CN-122017450-A - Cable fault distance measurement method, system, equipment and medium

CN122017450ACN 122017450 ACN122017450 ACN 122017450ACN-122017450-A

Abstract

The invention discloses a cable fault distance measurement method, system, equipment and medium, which comprise the steps of obtaining current signals of a normal signal section and a fault signal section of a cable, constructing an original data set according to the current signals, carrying out data processing on the original data set to obtain a sample matrix, calculating statistical characteristics of each channel of the normal signal section in the sample matrix, respectively constructing a time domain outlier matrix and a frequency domain outlier matrix according to the statistical characteristics, fusing the time domain outlier matrix and the frequency domain outlier matrix to obtain a time-frequency fusion outlier priori matrix, carrying out feature extraction on the sample matrix based on the time-frequency fusion outlier priori matrix to obtain priori attention enhancement features, calculating a fault distance measurement result based on the priori attention enhancement features, quantifying the distance measurement result to obtain reliability of the distance measurement result, optimizing the reliability, and outputting a final distance measurement result, thereby facilitating verification and decision of industrial field personnel.

Inventors

SUN RUIZE
NING NAN
CHEN HUAILIN
QIU HONGQI
CHE HONGBO
ZHOU YIRAN

Assignees

贵州电网有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251230

Claims (10)

1. A cable fault location method, comprising: Acquiring current signals of a normal signal section and a fault signal section of a cable, constructing an original data set according to the current signals, and performing data processing on the original data set to obtain a sample matrix; Calculating the statistical characteristics of each channel of a normal signal segment in the sample matrix, respectively constructing a time domain outlier matrix and a frequency domain outlier matrix according to the statistical characteristics, and fusing the time domain outlier matrix and the frequency domain outlier matrix to obtain a time-frequency fusion outlier prior matrix; based on the time-frequency fusion outlier prior matrix, extracting features of the sample matrix to obtain prior attention enhancement features; Based on the prior attention enhancement features, calculating to obtain a fault ranging result, quantifying the ranging result to obtain the reliability of the ranging result, optimizing the reliability, and outputting a final ranging result.
2. The cable fault location method of claim 1, wherein data processing the original data set comprises: acquiring multichannel current signals of two ends of a cable and midway monitoring points, discretizing according to a set sampling interval, and constructing an original data set containing sample value vectors of each channel and corresponding state labels at each discrete moment; And windowing the original data set, and stacking the data in each window according to the sequence to obtain a two-dimensional sample matrix, thereby obtaining the two-dimensional sample matrix.
3. The cable fault location method of claim 2, wherein constructing the time domain outlier matrix comprises: Calculating the current signal mean value and standard deviation of each channel based on the normal signal segment; carrying out first judgment on the current value of each data point in the sample matrix and the average value of the corresponding channel; Setting a first threshold according to the standard deviation and the confidence coefficient; if the deviation degree of the current value of each data point and the mean value of the corresponding channel is larger than a first threshold value, marking the corresponding data point as abnormal; Otherwise, marking the corresponding data point as normal; And generating a binary indication matrix with the same size as the sample matrix as a time domain outlier matrix by using the data points marked as anomalies.
4. A method of cable fault location as claimed in claim 3, wherein constructing a frequency domain outlier matrix comprises: frequency domain analysis is respectively carried out on each channel data in each sample matrix, and signal energy in a plurality of preset concerned frequency bands is calculated to obtain spectrum energy time tracks of each channel at different time points; calculating the mean value and standard deviation of the spectral energy of each channel based on the normal signal segment; setting a second threshold according to the standard deviation and the confidence coefficient of the spectral energy of each channel; if the deviation degree of the spectrum energy time track of each channel at different time points and the average value of the spectrum energy of each channel is larger than a second threshold value, marking the corresponding spectrum energy as abnormal, and generating a binary frequency domain outlier indication matrix; And linearly combining the time domain outlier indication matrix with the frequency domain outlier indication matrix according to preset weights, and comparing the time domain outlier indication matrix with a binarization threshold value to generate a time-frequency fusion outlier prior matrix.
5. The cable fault location method as claimed in claim 4, wherein performing feature extraction on the sample matrix comprises: Inputting a sample matrix into a neural network for forward propagation to obtain a total feature map formed by stacking feature maps of a plurality of channels; Taking the time-frequency fusion outlier prior matrix as a query matrix in an attention mechanism; Performing nonlinear mapping on the feature map of each channel in the total feature map, and performing aggregation operation along the channel dimension to generate a key matrix with the same size as the query matrix; calculating the distance between the query matrix and the key matrix, and calculating the attention weight matrix through a normalized exponential function based on the distance; Weighting the feature map of each channel in the total feature map element by using an attention weight matrix to obtain a priori attention enhancement feature; based on the prior attention enhancement features, intermediate results required for fault ranging are output through regression prediction heads.
6. The cable fault location method of claim 1 or 5, wherein calculating a fault location result comprises: Reducing the local index of the regression prediction head in each analysis window into a sample index on the global time sequence by combining the initial position offset of each window; Fusing global sample indexes of all windows to obtain global key point indexes; multiplying the global key point index by a sampling interval, and converting the sampling interval into a physical time value to obtain the actual arrival time of the fault key signal; By applying excitation to cable calibration sections with known physical lengths, measuring the actual arrival time difference of fault key signals and calculating to obtain equivalent wave speed; And calculating the distance between the fault point and the measuring port according to the traveling wave ranging principle based on the equivalent wave speed and the calculated actual arrival time difference of the fault key signal.
7. The cable fault location method of claim 6, wherein outputting the final location result comprises: calculating the consistency ratio between the attention weight matrix and the time-frequency fusion outlier prior matrix, and taking the consistency ratio as the confidence coefficient of the ranging result; Comparing the confidence coefficient with a preset confidence coefficient threshold value; if the confidence coefficient is lower than the confidence coefficient threshold value, a binarization threshold value used for generating a time-frequency fusion outlier prior matrix is adjusted, and based on the adjusted parameters, the construction of the time domain outlier matrix and the frequency domain outlier matrix is re-executed; and if the confidence coefficient reaches or exceeds the confidence coefficient threshold value, outputting a final ranging result and the corresponding confidence coefficient.
8. A cable fault location system employing a cable fault location method according to any one of claims 1 to 7, comprising: the data processing module is used for acquiring current signals of a normal signal section and a fault signal section of the cable, constructing an original data set according to the current signals, and performing data processing on the original data set to obtain a sample matrix; The matrix construction fusion module is used for calculating the statistical characteristics of each channel of the normal signal segment in the sample matrix, respectively constructing a time domain outlier matrix and a frequency domain outlier matrix according to the statistical characteristics, and fusing the time domain outlier matrix and the frequency domain outlier matrix to obtain a time-frequency fusion outlier prior matrix; The feature extraction module is used for extracting features of the sample matrix based on the time-frequency fusion outlier prior matrix to obtain prior attention enhancement features; And the calculation module is used for calculating a fault ranging result based on the prior attention enhancing characteristic, quantifying the ranging result to obtain the reliability of the ranging result, optimizing the reliability and outputting a final ranging result.
9. A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the steps of a cable fault location method as claimed in any one of claims 1 to 7 when the computer program is executed.
10. A computer-readable storage medium, characterized in that it has stored thereon a computer program which, when executed by a processor, implements the steps of a cable fault location method according to any one of claims 1 to 7.

