CN-122026609-A - Intelligent rapid wiring monitoring method for distribution branch box

CN122026609ACN 122026609 ACN122026609 ACN 122026609ACN-122026609-A

Abstract

The intelligent rapid wiring monitoring method for the distribution branch box comprises the steps of 1, carrying out multi-mode heterogeneous sensing collection and hierarchical heterogeneous feature purification pretreatment on electric parameters and physical form data related to wiring of the distribution branch box to generate a standardized multi-mode heterogeneous fusion feature vector, 2, carrying out cross-mode related analysis and intelligent wiring state research and judgment through an attention mechanism enhanced multi-branch convolutional neural network based on the standardized multi-mode heterogeneous fusion feature vector to generate wiring state research and judgment results meeting preset research and judgment standards, and 3, triggering a hierarchical closed-loop early warning mechanism and heterogeneous terminal to be pushed cooperatively according to the wiring state research and judgment results meeting the preset research and judgment standards to generate multi-dimensional wiring abnormality early warning information.

Inventors

CHEN JINGANG
HU YUTAO
GU CHUNYU
CHEN JIANWEI
LI YU
WU JIATAO
ZHANG WEIWEI
WANG BO
ZHANG JIACONG
ZHANG JUNHUI

Assignees

北京创万通伟科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260113

Claims (10)

