CN-121980170-A - Cable fault diagnosis method and system based on multi-mode data processing

CN121980170ACN 121980170 ACN121980170 ACN 121980170ACN-121980170-A

Abstract

The invention is applied to the field of data processing, and discloses a cable fault diagnosis method and system based on multi-mode data processing, wherein the method comprises the steps of obtaining dual-mode heterogeneous data of a target cable and analyzing the data quality condition of the dual-mode heterogeneous data; extracting numerical mode characteristics and visual mode characteristics in bimodal heterogeneous data, carrying out fusion processing on the bimodal characteristics based on data quality conditions to obtain bimodal fusion characteristics, carrying out fault prediction processing on the bimodal fusion characteristics, extracting confidence distinction degree of a fault prediction result, correcting the confidence distinction degree based on similarity degree between the numerical mode characteristics and the visual mode characteristics to obtain corrected confidence distinction degree, integrating the corrected confidence distinction degree, similarity degree and data quality conditions to obtain a decision risk result, screening the bimodal heterogeneous data based on the decision risk result, and obtaining a fault diagnosis result based on the screening result. The method improves the accuracy of cable fault diagnosis.

Inventors

YAN FENG
LI SHAOBIN
GUO ZHIGANG
QU TONG
LIU PENGYUE
YANG LINQING
Yao Jiachi
WU QIAN
JIA ZHANHAO
JIANG YANZHUO

Assignees

国网陕西省电力有限公司西安供电公司

Dates

Publication Date: 20260505
Application Date: 20251230

Claims (10)

