CN-121997252-A - Multi-mode data fusion method and system for extremely high wind of power transmission line

CN121997252ACN 121997252 ACN121997252 ACN 121997252ACN-121997252-A

Abstract

The invention discloses a multi-mode data fusion method and a system for extremely high wind of a power transmission line, which are characterized in that data of different modes of atmospheric image data, remote sensing images and terrain features are collected and projected and overlapped, and forming standardized data cubes, dividing each data cube into space-time image block cubes through zero padding and three-dimensional convolution operation, and converting the space-time image block cubes into characteristic unit sequences through linear transformation. And adding position codes for each characteristic sequence by taking the network coordinates of the data blocks as indexes to generate a final weather, remote sensing and topography characteristic unit sequence. And then, the model respectively extracts the common flow characteristics representing the shared background and the unique flow characteristics reflecting the characteristics of the data source from the sequences, and finally, intelligent fusion is carried out by taking the common flow characteristics as key values and the unique flow characteristics as queries, so that multi-mode data fusion is realized, and the reliability of the data fusion result is effectively improved.

Inventors

LIU DONGHUI
XUE JIARUI
YANG QIXUAN
Mu Weiguang
LI DANLI
ZHANG JIYONG
WANG HAO
XUE ZHENYU
ZHAO CHUNHUI
ZHANG AIYUAN
LV JINGGUO
YU GAO
WU XINPING

Assignees

国网经济技术研究院有限公司
北京建筑大学

Dates

Publication Date: 20260508
Application Date: 20260105

Claims (10)

1. The multi-mode data fusion method for extremely high wind of the power transmission line is characterized by comprising the following steps of: Acquiring atmospheric weather data, remote sensing data and topographic data of an area to be monitored; Respectively carrying out projection superposition on the meteorological data, the remote sensing data and the topographic data to obtain respective corresponding data cubes; Performing zero filling and convolution segmentation operation on each data cube to obtain space-time image block cubes corresponding to each data cube; performing linear transformation on each space-time image block cube to obtain a corresponding characteristic unit sequence; Adding position codes to the corresponding characteristic unit sequences by taking network coordinates of each data block in each space-time image block cube as a space-time index to obtain corresponding final characteristic unit sequences, wherein the corresponding final characteristic unit sequences comprise final weather characteristic unit sequences, final remote sensing characteristic unit sequences and final topography characteristic unit sequences; According to the obtained final feature unit sequences, respectively performing unique stream extraction and common stream extraction to obtain common stream features and unique stream features; And fusing the unique stream features serving as the queries by taking the common stream features as key values to obtain a fusion result.
2. The method for multi-modal data fusion for extreme wind of power transmission line according to claim 1, further comprising, when the corresponding final characteristic unit sequence is obtained: Carrying out terrain dynamic parameter projection on the terrain data cubes in each data cube to obtain high-order terrain dynamic characteristics; And adding the high-order topographic dynamic features to the corresponding feature unit sequences to obtain final topographic feature unit sequences.
3. The method for multi-modal data fusion for extreme wind of power transmission line according to claim 1, wherein the steps of performing unique flow extraction and common flow extraction according to the obtained final feature unit sequences, respectively, to obtain common flow features and unique flow features, include: Extracting unique flow from the final remote sensing characteristic unit sequence to obtain cloud cluster dynamic evolution characteristics; extracting the unique flow of the final topographic feature unit sequence to obtain topographic features; extracting the unique flow of the final meteorological feature unit sequence to obtain meteorological features; Determining the cloud dynamic evolution feature, the topography feature, and the meteorological feature as unique flow features; And carrying out common flow characteristic extraction based on the final remote sensing characteristic unit sequence, the final topography characteristic unit sequence and the final meteorological characteristic unit sequence to obtain common flow characteristics.
4. The method for multi-mode data fusion for extremely high wind of power transmission line according to claim 3, wherein the extracting the unique flow from the final remote sensing feature unit sequence to obtain cloud dynamic evolution features comprises the following steps: performing feature analysis on the final remote sensing feature unit sequence to obtain a new topography feature unit sequence; and extracting a cloud dynamic evolution rule in the new topography characteristic unit sequence to obtain a cloud dynamic evolution characteristic.
5. The method for multi-modal data fusion for extreme wind of power transmission line according to claim 3, wherein the extracting the unique flow of the final topographic feature unit sequence to obtain the topographic feature comprises: deconvolution is carried out on the final topography characteristic unit sequence to obtain a first space diagram; Performing gradient convolution in multiple directions by utilizing the first space diagram in multiple directions to obtain multiple convolution results, and obtaining gradient components according to each convolution result; splicing the gradient component and the Laplace elevation residual error to obtain a terrain two-dimensional physical vector; And carrying out convolution projection on the two-dimensional physical vector to obtain the topographic feature.
6. The method for multi-modal data fusion for extreme wind of power transmission line according to claim 3, wherein the extracting the unique stream of the final weather feature unit sequence to obtain weather features comprises: deconvolution is carried out on the final meteorological feature unit sequence to obtain a second space diagram; Based on the second space diagram, a meteorological two-dimensional physical vector is obtained; Performing gust jump feature extraction according to the meteorological two-dimensional physical vector to obtain attention force diagram; the attention is subjected to convolution projection to obtain meteorological features.
7. The method for multi-modal data fusion for extreme wind of power transmission line according to claim 3, wherein the performing the common flow feature extraction based on the final remote sensing feature unit sequence, the final topography feature unit sequence and the final meteorological feature unit sequence to obtain a common flow feature comprises: Carrying out space-time skeleton extraction on the final remote sensing characteristic unit sequence, the final topographic characteristic unit sequence and the final meteorological characteristic unit sequence to obtain a coordinate space-time skeleton; Performing cross-source covariance guiding weight operation based on the final remote sensing characteristic unit sequence, the final topography characteristic unit sequence and the final meteorological characteristic unit sequence to obtain gating weight; Multiplying the gating weight and the coordinate space-time skeleton element by element to obtain a common guide signal; The common pilot signal is determined to be a common flow characteristic.
8. The utility model provides a multimode data fusion system towards transmission line extremely strong wind which characterized in that includes: the acquisition module is used for acquiring atmospheric weather data, remote sensing data and topographic data of the area to be monitored; the projection superposition module is used for respectively carrying out projection superposition on the meteorological data, the remote sensing data and the topographic data to obtain respective corresponding data cubes; The convolution segmentation module is used for performing zero filling and convolution segmentation operation on each data cube to obtain space-time image block cubes corresponding to each data cube; The linear transformation module is used for carrying out linear transformation on each space-time image block cube to obtain a corresponding characteristic unit sequence; The position coding module is used for adding position codes to the corresponding characteristic unit sequences by taking the network coordinates of each data block in each space-time image block cube as a space-time index to obtain corresponding final characteristic unit sequences, wherein the corresponding final characteristic unit sequences comprise final weather characteristic unit sequences, final remote sensing characteristic unit sequences and final topography characteristic unit sequences; The feature extraction module is used for respectively extracting the unique flow and the common flow according to the obtained final feature unit sequences to obtain the common flow feature and the unique flow feature; and the fusion module is used for fusing the unique stream features serving as the queries by taking the common stream features as key values to obtain a fusion result.
9. A computer device, comprising: A memory for storing a computer program; A processor for implementing the multi-modal data fusion method for extreme wind of power transmission line according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the transmission line extreme wind oriented multi-modal data fusion method according to any one of claims 1 to 7.

