Power transmission line icing state analysis method, system, device and medium based on point cloud modeling and deep learning

Abstract

This invention belongs to the field of power system monitoring technology and discloses a method, system, equipment, and medium for analyzing the icing state of transmission lines based on point cloud modeling and deep learning, thereby overcoming the limitations of existing methods in time series processing. The method includes: dividing the time series point cloud of the transmission line into multiple time subsequences; generating a topological persistence graph for each subsequence through voxelization and persistent coherence analysis; calculating the dynamic topological distance between subsequences based on the persistence graph; extracting topological features to construct a feature matrix, and fusing topological distance, feature space, and feature distribution three views to construct a similarity matrix; using multi-view joint learning to obtain a consensus subspace and performing subsequence clustering within this space to obtain cluster labels; combining physical context information and predefined rules to assign physical pattern labels to the clustering results, generating an icing state classification report and visualizing it. This invention achieves temporal topological quantitative analysis and highly robust state identification of icing morphology.

Description

Technical Field

[0001] This invention belongs to the field of power system monitoring technology, specifically relating to a method, system, equipment, and medium for analyzing the icing status of transmission lines based on point cloud modeling and deep learning. Background Technology

[0002] With the continuous expansion of power systems and the increasingly widespread coverage of high-voltage transmission networks, the monitoring and early warning of icing disasters on transmission lines has become a key technical aspect of ensuring the safe and stable operation of the power grid. In recent years, spatial information acquisition technologies, represented by 3D lidar scanning and automated inspection by unmanned aerial vehicles (UAVs), have developed rapidly, making 3D digital modeling and icing status identification methods for transmission lines based on dense point cloud data a research and application hotspot.

[0003] In existing technologies, point cloud-based icing analysis methods mainly rely on traditional 3D geometric feature extraction and classification algorithms. For example, they calculate features such as local curvature, normal distribution, and density changes in the point cloud, or use morphological filtering to extract the icing contour, and then combine these with machine learning models (such as support vector machines and random forests) for icing identification and thickness estimation. Some studies have further introduced deep learning techniques, using convolutional neural networks (CNNs) or graph neural networks (GNNs) to learn features and classify point clouds or their projected images, aiming to improve the model's automatic recognition capabilities. However, these existing methods still have significant shortcomings in practical applications: on the one hand, these methods mostly focus on the geometric and statistical features of the point cloud, lacking effective quantification and utilization of the topological information inherent in the icing morphology, making it difficult to characterize the topological evolution laws during the icing formation process; on the other hand, these methods typically analyze data independently at a single point in time, failing to fully model the dynamic correlation of icing states over time, resulting in a lack of temporal consistency in the identification results and an inability to accurately reflect the continuous evolution of icing. These shortcomings limit the effectiveness of existing technologies in accurately identifying and reliably warning about icing states.

[0004] In summary, although existing technologies have made some progress in point cloud processing and icing recognition, they still have significant shortcomings in topological structure quantification, in-depth utilization of high-dimensional topological features, and temporal dynamic evolution modeling. Summary of the Invention

[0005] Based on the aforementioned shortcomings and deficiencies in the existing technology, one of the objectives of this invention is to at least solve one or more of the aforementioned problems in the existing technology. In other words, one of the objectives of this invention is to provide a method, system, device, and medium for analyzing the icing state of transmission lines based on point cloud modeling and deep learning that meets one or more of the aforementioned requirements, so as to achieve quantitative characterization of the topological evolution of icing morphology, fully explore the high-dimensional topological characteristics of point cloud data, and effectively characterize the dynamic correlation of icing state in time series, thereby improving the temporal consistency and accuracy of icing state identification results.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for analyzing the icing status of transmission lines based on point cloud modeling and deep learning, comprising the following steps: S1. Divide the time series point cloud of the target transmission line into multiple time subsequences; perform voxelization on the point cloud of each time subsequence to generate a voxel point set and perform persistent cohomology analysis to generate a persistent graph characterizing its topology; calculate the dynamic topological distance between different time subsequences based on the persistent graph. S2. Based on the dynamic topological distance and the persistent graph, extract and construct the topological feature vector of each subsequence to form the original topological feature matrix; construct similarity matrices under the three views of topological distance, feature space and feature distribution respectively; use the similarity matrices under the three views as multi-view input, obtain the consensus subspace through multi-view joint learning, and cluster all time subsequences in the consensus subspace to obtain cluster label vectors; S3. Based on the clustering label vector and its corresponding physical context information, assign a physical pattern label to each cluster according to predefined rules, generate an icing status classification report and display it visually.

