CN-122020101-A - Partial discharge data management method and device for high-voltage cable

CN122020101ACN 122020101 ACN122020101 ACN 122020101ACN-122020101-A

Abstract

The invention belongs to the technical field of power equipment state monitoring, and discloses a partial discharge data management method and device for a high-voltage cable. The data management method comprises the steps of obtaining a multi-mode original defect signal, carrying out feature extraction to obtain an initial defect feature, inputting the initial defect feature into a pre-built multi-scale noise self-adaptive separation network to be processed to obtain target noise probability and target noise feature, storing the target noise feature into a noise library when the target noise probability is larger than a preset noise probability threshold, otherwise judging whether the similarity between the discharge defect feature and each preset discharge type cluster in the discharge type library meets a preset similarity requirement, calculating the target discharge type probability of the discharge defect feature, or establishing a new discharge type cluster, so that intelligent management of partial discharge data is realized, and the recognition accuracy of partial discharge defects is improved.

Inventors

YANG RUI
LIANG JUNLING
LIU CUI
WANG LIANGJUN
WANG XUN
YANG HONGJIANG

Assignees

重庆泰昇智能电气有限公司
重庆泰山电缆有限公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. A method of partial discharge data management for a high voltage cable, comprising: S1, acquiring a multi-mode original defect signal, and extracting features of the original defect signal to obtain an initial defect feature; s2, inputting the initial defect characteristics into a pre-constructed multi-scale noise self-adaptive separation network for processing to obtain target noise characteristics, and calculating target noise probability of the target noise characteristics; S3, when the target noise probability is larger than a preset noise probability threshold, storing the target noise characteristic into a noise library, otherwise, performing signal purification on the initial defect characteristic according to a corresponding filtering strategy to obtain a discharge defect characteristic; S4, judging whether the similarity of the discharge defect characteristics and each preset discharge type cluster in a discharge type library meets preset similarity requirements, when the similarity meets the preset similarity requirements, calculating target discharge type probability of the discharge defect characteristics, determining the target discharge type cluster of the discharge defect characteristics based on the target discharge type probability, and outputting a corresponding target discharge type; And S5, calculating index priority scores according to the access information of the discharge type library, and carrying out hierarchical management on indexes according to the priority scores.
2. The method of claim 1, wherein the performing feature extraction on the original defect signal to obtain an initial defect feature comprises: performing wavelet threshold denoising on the original defect signal to obtain a denoised original defect signal; Performing cross-modal feature extraction on the denoised original defect signal through a cross-modal feature extraction module to obtain a multi-modal feature vector; and executing batch normalization and regularization processing on the multi-mode feature vector through a feature constraint module to obtain the initial defect feature.
3. The method according to claim 1, wherein step S2 comprises: S21, carrying out feature extraction on the initial defect feature according to a preset noise feature extraction dimension to obtain an initial noise feature vector; s22, processing the initial noise feature vector according to a multi-scale feature extraction module to obtain multi-scale noise features; S23, sequentially executing spatial attention and channel attention processing on the multi-scale noise features according to an attention feature extraction module to obtain key noise features; S24, carrying out feature processing on the key noise features according to a time sequence feature extraction module to obtain depth noise features; S25, carrying out feature processing on the depth noise features through a classification regression module to obtain the target noise features and noise probability distribution, wherein the noise probability distribution is used for representing noise probabilities of different noise types corresponding to the target noise features; the multi-scale noise self-adaptive separation network comprises the multi-scale feature extraction module, the attention feature extraction module, the time sequence feature extraction module and the classification regression module.
4. A method according to claim 3, characterized in that the method further comprises: And triggering the incremental training of the multi-scale noise self-adaptive separation network when the number of the target noise features which are newly added in the noise library reaches a preset threshold value, wherein the incremental training process comprises the step of executing fine adjustment of a preset learning rate on the classification regression module.
5. The method of claim 1, wherein the determining whether the similarity of the discharge defect feature to each of the predetermined discharge type clusters in the discharge type library meets a predetermined similarity requirement comprises: calculating cosine similarity between the discharge defect characteristics and the centers of the preset discharge type clusters respectively, and judging that the similarity meets the preset similarity requirement if the cosine similarity is in a first range; If the cosine similarity is in a second range, calculating the mahalanobis distance between the discharge defect characteristics and all preset discharge type clusters, and combining the mahalanobis distance of each preset discharge type cluster, calculating the inter-cluster similarity of each preset discharge type cluster, and if the inter-cluster similarity is in a third range, judging that the similarity meets the preset similarity requirement; If the cosine similarity is in a fourth range or the inter-cluster similarity is in the fourth range, judging that the similarity does not meet the preset similarity requirement, and sequentially reducing the values of the first range, the second range, the third range and the fourth range.