Description

Cable fault distance measurement method, system, equipment and medium Technical Field The invention relates to the technical field of cable fault location, in particular to a cable fault location method, a system, equipment and a medium. Background At present, in the existing cable fault ranging technology, a research hotspot is a method based on a Convolutional Neural Network (CNN), and the method has the following core defects in terms of being combined with related technologies, the interpretation is lacking, the traditional CNN belongs to a black box model, when the traditional CNN processes multichannel current signals of two ends of a cable and midway monitoring points, no method is available for defining fault key information of the model, such as attention areas of head wave arrival signals and fault point reflection signals, the ranging result lacks credibility verification basis, and the traceable requirements of industrial scenes on the diagnosis process are difficult to be met. And the traditional CNN has low recognition precision on key signals, and cable fault signals are easily interfered by power frequency harmonic waves and environmental noise. The existing methods are independent of time domains, such as signal amplitude threshold values or frequency domains, such as specific frequency band energy analysis, and do not integrate time-frequency domain outlier information, namely abnormal signal areas related to faults, and can cause large critical time extraction errors of head waves and reflected waves in complex noise environments to directly cause ranging deviation. In addition, the prior fusion efficiency of the traditional CNN is poor, part of methods try to integrate prior knowledge such as signal statistics characteristics, but pre-training is mostly adopted, such as model pre-training or pre-diagnosis based on failure frequency prior, for example, a shallow fusion mode of firstly carrying out signal anomaly detection and then inputting the signals into a network is adopted, deep embedding with the CNN characteristic extraction process is avoided, the calculation cost is increased, and the real-time signal processing requirement of multiple monitoring points of a cable is not met flexibly. The noise immunity of the traditional CNN is weak, and the traditional CNN faces a low signal-to-noise ratio scene, such as noise below-20 dB. The traditional CNN and the improved method, such as the extrusion excitation network SE-Net and the mask CNN, are easy to be influenced by interference signals without correlation, fault characteristic extraction offset and ranging performance are obviously reduced. Aiming at the defects of the prior cable fault location technology, the invention provides a cable fault location method, a system, equipment and a medium, which solve the core technical problems of the prior art scheme. Disclosure of Invention In view of the above existing problems, the present invention provides a cable fault location method, system, device and medium. The invention provides a cable fault distance measurement method, a system, equipment and a medium, which solve the problems of poor model interpretability, weak noise resistance, low prior fusion efficiency, insufficient key signal extraction precision and the like in the existing cable fault distance measurement technology. In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the present invention provides a cable fault location method, including: Acquiring current signals of a normal signal section and a fault signal section of a cable, constructing an original data set according to the current signals, and performing data processing on the original data set to obtain a sample matrix; Calculating the statistical characteristics of each channel of a normal signal segment in the sample matrix, respectively constructing a time domain outlier matrix and a frequency domain outlier matrix according to the statistical characteristics, and fusing the time domain outlier matrix and the frequency domain outlier matrix to obtain a time-frequency fusion outlier prior matrix; based on the time-frequency fusion outlier prior matrix, extracting features of the sample matrix to obtain prior attention enhancement features; Based on the prior attention enhancement features, calculating to obtain a fault ranging result, quantifying the ranging result to obtain the reliability of the ranging result, optimizing the reliability, and outputting a final ranging result. As a preferable scheme of the cable fault location method, the method comprises the following steps of: acquiring multichannel current signals of two ends of a cable and midway monitoring points, discretizing according to a set sampling interval, and constructing an original data set containing sample value vectors of each channel and corresponding state labels at each discrete moment; And windowing the original data set, a