[0007] As a preferred approach, step S1 generates a persistent graph representing its topological structure, specifically as follows: The point clouds of each time subsequence are fused and downsampled to generate a voxel point set; Construct a multi-scale Vietoris-Rips complex filtering sequence based on the voxel point set; Persistent cohomology computation is performed on the Vietoris-Rips complex filtering sequence to generate 0-dimensional persistent graphs and 1-dimensional persistent graphs.

[0008] As a preferred approach, the dynamic topological distance between different time subsequences in step S1 is calculated by calculating the multi-feature weighted temporal Wasserstein distance, and the calculation formula is as follows: , In the formula, iand j Indices for the number of two distinct subsequences. and There are two subsequences. For indexes of time points, The time length of the subsequence. For Wasserstein distance, For subsequence and Multi-feature weighted temporal Wasserstein distance For time series weights based on Gaussian distribution, and Representing subsequences respectively and In the A 0-dimensional continuous plot at each time point and Representing subsequences respectively and In the A 1D continuous plot at each time point.

[0009] As a preferred approach, step S2 extracts and constructs the topological feature vector for each subsequence to form the original topological feature matrix, specifically as follows: Primary topological feature statistics are extracted from the 0-dimensional and 1-dimensional persistence plots of each time subsequence and concatenated to form the topological feature vector of that subsequence. The original topological feature matrix is ​​constructed by treating the topological feature vectors of all subsequences as rows.

[0010] As a preferred approach, the consensus subspace is obtained in step S2 through multi-view joint learning, specifically as follows: The similarity matrices under the three views are used as adjacency matrices to smoothly propagate the original topological feature matrix; A multi-view joint objective function is constructed and optimized, which includes a cooperative regularization term to constrain the consistency of projection of different views into the consensus subspace, so as to obtain the consensus subspace basis matrix.

[0011] As a preferred approach, step S2 involves clustering all time subsequences within this space, specifically as follows: Project the original topological features onto the consensus subspace; Spectral embedding is performed on the projected features to obtain a low-dimensional embedding clustering matrix; K-means clustering is performed on the embedded clustering matrix to obtain the cluster label vector.

[0012] As a preferred option: The physical context information includes average icing thickness, average ambient temperature, and average relative humidity. The predefined rules are matching rules set based on the icing physical mechanism, which combine topological features and physical context information with a threshold for comparison.

[0013] Secondly, the present invention provides a transmission line icing state analysis system based on point cloud modeling and deep learning, used to implement the transmission line icing state analysis method as described in the first aspect, including: The data acquisition and topology feature extraction module is used to divide the time series point cloud of the target transmission line into multiple time subsequences; to voxelize the point cloud of each time subsequence to generate a voxel point set and perform persistent cohomology analysis to generate a persistent graph characterizing its topology; and to calculate the dynamic topological distance between different time subsequences based on the persistent graph. The multi-view fusion clustering module is used to extract and construct the topological feature vector of each subsequence based on the dynamic topological distance and the persistent graph, forming the original topological feature matrix; construct similarity matrices under the three views of topological distance, feature space and feature distribution respectively; use the similarity matrices under the three views as multi-view input, obtain the consensus subspace through multi-view joint learning, and cluster all time subsequences in the consensus subspace to obtain cluster label vectors; The state mapping and visualization module is used to assign physical pattern labels to each cluster according to the cluster label vector and its corresponding physical context information, based on predefined rules, to generate an icing state classification report and perform visualization.

[0014] Thirdly, the present invention provides an electronic device, the computer device including a memory, a processor and a computer program, wherein when the computer program is executed by the processor, it implements the transmission line icing status analysis method as described in the first aspect.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the transmission line icing status analysis method as described in the first aspect.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention introduces a persistent cohomology analysis method based on Vietoris-Rips complexes to rigorously quantify the topological structure of point cloud data, overcoming the limitations of traditional methods that rely solely on geometric and statistical features. It can accurately capture the dynamic topological evolution of icing morphology during its formation, development, and melting.

[0017] 2. This invention proposes a multi-feature weighted temporal Wasserstein distance calculation method, which effectively measures the similarity and difference in topological structure of icing morphology in different time periods, realizes the modeling of dynamic correlation between time series point cloud data, and improves the temporal consistency and continuity of icing state identification.

