A lead state fusion perception method based on multi-modal learning

By fusing electrical, environmental, and visual data through a multimodal learning method and using C2f_DCNv3 convolutional networks and dynamic head DyHead networks for feature extraction, the problem of insufficient sensing capability in transmission line conductor condition monitoring is solved, and more accurate condition identification and fault location are achieved.

CN122196901APending Publication Date: 2026-06-12JREN INFORMATION TECH LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
JREN INFORMATION TECH LTD
Filing Date
2026-03-12
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

In existing technologies, transmission line conductor condition monitoring mainly relies on a single type of data, which is difficult to fully reflect complex changes in operating status, resulting in limited sensing capabilities and difficulty in achieving accurate fault location and early warning.

⚗Method used

A multimodal learning approach is adopted, combining electrical data, environmental data, visual data, and fault location data. Feature extraction and fusion processing are performed through a joint detection network of C2f_DCNv3 convolution and dynamic head DyHead, so as to achieve comprehensive perception of conductor status and fault early warning.

🎯Benefits of technology

It improves the integrity and reliability of conductor status perception, enhances the ability to extract visual features, improves the accuracy of status recognition and the precision of fault location, and realizes integrated processing of conductor status monitoring, fault location and early warning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122196901A_ABST
    Figure CN122196901A_ABST
Patent Text Reader

Abstract

The application discloses a kind of based on multi-modal learning's wire state fusion perception method, including the following steps: generating multi-modal wire state dataset;Carrying out feature extraction, introduce the joint detection network of C2f_DCNv3 convolution and dynamic head DyHead, generate multi-modal feature data;Utilize multi-modal learning model to carry out fusion processing, obtain wire state fusion feature data;Carrying out wire state recognition, form wire state recognition result;High-speed sampling and waveform recording are carried out to fault location data, and distributed fault location calculation is carried out, obtain fault interval positioning result and fault position result;Generating wire state monitoring information and fault early warning information;Wire state monitoring information and fault early warning information are uploaded to background main station system and are stored and displayed, realize the comprehensive perception of wire operating state, state recognition, fault location and early warning information generation, improve the accuracy of wire state recognition.
Need to check novelty before this filing date? Find Prior Art