1. The intelligent rapid wiring monitoring method for the distribution branch box is characterized by comprising the steps of 1, carrying out multimode heterogeneous sensing collection and layering heterogeneous feature purification pretreatment on electrical parameters and physical form data related to wiring of the distribution branch box to generate a standardized multimode heterogeneous fusion feature vector, 2, carrying out cross-modal related analysis and intelligent wiring state research and judgment through an attention mechanism enhanced multi-branch convolutional neural network based on the standardized multimode heterogeneous fusion feature vector to generate a wiring state research and judgment result meeting a preset research and judgment standard, and 3, triggering a hierarchical closed-loop early warning mechanism and heterogeneous terminal to be pushed in a cooperative mode according to the wiring state research and judgment result meeting the preset research and judgment standard so as to generate multi-dimensional wiring abnormality early warning information.
2. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 1, wherein step 1 comprises: step 11, performing synchronous sensing acquisition of electrical parameters, which meet preset acquisition specifications, on connection positions of connecting terminals and wires of a distribution branch box through a heterogeneous sensor differential deployment strategy so as to generate a heterogeneous electrical parameter bimodal original sensing matrix; Step 12, performing environment adaptation acquisition on an internal wiring area of a distribution branch box through a multi-angle heterogeneous image sensor so as to generate a heterogeneous physical form multi-view original imaging array; and 13, carrying out layered filtering denoising, heterogeneous feature extraction and normalization calibration processing on the heterogeneous electric parameter bimodal original perception matrix and the heterogeneous physical form multi-view original imaging array so as to generate a standardized multimodal heterogeneous fusion feature vector.
3. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 2, wherein step 11 comprises: step 111, forming a probe type contact resistance sensor which contacts each wiring terminal of the distribution branch box and meets the requirement of a preset contact gap, and executing contact resistance signal sensing acquisition according to a rule to generate terminal contact resistance raw data; step 112, executing non-invasive type micro-current signal synchronous acquisition on the buckle type micro-current sensor clamped at the connection part of each wiring loop wire of the distribution branch box so as to generate loop micro-current original data; And 113, performing heterogeneous association labeling and data integration on the terminal contact resistance raw data and the loop micro-current raw data according to the electrical parameter type so as to generate a heterogeneous electrical parameter bimodal raw sensing matrix.
4. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 2, wherein step 12 comprises: Step 121, performing multi-angle full coverage shooting by using an industrial image sensor deployed in an internal wiring area of a distribution branch box so as to generate initial wiring form image data; Step 122, based on the initial wiring form image data, starting an infrared light supplementing self-adaptive adjustment and anti-fog and noise removal processing mechanism built in the industrial image sensor, and executing environment adaptive image optimization on the initial wiring form image data to generate wiring form image data meeting a preset definition standard; And step 123, performing time sequence heterogeneous association storage and integration on the initial wiring form image data and the wiring form image data meeting the preset definition standard according to the shooting angle so as to generate a heterogeneous physical form multi-view original imaging array.
5. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 2, wherein step 13 comprises: Step 131, performing sliding window self-adaptive filtering interference elimination processing on the heterogeneous electrical parameter bimodal original sensing matrix, and extracting key electrical heterogeneous characteristics from the filtered heterogeneous electrical parameter bimodal original sensing matrix through a dynamic characteristic quantity extraction algorithm to generate a standardized electrical heterogeneous characteristic vector; Step 132, performing Gaussian adaptive filtering noise reduction processing on the heterogeneous physical form multi-view original imaging array, and extracting key form heterogeneous features from the noise-reduced heterogeneous physical form multi-view original imaging array through an edge detection and contour rule extraction algorithm to generate standardized form heterogeneous feature vectors; And 133, performing dimension unified calibration and interval normalization processing on the standardized electrical heterogeneous feature vector and the standardized morphological heterogeneous feature vector to generate a standardized multi-modal heterogeneous fusion feature vector.
6. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 1, wherein step 2 comprises: Step 21, based on the constructed attention mechanism enhanced multi-branch convolutional neural network architecture, performing differential mapping and dynamic weight distribution on the electric heterogeneous characteristics and the morphological heterogeneous characteristics to generate a cross-modal fusion research and judgment model framework; Step 22, based on the multi-scene heterogeneous wiring sample library, performing batch offline training and parameter optimization on the cross-modal fusion research model framework to generate an initial wiring state intelligent research model; And step 23, inputting the standardized multi-mode heterogeneous fusion feature vector into an optimized intelligent wiring state research and judgment model, and executing intelligent wiring state research and judgment on the standardized multi-mode heterogeneous fusion feature vector by combining a dual mechanism of rule base logic verification and historical baseline feature comparison so as to generate a wiring state research and judgment result which accords with a preset research and judgment standard.
7. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 6, wherein step 21 comprises: Step 211, performing feature dimension mapping and strengthening treatment on the standardized electrical heterogeneous feature vector through the constructed full connection layer to generate an electrical heterogeneous feature mapping vector; Step 212, performing morphological feature mining processing on the standardized morphological feature vector through the constructed convolution layer-pool layer combination branch to generate a morphological feature mapping vector; And 213, performing dynamic weight distribution on the electric heterogeneous feature mapping vector and the morphological heterogeneous feature mapping vector through the attention mechanism calculation module, and then performing cross-modal feature stitching fusion to generate a cross-modal fusion research model framework.
8. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 6, wherein step 22 comprises: Step 221, collecting multi-mode heterogeneous monitoring data under different wire types, wiring modes and environmental conditions, and executing wiring state association labeling on the multi-mode heterogeneous monitoring data to generate a multi-scene heterogeneous wiring sample library; Step 222, based on the multi-scene heterogeneous wiring sample library, performing batch offline training and parameter optimization on the cross-modal fusion research model framework to generate an initial wiring state intelligent research model; and 223, receiving the multi-mode heterogeneous monitoring data acquired in real time and the manual labeling result, and executing parameter increment update on the initial wiring state intelligent research and judgment model by adopting a small-batch gradient descent iteration method so as to generate an optimized wiring state intelligent research and judgment model.
9. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 6, wherein step 23 comprises: Step 231, inputting the standardized multi-mode heterogeneous fusion feature vector into an optimized intelligent wiring state research model, and performing wiring state probability reasoning and research on the standardized multi-mode heterogeneous fusion feature vector to generate a preliminary wiring state research result; step 232, calling a preset wiring state judgment rule base, performing logic verification on the preliminary wiring state judgment result, and eliminating reasoning conflict to generate a wiring state result after rule verification; And 233, calling historical baseline characteristic data of similar wiring of the distribution branch box, and performing similarity comparison on the characteristic corresponding to the wiring state result after rule verification and the historical baseline characteristic to generate a wiring state judging result which meets the preset judging standard.
10. The intelligent rapid wiring monitoring method of a distribution branch box according to claim 1, wherein step 3 comprises: Step 31, triggering an audible and visual linkage alarm and state corresponding indication device arranged in the distribution branch box according to the state category in the wiring state research result which accords with the preset research standard so as to generate a local closed-loop early warning signal; Step 32, based on the local closed-loop early warning signal, generating a remote cooperative early warning signal with a feedback confirmation mark through a double-link redundancy transmission architecture formed by the industrial Ethernet and wireless communication; And step 33, fusing the local closed-loop early warning signal and the remote cooperative early warning signal with the feedback confirmation mark, and executing differential integration and pushing strategy adaptation on the abnormal information according to a preset abnormal grade division rule so as to generate multi-dimensional wiring abnormal early warning information.