[0018] 3. A multi-view joint learning framework is constructed based on topological distance similarity, feature space similarity, and feature distribution similarity. Through collaborative optimization, the deep fusion of multi-source feature information is achieved, overcoming the recognition limitations under a single feature dimension and significantly improving the accuracy and robustness of icing state classification.

[0019] 4. By using an adaptive clustering and physical context information fusion mechanism, the data-driven clustering results are transformed into icing status labels with clear physical meaning, realizing a reliable mapping from topological features to practical engineering applications, and providing a more interpretable and practical analysis method for transmission line icing monitoring.

[0020] Further or more detailed beneficial effects will be described in conjunction with specific embodiments in the detailed implementation. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the method for analyzing the icing status of transmission lines according to Embodiment 1 of the present invention.

[0023] Figure 2 This is a schematic diagram of the transmission line icing status analysis system described in Embodiment 2 of the present invention.

[0024] Figure 3 This is a structural diagram of the electronic device described in Embodiment 3 of the present invention.

[0025] Icon labels: 300. Electronic devices; 301. Processor; 302. Communication bus; 303. User interface; 304. Network interface; 305. Memory. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0027] In the following description, several embodiments of the present invention are provided. Different embodiments can be substituted or combined. Therefore, the present invention can also be considered to include all possible combinations of the same and / or different embodiments described. Thus, if one embodiment includes features A, B, and C, and another embodiment includes features B and D, then the present invention should also be considered to include embodiments containing one or more other possible combinations of A, B, C, and D, even if such embodiments are not explicitly described in the following text.

[0028] The following description provides examples and does not limit the scope, applicability, or examples set forth in the claims. Changes may be made to the function and arrangement of the described elements without departing from the scope of the invention. Various processes or components may be appropriately omitted, substituted, or added to the various examples. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined into other examples.

[0029] To facilitate a better understanding of the embodiments of the present invention, its application scenarios will be explained before providing a detailed explanation of the specific implementation methods.

[0030] The transmission line icing status analysis method described in this specification is applied to intelligent inspection and icing disaster prevention scenarios for transmission lines, especially in the operation and maintenance management of overhead transmission lines in high-altitude, high-humidity, and icing-prone areas. In these scenarios, the application of this transmission line icing status analysis method aims to achieve automated, refined, and intelligent monitoring and early warning of the icing status of transmission lines. Specifically, this method aims to automatically identify the morphological evolution process of icing by processing periodically collected three-dimensional point cloud data, accurately classify the icing development stages, and output an icing status report with clear physical meaning. This provides timely decision support for power grid operation and maintenance personnel to prevent accidents such as line galloping, flashover, and line breakage caused by icing overload, ensuring the safe and stable operation of the power grid.

[0031] The following is a brief explanation of the point cloud, persistent graph, weighted temporal Wasserstein distance, consensus subspace, physical context information, Vietoris-Rips complex filtering sequence, and embedded clustering matrix involved in several embodiments of this specification: Point cloud refers to a collection of a large number of discrete three-dimensional coordinate points (usually containing X, Y, and Z coordinate values), obtained through techniques such as 3D lidar scanning or photogrammetry. In this invention, it specifically refers to the raw three-dimensional spatial data obtained by periodically and temporally scanning the target transmission line and its attached icing, which is the basic data source for subsequent topology analysis and condition identification.

[0032] Persistence graphs are a visual and mathematical representation of persistent homology analysis. They record the distribution of topological features (such as connected components and voids) at their generation scale (“birth”) and disappearance scale (“death”). In this invention, 0-dimensional persistence graphs represent the evolution of connected components in a point cloud, while 1-dimensional persistence graphs represent the evolution of void structures (such as one-dimensional rings). Points far from the diagonal represent long-lived, structurally stable topological features, serving as key fingerprints for distinguishing different icing morphologies.

[0033] Weighted Time Series Wasserstein Distance: A comprehensive distance used to measure the differences between two time series data in the topological feature space. It calculates the differences in the 0-dimensional and 1-dimensional persistence graphs of the two series at corresponding time points using the Wasserstein distance, and introduces time-based Gaussian distribution weights to emphasize the contribution of the stable state at the center of the time window while weakening the influence of instantaneous fluctuations at both ends. This distance is used to quantify the dynamic evolution differences in the topological structure of icing morphology across different time periods.