6. The method of claim 5, wherein the calculating the target discharge type probability for the discharge defect feature comprises: generating a time-frequency spectrum corresponding to the discharge defect characteristics, and extracting the characteristics of the time-frequency spectrum to obtain initial spectrum characteristics; Residual processing is carried out on the initial map features through a residual module embedded with a feature enhancement type residual attention mechanism, so that enhancement map features are obtained; And sequentially performing global average pooling, full connection and normalization on the enhanced spectrum features to obtain discharge type probability distribution, wherein the discharge type probability distribution is used for representing the discharge type probability of each preset discharge type corresponding to the discharge defect feature, and determining that the maximum discharge type probability in the discharge type probability distribution is the target discharge type probability.
7. The method according to claim 1, further comprising, after step S4: Triggering a redundancy cleaning mechanism whenever the number of the discharge defect characteristics newly added by the same discharge type cluster reaches a preset value or at regular time, wherein the discharge type cluster comprises the preset discharge type cluster and the new discharge type cluster; The redundancy cleaning mechanism comprises the following steps: Analyzing the distribution density of defect data in the same discharge type cluster by utilizing nuclear density estimation, dividing the same discharge type cluster into areas with different densities based on the peak value of the distribution density, wherein the areas with different densities comprise a dense area, a transition area and a sparse area with the densities sequentially reduced; setting differential redundancy reduction threshold values for areas with different densities, and calculating weighted Euclidean distance of the defect data; And clustering the defect data in the same discharge type cluster according to the weighted Euclidean distance, and performing redundancy elimination according to a clustering result and a corresponding redundancy reduction threshold value.
8. The method according to claim 1, wherein step S5 comprises: analyzing access information of different indexes in real time, and calculating the priority score according to weight coefficients of the access information, wherein the access information comprises index access frequency, inquiry time consumption and hit rate; And carrying out scheduling management on different indexes according to the priority score, wherein the scheduling management comprises corresponding storage management on the different indexes based on the priority score.
9. The method of claim 8, wherein the corresponding storage management of different indexes based on the priority score comprises: arranging the different indexes in descending order of the priority scores; marking the first N indexes in the sequencing result as high-priority indexes, and loading the high-priority indexes into a memory cache; And storing the index (n+1) and the indexes after the n+1 in the sequencing result in a disk.
10. A partial discharge data management apparatus for a high voltage cable, comprising: The acquisition module is configured to acquire multi-mode original defect signals, and perform feature extraction on the original defect signals to obtain initial defect features; The noise probability calculation module is configured to input the initial defect characteristics into a pre-constructed multi-scale noise self-adaptive separation network for processing to obtain target noise characteristics, and calculate target noise probability of the target noise characteristics; The noise judgment module is configured to store the target noise characteristics into a noise library when the target noise probability is larger than a preset noise probability threshold value, and otherwise, perform signal purification on the initial defect characteristics according to a corresponding filtering strategy to obtain discharge defect characteristics; The defect data processing module is configured to judge whether the similarity of the discharge defect characteristics and each preset discharge type cluster in the discharge type library meets preset similarity requirements, calculate target discharge type probability of the discharge defect characteristics when the similarity meets the preset similarity requirements, determine the target discharge type cluster of the discharge defect characteristics based on the target discharge type probability and output corresponding target discharge types; otherwise, storing the discharge defect characteristics into a buffer area to be verified, establishing a new discharge type cluster according to the discharge defect characteristics when the characteristic data in the buffer area to be verified meets a preset new cluster creation condition, and dynamically updating the discharge type library according to the data increment and the characteristic cluster similarity in the new discharge type cluster; And the index management module is configured to calculate an index priority score according to the access information of the discharge type library and conduct hierarchical management on indexes according to the priority score.