Description

Intelligent rapid wiring monitoring method for distribution branch box Technical Field The application relates to the technical field of monitoring of power distribution equipment, in particular to an intelligent rapid wiring monitoring method for a power distribution branch box. Background Along with the continuous expansion of the construction scale of the power distribution network, the distribution branch box is used as a key node of the power distribution network, and the safety and reliability of the wiring state of the distribution branch box directly influence the stable operation of the power distribution network. The wiring terminal of the distribution branch box is easily affected by factors such as environmental erosion, mechanical vibration and the like to have the problems of loosening, poor contact and the like, so that the wiring state of the distribution branch box needs to be monitored accurately in real time, and accidents such as line faults, power supply interruption and the like caused by abnormal wiring are avoided. In the existing distribution branch box wiring monitoring scheme, a single electrical parameter monitoring architecture is generally adopted, electrical data of a wiring loop are collected through deployment of current and voltage sensors, and whether the wiring state is abnormal or not is judged based on a fixed threshold value. The scheme includes that current and voltage sensors are deployed in a key loop of a distribution branch box, loop electric parameters are collected in real time, collected data are uploaded to an area monitoring terminal and compared with a preset parameter threshold value, whether wiring is abnormal or not is judged according to a comparison result, and a simple alarm signal is generated. However, this single electrical parameter monitoring scheme has significant technical drawbacks. Meanwhile, the abnormal judgment of the scheme depends on a fixed threshold value, is difficult to adapt to parameter fluctuation under different operation conditions, lacks a multi-dimensional early warning mechanism, has low timeliness and effectiveness of abnormal early warning, and cannot meet the requirement of accurate monitoring of the wiring state of the distribution branch box. Disclosure of Invention In order to solve the technical problems, the application provides an intelligent rapid wiring monitoring method for a distribution branch box, which is used for at least alleviating the technical problems. An intelligent rapid wiring monitoring method for a distribution branch box comprises the following steps: Step 1, carrying out multimode heterogeneous sensing acquisition and hierarchical heterogeneous feature purification pretreatment on electric parameters and physical form data associated with wiring of a distribution branch box so as to generate a standardized multimode heterogeneous fusion feature vector; Step 2, based on the standardized multi-mode heterogeneous fusion feature vector, performing cross-mode association analysis and intelligent wiring state research and judgment through an attention mechanism enhanced multi-branch convolutional neural network to generate a wiring state research and judgment result which accords with a preset research and judgment standard; And step 3, triggering a hierarchical closed-loop early warning mechanism to be pushed cooperatively with the heterogeneous terminal according to a wiring state research and judgment result meeting a preset research and judgment standard so as to generate multidimensional wiring abnormality early warning information. Optionally, step 1 includes: step 11, performing synchronous sensing acquisition of electrical parameters, which meet preset acquisition specifications, on connection positions of connecting terminals and wires of a distribution branch box through a heterogeneous sensor differential deployment strategy so as to generate a heterogeneous electrical parameter bimodal original sensing matrix; Step 12, performing environment adaptation acquisition on an internal wiring area of a distribution branch box through a multi-angle heterogeneous image sensor so as to generate a heterogeneous physical form multi-view original imaging array; and 13, carrying out layered filtering denoising, heterogeneous feature extraction and normalization calibration processing on the heterogeneous electric parameter bimodal original perception matrix and the heterogeneous physical form multi-view original imaging array so as to generate a standardized multimodal heterogeneous fusion feature vector. Optionally, step 11 includes: step 111, forming a probe type contact resistance sensor which contacts each wiring terminal of the distribution branch box and meets the requirement of a preset contact gap, and executing contact resistance signal sensing acquisition according to a rule to generate terminal contact resistance raw data; step 112, executing non-invasive type micro-current signal synchronous ac