[0034] Consensus Subspace: In multi-view learning, a common, low-dimensional representation space is learned from data from multiple different perspectives (views) through joint optimization. In this invention, this space integrates information from three views: topological distance, feature space, and feature distribution. The aim is to preserve consistent and discriminative features across different views, thereby providing a better, less noisy feature representation for subsequent clustering.

[0035] Physical context information refers to the external environment and measurement parameters directly related to the icing state. In this invention, it specifically refers to the average icing thickness, average ambient temperature, and average relative humidity corresponding to each cluster (or subsequence). This information links the data-driven clustering results with the actual physical process and is the key basis for realizing the mapping from topological features to icing physical state labels.

[0036] Vietoris-Rips complex filtering sequence: a standard method for constructing multi-scale topological structures from point cloud data. For a given scale parameter, it connects all points whose distance to each other is less than the scale parameter to form edges and fills them with corresponding simplexes (such as triangles). As the scale parameter gradually increases from 0, a series of nested complexes are generated, forming a "filtering sequence". This sequence records the complete lifecycle of topological features as the scale changes and is the basis for persistent cohomology computation.

[0037] Embedded clustering matrix: In spectral clustering, the matrix is ​​formed by calculating the Laplacian matrix of the affinity matrix and performing eigenvalue decomposition, then arranging the eigenvectors corresponding to the k smallest eigenvalues ​​column-wise. Each row of this matrix corresponds to a new coordinate representation of an original data sample in a low-dimensional spectral space. In this low-dimensional space, the cluster structure of the data becomes more obvious and linearly separable, thus facilitating efficient and accurate clustering using algorithms such as K-means.

[0038] Example 1: This embodiment provides a method for analyzing the icing status of transmission lines based on point cloud modeling and deep learning, such as... Figure 1 As shown, this method systematically realizes the entire process analysis from point cloud data to classification of icing physical states, specifically including the following steps: Step S1 involves constructing the temporal topological features and dynamic difference matrix of the point cloud. This step aims to transform the original 3D point cloud sequence into a mathematical model capable of quantifying the topological evolution of icing morphology. Step S1 specifically includes the following sub-steps: S11. Point cloud data acquisition and subsequence partitioning: Periodic lidar scans were performed on the target transmission line to acquire raw 3D point cloud data over time. The raw point cloud at each time point was sequentially filtered, denoised, and subjected to multi-period coordinate registration to eliminate noise and spatial offset, ensuring geometric consistency between the time-series point clouds. Subsequently, a sliding window was used to divide the preprocessed time-series point cloud into multiple consecutive time subsequences, thus transforming the continuous icing evolution process into structured analysis units.

[0039] S12, Voxelization and Multiscale Topology Modeling: For each time subsequence, the point clouds of all times contained therein are merged into a unified point cloud set. The global density of the point cloud set is calculated, and the voxel grid size is adaptively determined according to the cube root scaling law to construct a regular 3D grid covering the entire point cloud. The centroids of all points in the non-empty grid are used as representative points and downsampled to generate a simplified voxel point set, providing a stable and efficient input for subsequent topology calculations.

[0040] S13. Persistent cohomology analysis generates persistence graphs: Using a set of voxels as the vertex set, the maximum Euclidean distance between any two dissimilar points within the set is calculated and used as the maximum value of the scale parameter. A scaling factor is set based on the average nearest neighbor distance (which can be set empirically; in this embodiment, it is preferably 5), generating an equally spaced scale sequence from 0 to the maximum value of the scale parameter. The equally spaced scale sequence is traversed, and for each interval scale value, all vertex pairs whose Euclidean distance does not exceed the interval scale value are connected. The corresponding simplex (such as edges or triangles) is generated according to the Vietoris-Rips rule, forming a Vietoris-Rips complex filtering sequence that grows with the time scale.

[0041] Based on the Vietoris-Rips complex filtering sequence, a boundary matrix is ​​constructed in the binary domain, with its rows and columns arranged in ascending order of the scale of the first occurrence of the simplex. The element of the boundary matrix is ​​defined as follows: if the simplex corresponding to a row is a direct plane of the simplex corresponding to a column (differing in dimension by 1), then the element is 1; otherwise, it is 0. Subsequently, the boundary matrix is ​​column-wise simplified using a standard persistent pair algorithm. By analyzing the zero and non-zero columns of the simplified matrix, the "birth" and "death" scales of topological features (such as connectors and holes) are identified. All (birth, death) scale pairs of features are collected, and 0-dimensional persistent maps representing the evolution of connectors and 1-dimensional persistent maps representing the evolution of hole structures are plotted. These two maps together constitute the "fingerprint" of the subsequence point cloud topology.