Description

Partial discharge data management method and device for high-voltage cable Technical Field The invention relates to the technical field of power equipment state monitoring, in particular to a partial discharge data management method and device for a high-voltage cable. Background The partial discharge signal (PARTIAL DISCHARGE, PD) is an early feature of high voltage cable insulation defects. However, with the development of smart grids, partial discharge data has a tendency of "mass growth", and the conventional partial discharge data management mode has a significant limitation when facing such a new data form: At the data characteristic level, partial discharge signals are easy to be subjected to mixed interference of multiple types of noise such as field complex electromagnetic environment, environmental background sound wave and the like, the time domain and frequency domain characteristics of different noise are obvious, and the partial discharge signals and effective signals are often highly coupled on the time domain and the frequency domain, so that the traditional filtering method is difficult to realize accurate separation, meanwhile, the patterns of similar defects under different working conditions are different, and the complexity of sample classification is increased. In the data storage and management layer, a large number of similarity maps exist in mass data (like a defect that data are repeatedly monitored under the same working condition), a traditional database lacks a targeted redundancy processing mechanism, so that storage resources are occupied inefficiently, and as data are accumulated, the redundancy is continuously expanded, and a large amount of storage resources and calculation cost are consumed. At the intelligent analysis model level, the existing defect recognition model depends on historical sample training, but in practice, the cable discharge type may have new characteristics along with the change of running time and environment, the traditional model needs to be updated manually and regularly, and is difficult to adapt to the monitoring requirement of dynamic change, so that the recognition accuracy is gradually reduced. Therefore, those skilled in the art are working to develop a high-voltage cable partial discharge data management method and an index management method with adaptive separation of multiple types of noise, free storage of massive redundant data and online iterative evolution capability of a discharge type identification model. Disclosure of Invention In view of the above-mentioned drawbacks of the prior art, the present application provides a method and apparatus for managing partial discharge data of a high voltage cable, so as to comprehensively improve the quality, storage efficiency and intelligent analysis accuracy of the partial discharge data, and provide a solid data base and technical support for lean operation and maintenance of power grid equipment. The application provides a partial discharge data management method of a high-voltage cable, which comprises the steps of obtaining a multi-mode original defect signal, carrying out feature extraction on the original defect signal to obtain an original defect feature, inputting the original defect feature into a pre-built multi-scale noise self-adaptive separation network to be processed to obtain a target noise feature, calculating the target noise probability of the target noise feature, storing the target noise feature into a noise library when the target noise probability is larger than a preset noise probability threshold value, otherwise, carrying out signal purification on the original defect feature according to a corresponding filtering strategy to obtain a discharge defect feature, judging whether the similarity of the discharge defect feature and each preset discharge type cluster in the discharge type library meets the preset similarity requirement, calculating the target discharge type probability of the discharge defect feature when the similarity meets the preset similarity requirement, determining the target discharge type cluster of the discharge defect feature based on the target discharge type probability, outputting the corresponding target discharge type, otherwise, storing the discharge defect feature into a buffer to be verified, establishing a new discharge type cluster according to the discharge defect feature when the feature data in the buffer to be verified meets the preset new cluster creation condition, calculating the access information of the priority index and carrying out priority index score management. The application provides a local discharge data management device of a high-voltage cable, which is characterized by comprising an acquisition module, a noise probability calculation module, a noise judgment module, a defect data processing module and a priority index management module, wherein the acquisition module is configured to acquire multi-mode original defe