CN-122023874-A - Method, system, equipment and medium for identifying micro-topography of power transmission line based on multi-mode attention mechanism

CN122023874ACN 122023874 ACN122023874 ACN 122023874ACN-122023874-A

Abstract

The invention discloses a method, a system, equipment and a medium for identifying micro-topography of a power transmission line based on a multi-modal attention mechanism, which comprise the steps of acquiring power transmission line space corridor data and multi-source geographic information data of a target area, preprocessing the data, and respectively generating a line characteristic diagram and a multi-channel ground characteristic diagram; the method comprises the steps of constructing a double-branch deep learning network model, inputting a line characteristic diagram into a topographic characteristic branch, inputting a multichannel ground characteristic diagram into the line characteristic branch, respectively extracting deep characteristic information, weighting the characteristics extracted by the topographic characteristic branch by utilizing a space attention weight generated by extracting the characteristics of the line characteristic branch through a cross-branch space attention module to obtain a fusion characteristic, inputting the fusion characteristic into a decoder for up-sampling and characteristic reconstruction, and outputting a micro-topography classification recognition result aiming at the line of a power transmission line corridor. The invention realizes full-flow intelligent processing, remarkably improves the recognition efficiency, and is suitable for large-scale line census and risk assessment.

Inventors

Nie Xianglun
HUANG TIANYING
HUANG HUAJIANG
JIANG LONG
WEI LUJUN
HOU YONGHONG
JIANG JIBIN
LI YUHANG
QI YANBO
WANG YOUJUN
ZHAO WEI
WU YU
Xiong Ganggang
Ou Liangcheng
YIN CHENGYI
ZHOU ZHENLIN
MAO WEI
LI YI
ZHANG YAO
NIE JING
ZHENG XIAOHU
LIU QING
CAI DENGSHENG
GUO YANZHAO

Assignees

贵州电网有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251226

Claims (10)

1. The method for identifying the micro-topography of the power transmission line based on the multi-mode attention mechanism is characterized by comprising the following steps of: acquiring transmission line space corridor data and multisource geographic information data of a target area, preprocessing, and respectively generating a line characteristic map and a multichannel ground characteristic map; Constructing a double-branch deep learning network model, inputting the line feature map into a topographic feature branch, inputting the multi-channel ground feature map into the line feature branch, and respectively extracting deep feature information; Weighting the features extracted by the topographic feature branches by using a cross-branch spatial attention module and using spatial attention weights generated by the line feature branch extraction features to obtain fusion features; And inputting the fusion characteristics into a decoder for up-sampling and characteristic reconstruction, and outputting micro-topography classification recognition results aiming at the power transmission line corridor along the line.
2. The method for identifying micro-topography of a power transmission line based on a multi-modal attentiveness mechanism as defined in claim 1, wherein the generating of the line signature includes: carrying out rasterization processing on vector data corresponding to the power transmission line space corridor data to generate a binary mask image representing a line path; Calculating Euclidean distance from each pixel point to the nearest line pixel based on the binary mask image, and generating a distance transformation graph; and taking the binary mask image or the distance transformation graph as the line characteristic graph.
3. The method for identifying micro-topography of a power transmission line based on a multi-modal attentiveness mechanism as defined in claim 2, wherein the generating of the multi-channel ground feature map includes: calculating a gradient map, a section curvature map and an elevation variation coefficient map through geographic information system software; And carrying out spatial alignment and normalization processing on the multi-source geographic information data containing the digital elevation model, the slope map, the section curvature map and the elevation variation coefficient map, and then stacking along the channel dimension to generate the multi-channel ground characteristic map.
4. The method for identifying micro-topography of a power transmission line based on a multi-modal attention mechanism as set forth in claim 3 wherein said constructing a dual-branch deep learning network model includes: the dual-branch deep learning network model adopts an encoder-decoder structure, and the topographic feature branch and the line feature branch jointly form an encoder part; parallel convolution branches containing convolution kernels with different sizes are arranged in the topographic feature branches and are used for extracting multi-scale topographic features through step-by-step convolution and downsampling operation; the line characteristic branches are identical to the topographic characteristic branches in structure and are used for extracting the spatial distribution and trend characteristics of the power transmission line through step-by-step convolution and downsampling operation.
5. The method for identifying micro-topography of a power transmission line based on a multi-modal attentiveness mechanism as defined in claim 4, wherein said obtaining a fusion feature includes: At each level of the encoder, carrying out channel compression and nonlinear mapping on the characteristic map output by the line characteristic branch, and generating a space attention weight map consistent with the size of the topographic characteristic map; And multiplying the space attention weight graph with the corresponding level characteristic graph output by the topographic characteristic branch element by element to obtain a weighted fusion characteristic graph.
6. The method for identifying micro-topography of a transmission line based on a multi-modal attentiveness mechanism as defined in claim 5, wherein said inputting said fused features into a decoder for upsampling and feature reconstruction includes: Inputting the weighted fusion feature images of all the layers into a decoder through jump connection; The spatial resolution of the feature map is restored by gradually performing up-sampling operation and fusing shallow features; And adjusting the channel number to the preset micro-topography class number through convolution at the tail layer of the decoder, and outputting a pixel-level multi-class probability map after the processing of an activation function.
7. The method for identifying micro-topography of a power transmission line based on a multi-modal attentiveness mechanism as defined in claim 6, wherein said outputting the identification result for the micro-topography classification along the power transmission line corridor includes: Carrying out maximum value judgment on the pixel-level multi-category probability map along the channel dimension, and determining the micro-topography category to which each pixel belongs; and generating a pixel-level semantic segmentation map according to the micro-topography category to which each pixel belongs, wherein different pixel values correspond to at least one category of saddle, canyon, mountain watershed, lifting topography and atypical micro-topography.
8. The transmission line micro-topography recognition system based on the multi-mode attention mechanism, which applies the transmission line micro-topography recognition method based on the multi-mode attention mechanism as claimed in any one of claims 1 to 7, is characterized by comprising the following steps: The data processing module is used for acquiring the space corridor data and the multisource geographic information data of the power transmission line in the target area and preprocessing the space corridor data and the multisource geographic information data to respectively generate a line characteristic diagram and a multichannel ground characteristic diagram; The deep feature extraction module is used for constructing a double-branch deep learning network model, inputting the line feature map into a topographic feature branch, inputting the multichannel ground feature map into the line feature branch, and respectively extracting deep feature information; the attention fusion module is used for weighting the features extracted by the topographic feature branches by using the spatial attention weights generated by the line feature branch extraction features through the cross-branch spatial attention module to obtain fusion features; And the identification output module is used for inputting the fusion characteristics into a decoder for up-sampling and characteristic reconstruction and outputting micro-topography classification identification results aiming at the line of the power transmission line corridor.
9. An electronic device comprising a memory and a processor, wherein the memory is configured to store computer-executable instructions, and the processor, when executing the computer-executable instructions, implements the steps of the method for identifying micro-topography of a power transmission line based on a multimodal attention mechanism according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon computer executable instructions, wherein the computer executable instructions when executed by a processor implement the steps of the multimode attention mechanism based transmission line micro topography identification method as claimed in any one of claims 1 to 7.