[0042] Step S14: Calculate the dynamic topological distance matrix: To quantify the evolutionary differences in topological structure of icing morphology across different time periods, for any two time subsequences, the multi-feature weighted temporal Wasserstein distance is calculated. This distance comprehensively considers the differences between the 0-dimensional and 1-dimensional persistence maps of the two sequences at each corresponding time point and applies temporal weighting. The specific calculation formula is as follows: , in, and There are two subsequences. i and j Indices for the number of two distinct subsequences. For indexes of time points, The time length of the subsequence. For Wasserstein distance, For subsequence and Multi-feature weighted temporal Wasserstein distance As weight, and Representing subsequences respectively and In the A 0-dimensional continuous plot at each time point and Representing subsequences respectively and In the A 1D continuous plot at each time point.

[0043] Weight Calculated using Gaussian distribution: , where the mean Standard deviation This assigns higher weight to the central moment of the time window to emphasize a stable icing pattern.

[0044] Calculate the distance between all subsequence pairs and construct a dynamic topological distance matrix, which fully characterizes the similarity and differences of the icing topology in the time dimension.

[0045] Step S2 involves multi-view joint learning and adaptive clustering of icing states. Based on the topological features generated in S1, this step integrates multi-source similarity information to accurately classify icing evolution patterns within the consensus subspace. Step S2 specifically includes the following sub-steps: S21. Construct the matrix, including: Topological distance similarity matrix: The dynamic topological distance matrix is ​​transformed by applying the RBF kernel function to map its elements to a similarity measure in the interval [0,1], thereby achieving smooth normalization of topological differences.

[0046] The original topological feature matrix is ​​obtained by extracting primary statistical features such as the total persistence, maximum persistence, persistence variance, and number of holes from the 1D persistence graph of each time subsequence. After filtering out points with infinite lifetimes in the 0D persistence graph, the same statistical features are calculated for their finite lifetime portions. The two sets of feature vectors are concatenated to obtain the final topological feature vector for each subsequence, and a matrix is ​​constructed based on this vector.

[0047] Feature space similarity matrix: Calculates the Euclidean distance between the final topological feature vectors of different time subsequences and converts it into a similarity measure, thus characterizing similarity from the macroscopic level of feature statistics.

[0048] Feature Distribution Similarity Matrix: For each time subsequence, the first moment (mean), second central moment (variance), and third standard moment (skewness) of the persistent lifetime distribution in its 0-dimensional and 1-dimensional persistence graphs are calculated to form a distribution descriptor vector. The Euclidean distance between the descriptor vectors of different time subsequences is calculated and converted into similarity, revealing deep-level pattern similarities from the overall morphological level of the lifetime distribution.

[0049] S22, Multi-view Smooth Propagation and Consensus Subspace Learning: The topological distance similarity matrix, feature space similarity matrix, and feature distribution similarity matrix are symmetrically normalized to obtain the adjacency matrix. This is then calculated using the formula... (in, Let m be the topological feature matrix of the smoothed view. It is an adjacency matrix. The number of propagations is denoted by m, where m is the view type index, and the matrix includes the topological distance similarity matrix, the feature space similarity matrix, and the feature distribution similarity matrix. The original topological feature matrix is ​​smoothed by propagating it through a smoothing process to obtain smoothed feature matrices for each view. This operation effectively suppresses noise and local anomalies, enhancing feature robustness.

[0050] Construct a joint objective function for multiple views: , in, For the joint objective function of multiple views, Let be the basis matrix of the consensus subspace to be found. For the m-th view in the subspace basis matrix The coefficient matrix below, The co-regulation term aims to make the results projected from each view's data onto the subspace as consistent as possible with the view's own coefficients. The square of the Frobenius norm. This is the regularization parameter (hyperparameter, whose value is set by human experience).

[0051] The function is solved by an alternating optimization method, resulting in a low-dimensional consensus subspace that comprehensively reflects the commonalities of multiple views and is discriminative.

[0052] S23. Spectral clustering generates state labels: Spectral embedding: The original topological feature matrix is ​​projected onto the consensus subspace to obtain the consensus embedding matrix. A second-order polynomial kernel is used to perform a nonlinear mapping on the row vectors, and the affinity matrix is ​​constructed using the square of the inner product of the mapped vectors. The symmetric normalized Laplacian matrix is ​​calculated and eigenvalue decomposition is performed. The eigenvectors corresponding to the first few smallest eigenvalues ​​are selected to form the embedding clustering matrix.