Description

Multi-mode data fusion method and system for extremely high wind of power transmission line Technical Field The invention relates to the technical field of disaster prevention of power systems, in particular to a multi-mode data fusion method and system for extremely high wind of a power transmission line. Background The traditional method and the machine learning-based method, such as the traditional method for fusing the high wind data, mainly adopts optimal interpolation or three-dimensional variation to fuse the sea surface wind field of the scatterometer with the ground weather station observation, and outputs 0.25-0.5 degree lattice point products. Or adopting shallow models such as random forests, extreme gradient lifting, supervised learning algorithms and the like, carrying out nonlinear regression on satellite cloud top bright temperature, radar reflectivity, climate analysis data sets and site wind gusts, and realizing thunderstorm high wind identification with a hit rate of about 80%. The method can effectively integrate the characteristics of the strong wind multi-source data. However, when the method is oriented to mountainous terrain, due to the characteristics of large terrain drop and small site density, the method mainly uses weighted average or splicing to directly fuse, ignores the unique characteristics and shared characteristics of multi-source data, and therefore results in the loss of partial information of the data fusion result, the fusion result deviates from the actual situation, and the reliability is reduced. Disclosure of Invention In order to solve the technical problems, the embodiment of the invention provides a multi-mode data fusion method and a system for extremely high wind of a power transmission line, which aim to solve the technical problem that the reliability of a data fusion result obtained by the existing multi-mode data fusion method for extremely high wind of the power transmission line is low. The first aspect of the embodiment of the invention provides a multi-mode data fusion method for extremely high wind of a power transmission line, which comprises the following steps: Acquiring atmospheric weather data, remote sensing data and topographic data of an area to be monitored; respectively carrying out projection superposition on meteorological data, remote sensing data and topographic data to obtain respective corresponding data cubes; Performing zero filling and convolution segmentation operation on each data cube to obtain space-time image block cubes corresponding to each data cube; Performing linear transformation on each space-time image block cube to obtain a corresponding characteristic unit sequence; Adding position codes to the corresponding characteristic unit sequences by taking network coordinates of each data block in each space-time image block cube as a space-time index to obtain corresponding final characteristic unit sequences, wherein the corresponding final characteristic unit sequences comprise final weather characteristic unit sequences, final remote sensing characteristic unit sequences and final topography characteristic unit sequences; according to the obtained final feature unit sequences, respectively carrying out unique stream extraction and common stream extraction to obtain common stream features and unique stream features; and fusing the shared stream features serving as key values and the unique stream features serving as queries to obtain a fusion result. In a possible implementation manner of the first aspect, when obtaining the corresponding final feature unit sequence, the method further includes: Carrying out terrain dynamic parameter projection on the terrain data cubes in each data cube to obtain high-order terrain dynamic characteristics; And adding the high-order topographic dynamic features to the corresponding feature unit sequences to obtain final topographic feature unit sequences. In a possible implementation manner of the first aspect, according to the obtained final feature unit sequences, performing unique stream extraction and common stream extraction respectively, to obtain common stream features and unique stream features, including: Extracting unique streams of the final remote sensing characteristic unit sequence to obtain cloud cluster dynamic evolution characteristics; extracting unique flow from the final topographic feature unit sequence to obtain topographic features; Extracting unique streams of the final meteorological feature unit sequence to obtain meteorological features; Determining cloud dynamic evolution characteristics, topography characteristics and meteorological characteristics as unique flow characteristics; and carrying out common flow characteristic extraction based on the final remote sensing characteristic unit sequence, the final topography characteristic unit sequence and the final meteorological characteristic unit sequence to obtain common flow characteristics. In a possible implem