Description

Method, system, equipment and medium for identifying micro-topography of power transmission line based on multi-mode attention mechanism Technical Field The invention relates to the technical field of power system safety, in particular to a method, a system, equipment and a medium for identifying micro-topography of a power transmission line based on a multi-mode attention mechanism. Background The transmission line forms an important part of the power grid system, and the safe and stable operation of the transmission line is directly related to the reliability research of the power grid system. However, as the scale of the power grid construction continues to expand, the newly built power transmission line often needs to traverse mountainous areas with complex terrains and changeable meteorological conditions. However, the micro-topography of the mountain area can have obvious influence on the local wind field, humidity and temperature distribution, and meteorological disasters such as wind deflection, galloping and icing of the transmission conductor are easy to cause, so that the line safety is threatened. The existing method for identifying the micro-topography along the transmission line mainly comprises manual field investigation and geographic information system analysis based on rules. The identification mode of manual investigation is visual, but low in efficiency, high in cost, high in risk, and the result is greatly influenced by human factors, so that the large-range application requirement is difficult to meet. The geographic information system analysis method based on rules usually carries out automatic judgment by setting a threshold value of a terrain factor, and the method has the problems of dependence on expert experience and difficulty in unification of rules although the identification efficiency can be improved. In addition, the geographic information system analysis method based on rules is difficult to effectively distinguish between saddle and canyon terrain types with similar morphologies, and has limited recognition accuracy. In addition, the existing identification method is mainly used for carrying out independent analysis on the topography and the topography, and consideration of the spatial coupling relation between the trend of the power transmission line and the micro-topography is lacking. In fact, the impact of micro-topography on the safety of a line is closely related to its relative position and orientation, such as the line crossing saddles is subject to a large difference in wind conditions from the line laid along valleys. Neglecting the coupling relation can lead to insufficient relevance between the identification result and the actual risk of the line, and is difficult to effectively support disaster prevention, disaster reduction and operation and maintenance work of the power grid. Therefore, a method for identifying the micro-topography of the power transmission line based on a multi-mode attention mechanism is urgently needed to improve the precision of the identification of the micro-topography. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides the method, the system, the equipment and the medium for identifying the micro-topography of the power transmission line based on the multi-mode attention mechanism, which solve the problems that the existing method for identifying the micro-topography excessively depends on manual experience or fixed rules, has low degree of automation and insufficient generalization capability, is difficult to accurately distinguish the micro-topography types similar to the saddle, the canyon and the like, and simultaneously omits the spatial coupling relation between the trend of the power transmission line and the topography, so that the relevance between the identification result and the actual risk of the line is weaker. In order to solve the technical problems, the invention provides the following technical scheme: In a first aspect, the present invention provides a method for identifying micro-topography of a power transmission line based on a multi-modal attention mechanism, including: acquiring transmission line space corridor data and multisource geographic information data of a target area, preprocessing, and respectively generating a line characteristic map and a multichannel ground characteristic map; Constructing a double-branch deep learning network model, inputting the line feature map into a topographic feature branch, inputting the multi-channel ground feature map into the line feature branch, and respectively extracting deep feature information; Weighting the features extracted by the topographic feature branches by using a cross-branch spatial attention module and using spatial attention weights generated by the line feature branch extraction features to obtain fusion features; And inputting the fusion ch