain a time-frequency correlation feature graph of the current signal; Decentralizing the time-frequency correlation feature map of the current signal along the channel dimension to obtain a set of time-frequency correlation feature vectors of the current signal pixel granularity; Inputting the set of the current signal pixel granularity time-frequency correlation feature vectors into a feature distribution reverse mask clustering analysis network to obtain a current signal global time-frequency clustering representation vector as a current signal global time-frequency clustering representation feature; Based on the global time-frequency clustering representation feature of the current signal, a monitoring result is obtained, wherein the monitoring result is used to indicate whether there is an abnormality; Inputting the set of the current signal pixel granularity time-frequency correlation feature vectors into a feature distribution reverse mask clustering analysis network to obtain a global time-frequency clustering representation vector of the current signal, including: Clustering the set of the current signal pixel granularity time-frequency correlation feature vectors to obtain a current signal pixel granularity time-frequency correlation cluster center vector; Calculating the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of the current signal pixel granularity time-frequency correlation feature vectors and the current signal pixel granularity time-frequency correlation cluster center vector to obtain a set of current signal pixel granularity time-frequency correlation coefficients; Taking the inverse of each current signal pixel granularity time-frequency correlation coefficient in the set of current signal pixel granularity time-frequency correlation coefficients to obtain a set of current signal pixel granularity time-frequency anti-correlation coefficients; Using a Sigmoid function to normalize the set of the current signal pixel granularity time-frequency anti-correlation coefficients to obtain a set of current signal pixel granularity time-frequency anti-correlation weight coefficients; Taking each current signal pixel granularity time-frequency anti-correlation weight coefficient in the set of current signal pixel granularity time-frequency anti-correlation weight coefficients as a weight, weighting each current signal pixel granularity time-frequency association feature vector in the set of current signal pixel granularity time-frequency association feature vectors to obtain a set of suppressed current signal pixel granularity time-frequency association feature vectors; Calculating the set of the current signal pixel granularity time-frequency correlation feature vectors and the set of the suppressed current signal pixel granularity time-frequency correlation feature vectors, each group of corresponding current signal pixel granularity time-frequency correlation feature vectors and the suppressed current signal pixel granularity time-frequency correlation feature vectors are differentiated position by position to obtain a set of enhanced current signal pixel granularity time-frequency correlation feature vectors; The position-by-position mean vector of the set of pixel-granularity time-frequency associated feature vectors of the enhanced current signal is calculated to obtain the global time-frequency clustering representation vector of the current signal.

2. The intelligent monitoring method for grounding protection of feeder cables in electrified railways according to claim 1 is characterized in that: Performing time-frequency conversion on the current monitoring signal to obtain a two-dimensional time-frequency diagram of the current signal includes: performing wavelet transformation on the current monitoring signal to obtain the two-dimensional time-frequency diagram of the current signal.

3. The intelligent monitoring method for grounding protection of feeder cables in electrified railways according to claim 2 is characterized in that: The two-dimensional time-frequency diagram of the current signal is subjected to current signal time-frequency feature extraction to obtain a current signal time-frequency correlation feature diagram, including: inputting the two-dimensional time-frequency diagram of the current signal into a current signal time-frequency feature extractor based on a hole convolutional neural network model to obtain the current signal time-frequency correlation feature diagram.

4. The intelligent monitoring method for grounding protection of feeder cables in electrified railways according to claim 3 is characterized in that: Clustering the set of the current signal pixel granularity time-frequency correlation feature vectors to obtain the current signal pixel granularity time-frequency correlation cluster center vector, including: calculating the mean vector of the set of the current signal pixel granularity time-frequency correlation feature vectors to obtain the current signal pixel granularity time-frequency correlation cluster center vector.

5. The intelligent monitoring method for grounding protection of feeder cables in electrified railways according to claim 4 is characterized in that: Calculating the correlation coefficient between each current signal pixel granularity time-frequency correlation feature vector in the set of the current signal pixel granularity time-frequency correlation feature vectors and the current signal pixel granularity time-frequency correlation cluster center vector to obtain a set of current signal pixel granularity time-frequency correlation coefficients, including: Performing vector multiplication of the transposed vector of the current signal pixel granularity time-frequency correlation feature vector and the current signal pixel granularity time-frequency correlation cluster center vector to obtain a current signal pixel granularity time-frequency correlation value; Respectively calculating the time-frequency association characteristic vector of the current signal pixel granularity and the norm of the time-frequency association cluster center vector of the current signal pixel granularity to obtain the time-frequency association norm value of the current signal pixel granularity and the time-frequency association cluster norm value of the current signal pixel granularity; Multiplying the current signal pixel granularity time-frequency correlation norm value and the current signal pixel granularity time-frequency correlation cluster norm value to obtain a current signal pixel granularity time-frequency interaction value; The current signal pixel granularity time-frequency correlation value is divided by the current signal pixel granularity time-frequency interaction value to obtain the current signal pixel granularity time-frequency correlation coefficient corresponding to the current signal pixel granularity time-frequency correlation feature vector.

6. The method for intelligent monitoring of grounding protection of feeder cables in electrified railways according to claim 5, characterized in that: The set of the time-frequency anti-correlation coefficients of the pixel granularity of the current signal is normalized by using a Sigmoid function to obtain a set of the time-frequency anti-correlation weight coefficients of the pixel granularity of the current signal, including: Taking the negative number of each current signal pixel granularity time-frequency anti-correlation coefficient in the set of current signal pixel granularity time-frequency anti-correlation coefficients as the exponent of the natural constant, to calculate the exponential function value with the natural constant as the base for each position, to obtain the set of current signal pixel granularity time-frequency anti-correlation class support coefficients; The reciprocal of the sum of each current signal pixel granularity time-frequency anti-correlation class support coefficient in the set of the current signal pixel granularity time-frequency anti-correlation class support coefficient and a constant one is calculated to obtain the set of the current signal pixel granularity time-frequency anti-correlation weight coefficients.

7. The intelligent monitoring method for grounding protection of feeder cables in electrified railways according to claim 6 is characterized in that: Based on the global time-frequency clustering representation feature of the current signal, a monitoring result is obtained, including: inputting the global time-frequency clustering representation vector of the current signal into a monitoring result generation module based on a classifier to obtain the monitoring result.