1. A cable fault diagnosis method based on multi-modal data processing, comprising: acquiring bimodal heterogeneous data of a target cable, and analyzing and obtaining the data quality condition of the bimodal heterogeneous data; Extracting numerical modal characteristics and visual modal characteristics from the bimodal heterogeneous data, and carrying out fusion processing on the numerical modal characteristics and the visual modal characteristics based on the data quality condition to obtain bimodal fusion characteristics; Performing fault prediction processing on the bimodal fusion characteristics, and extracting confidence distinction degree corresponding to the cable state analysis condition in a fault prediction result; determining the similarity degree between the numerical mode characteristics and the visual mode characteristics, and correcting the confidence differentiation degree based on the similarity degree to obtain a corrected confidence differentiation degree; Integrating the correction confidence distinction degree, the similarity degree and the data quality condition to obtain a decision risk result; And screening the bimodal heterogeneous data based on the decision risk result, and obtaining a fault diagnosis result of the target cable based on the screening result.
2. The method for diagnosing a cable fault based on multi-modal data processing as claimed in claim 1, wherein said acquiring the dual-modal heterogeneous data of the target cable includes: mapping the bimodal heterogeneous data to a unified time axis, and selecting candidate numerical value monitoring data and candidate visual inspection data within a preset time range of the unified time axis; Sequentially carrying out robust standardization and abnormal peak clipping treatment on the candidate numerical value monitoring data to obtain standardized numerical value modal data; normalizing the candidate visual inspection data to obtain standardized visual mode data; And integrating the standardized numerical mode data and the standardized visual mode data to generate bimodal heterogeneous data of the target cable.
3. The method for diagnosing a cable fault based on multi-modal data processing as set forth in claim 1, wherein said extracting numerical mode features and visual mode features from the bimodal heterogeneous data includes: Carrying out numerical coding processing on standardized numerical mode data in the bimodal heterogeneous data to obtain numerical mode characteristics; Performing visual coding processing on standardized visual mode data in the bimodal heterogeneous data to obtain initial visual mode characteristics; And executing time aggregation processing within a preset time range on the initial visual mode characteristics to obtain visual mode characteristics, wherein the time aggregation processing is designed to dynamically allocate the weight of the sample characteristics based on the time distance between the sample characteristics and a target time step so as to weaken long-time sample characteristic interference.
4. The method for diagnosing cable faults based on multi-modal data processing as claimed in claim 1, wherein the performing fault prediction processing on the dual-modal fusion features, extracting confidence distinction corresponding to cable state analysis conditions in fault prediction results, comprises: Inputting the bimodal fusion characteristics into a pre-trained cable fault prediction model for linear mapping to obtain initial category posterior probability reflecting each health level of a target cable and corresponding original prediction scores; Determining a calibration coefficient based on the data quality condition, scaling and calibrating the original prediction score according to the calibration coefficient to obtain a calibration prediction score, and generating a calibration category posterior probability based on the calibration prediction score; and carrying out posterior probability interval calculation processing according to the calibration category posterior probability, and extracting probability distribution difference values in the posterior probability interval calculation result as confidence distinction degree of the cable state analysis condition, wherein the confidence distinction degree reflects the determinable strength of the bimodal fusion characteristic.
5. The cable fault diagnosis method based on multi-modal data processing as set forth in claim 1, wherein the screening the dual-modal heterogeneous data based on the decision risk result and obtaining the fault diagnosis result of the target cable based on the screening result includes: Performing threshold optimization processing on the decision risk result based on a preset coverage rate to obtain an upper threshold and a lower threshold, wherein the upper threshold and the lower threshold are used for determining the confidence level of the bimodal heterogeneous data; classifying bimodal heterogeneous data corresponding to the decision risk result according to the upper and lower double thresholds to obtain comprehensive confidence coefficient of the bimodal heterogeneous data; And performing fault analysis on the bimodal heterogeneous data corresponding to the comprehensive confidence coefficient meeting the preset confidence coefficient to obtain a fault diagnosis result of the target cable.
6. A cable fault diagnosis system based on multi-modal data processing, comprising: the acquisition module is used for acquiring bimodal heterogeneous data of the target cable and analyzing and obtaining the data quality condition of the bimodal heterogeneous data; The extraction module is used for extracting numerical modal characteristics and visual modal characteristics in the bimodal heterogeneous data, and carrying out fusion processing on the numerical modal characteristics and the visual modal characteristics based on the data quality condition to obtain bimodal fusion characteristics; The prediction module is used for carrying out fault prediction processing on the bimodal fusion characteristics and extracting confidence distinction degree corresponding to the cable state analysis condition in a fault prediction result; The correction module is used for determining the similarity degree between the numerical mode characteristics and the visual mode characteristics, correcting the confidence differentiation degree based on the similarity degree and obtaining a corrected confidence differentiation degree; The integration module is used for integrating the correction confidence distinction degree, the similarity degree and the data quality condition to obtain a decision risk result; and the generation module is used for screening the bimodal heterogeneous data based on the decision risk result and obtaining a fault diagnosis result of the target cable based on the screening result.
7. The multi-modal data processing based cable fault diagnosis system of claim 6, wherein the acquisition module is specifically configured to: mapping the bimodal heterogeneous data to a unified time axis, and selecting candidate numerical value monitoring data and candidate visual inspection data within a preset time range of the unified time axis; Sequentially carrying out robust standardization and abnormal peak clipping treatment on the candidate numerical value monitoring data to obtain standardized numerical value modal data; normalizing the candidate visual inspection data to obtain standardized visual mode data; And integrating the standardized numerical mode data and the standardized visual mode data to generate bimodal heterogeneous data of the target cable.
8. The multi-modal data processing based cable fault diagnosis system of claim 6, wherein the extraction module is specifically configured to: Carrying out numerical coding processing on standardized numerical mode data in the bimodal heterogeneous data to obtain numerical mode characteristics; Performing visual coding processing on standardized visual mode data in the bimodal heterogeneous data to obtain initial visual mode characteristics; And executing time aggregation processing within a preset time range on the initial visual mode characteristics to obtain visual mode characteristics, wherein the time aggregation processing is designed to dynamically allocate the weight of the sample characteristics based on the time distance between the sample characteristics and a target time step so as to weaken long-time sample characteristic interference.
9. The multi-modal data processing based cable fault diagnosis system of claim 6, wherein the prediction module is specifically configured to: Inputting the bimodal fusion characteristics into a pre-trained cable fault prediction model for linear mapping to obtain initial category posterior probability reflecting each health level of a target cable and corresponding original prediction scores; Determining a calibration coefficient based on the data quality condition, scaling and calibrating the original prediction score according to the calibration coefficient to obtain a calibration prediction score, and generating a calibration category posterior probability based on the calibration prediction score; and carrying out posterior probability interval calculation processing according to the calibration category posterior probability, and extracting probability distribution difference values in the posterior probability interval calculation result as confidence distinction degree of the cable state analysis condition, wherein the confidence distinction degree reflects the determinable strength of the bimodal fusion characteristic.
10. The cable fault diagnosis system based on multi-modal data processing as set forth in claim 6, wherein said generation module is specifically configured to: Performing threshold optimization processing on the decision risk result based on a preset coverage rate to obtain an upper threshold and a lower threshold, wherein the upper threshold and the lower threshold are used for determining the confidence level of the bimodal heterogeneous data; classifying bimodal heterogeneous data corresponding to the decision risk result according to the upper and lower double thresholds to obtain comprehensive confidence coefficient of the bimodal heterogeneous data; And performing fault analysis on the bimodal heterogeneous data corresponding to the comprehensive confidence coefficient meeting the preset confidence coefficient to obtain a fault diagnosis result of the target cable.

Description

Cable fault diagnosis method and system based on multi-mode data processing Technical Field The invention relates to the technical field of data processing, in particular to a cable fault diagnosis method and system based on multi-mode data processing. Background The high-voltage cable system is an 'energy artery' for modern urban power grid and clean energy transmission, and the operation reliability of the high-voltage cable system is directly related to the safety and stability of the whole power system. The power cable operates in a long-term buried and wet load environment, and factors such as insulation aging, joint defects, sheath corrosion, partial discharge and the like are interwoven and overlapped, so that a failure mechanism presents multi-scale, multi-stage and multi-mode characteristics. Therefore, an accurate and efficient cable fault diagnosis technology becomes a research focus and urgent need in the field of power operation and maintenance. The existing cable fault diagnosis method is to directly conduct fault analysis on detected data, for example, the fault diagnosis is directly conducted on the detected data by depending on a fixed threshold value and an empirical rule, however, the fixed threshold value lacks dynamic adaptability and cannot be matched with nonlinear fault characteristics caused by different working conditions, aging stages and distributed power grid coupling of a cable, the cable is susceptible to failure caused by the influence of measurement precision and external interference, the empirical rule depends on manual subjective judgment, the generalization capability is weak, and complex fault scenes of multi-factor interleaving are difficult to deal with. These defects directly lead to high false alarm rate and false missing rate of diagnosis results, and early faults with short duration and weak signals cannot be accurately identified, so that operation and maintenance decision blindness is increased, and the safe and stable operation of the power system is threatened. Disclosure of Invention The invention provides a cable fault diagnosis method and system based on multi-mode data processing, which aim to solve the technical problem of how to improve the existing cable fault diagnosis method and achieve the technical effect of improving the cable diagnosis accuracy. In order to solve the technical problems, an aspect of the present invention provides a cable fault diagnosis method based on multi-mode data processing, including: acquiring bimodal heterogeneous data of a target cable, and analyzing and obtaining the data quality condition of the bimodal heterogeneous data; Extracting numerical modal characteristics and visual modal characteristics from the bimodal heterogeneous data, and carrying out fusion processing on the numerical modal characteristics and the visual modal characteristics based on the data quality condition to obtain bimodal fusion characteristics; Performing fault prediction processing on the bimodal fusion characteristics, and extracting confidence distinction degree corresponding to the cable state analysis condition in a fault prediction result; determining the similarity degree between the numerical mode characteristics and the visual mode characteristics, and correcting the confidence differentiation degree based on the similarity degree to obtain a corrected confidence differentiation degree; Integrating the correction confidence distinction degree, the similarity degree and the data quality condition to obtain a decision risk result; And screening the bimodal heterogeneous data based on the decision risk result, and obtaining a fault diagnosis result of the target cable based on the screening result. As one preferable solution, the acquiring the bimodal heterogeneous data of the target cable includes: mapping the bimodal heterogeneous data to a unified time axis, and selecting candidate numerical value monitoring data and candidate visual inspection data within a preset time range of the unified time axis; Sequentially carrying out robust standardization and abnormal peak clipping treatment on the candidate numerical value monitoring data to obtain standardized numerical value modal data; normalizing the candidate visual inspection data to obtain standardized visual mode data; And integrating the standardized numerical mode data and the standardized visual mode data to generate bimodal heterogeneous data of the target cable. As one preferable solution, the extracting numerical mode characteristics and visual mode characteristics in the bimodal heterogeneous data includes: Carrying out numerical coding processing on standardized numerical mode data in the bimodal heterogeneous data to obtain numerical mode characteristics; Performing visual coding processing on standardized visual mode data in the bimodal heterogeneous data to obtain initial visual mode characteristics; And executing time aggregation processing within a preset time range on t