[0053] Final clustering: The K-means clustering algorithm is performed on the row vectors of the embedded clustering matrix to output the final cluster label vector. This vector divides all time subsequences into several groups, each representing a typical icing topological evolution pattern.

[0054] Step S3 involves mapping and visualizing the physical state of icing. This step transforms the data-driven clustering results into an icing state report with clear physical meaning. Step S3 specifically includes the following sub-steps: S31. Generate an icing status classification report: The cluster label vectors are mapped back to the original topological features, and the feature mean of each cluster is calculated as its "representative center". For each subsequence within a cluster, the average icing thickness is extracted from the corresponding point cloud data, and the average ambient temperature and average relative humidity for the corresponding time window are obtained from an external meteorological database to form a contextual information triple (thickness, temperature, humidity) for that cluster. Combining the icing physical mechanism with predefined rules of domain knowledge (e.g., if the 1D maximum persistence feature value of a cluster's "representative center" is high and the average temperature in its context triple is below -5°C, it is determined to be in the "rapid icing growth period"), each cluster is matched and assigned a physical pattern label, ultimately generating a structured icing state classification report.

[0055] S32. Construct a visual monitoring interface: Use the React framework to build interactive visual interfaces.

[0056] The visual interface includes the following modules: Continuity Graph Module: Dynamically displays the 0-dimensional and 1-dimensional continuity graphs of the currently selected sequence, where points far from the diagonal represent significant and stable topological features; Topological Evolution Heatmap Module: The dynamic topological distance matrix is ​​presented in the form of a heatmap. The horizontal and vertical axes are time / subsequence indices, and the color intensity intuitively reflects the topological differences in icing morphology at different times. Status Report Module: Clearly displays icing status classification reports and their corresponding physical labels.

[0057] The visualization interface provides maintenance personnel with an intuitive and comprehensive view of icing status monitoring and decision support.

[0058] Example 2: like Figure 2 As shown, this embodiment provides a transmission line icing state analysis system based on point cloud modeling and deep learning, used to implement the transmission line icing state analysis method as described in Embodiment 1, including: The data acquisition and topology feature extraction module is used to divide the time series point cloud of the target transmission line into multiple time subsequences; to voxelize the point cloud of each time subsequence to generate a voxel point set and perform persistent cohomology analysis to generate a persistent graph characterizing its topology; and to calculate the dynamic topological distance between different time subsequences based on the persistent graph. The multi-view fusion clustering module is used to extract and construct the topological feature vector of each subsequence based on the dynamic topological distance and the persistent graph, forming the original topological feature matrix; construct similarity matrices under the three views of topological distance, feature space and feature distribution respectively; use the similarity matrices under the three views as multi-view input, obtain the consensus subspace through multi-view joint learning, and cluster all time subsequences in the consensus subspace to obtain cluster label vectors; The state mapping and visualization module is used to assign physical pattern labels to each cluster according to the cluster label vector and its corresponding physical context information, based on predefined rules, to generate an icing state classification report and perform visualization.

[0059] Example 3: like Figure 3 As shown, this embodiment provides an electronic device, which may include: at least one processor, at least one network interface, a user interface, a memory, and at least one communication bus.

[0060] The communication bus can be used to enable communication between the various components mentioned above.

[0061] The user interface may include buttons, and optional user interfaces may also include standard wired interfaces and wireless interfaces.

[0062] The network interface may include, but is not limited to, Bluetooth modules, NFC modules, Wi-Fi modules, etc.

[0063] The processor may include one or more processing cores. It connects various parts of the electronic device via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in memory, and accessing data stored in memory to perform various functions and process data. Optionally, the processor can be implemented using at least one hardware form of DSP, FPGA, or PLA. The processor may integrate one or more of the following: CPU, GPU, and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor.

[0064] The memory may include RAM or ROM. Optionally, the memory may include a non-transitory computer-readable medium. The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor. The memory, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and analysis applications. The processor can be used to call the analysis applications stored in the memory and execute the steps of the transmission line icing state analysis method mentioned in the foregoing embodiments.

[0065] Example 4: This embodiment provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the above-described instructions. Figure 1 One or more steps in the illustrated embodiment. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0066] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0067] Those skilled in the art will understand that all or part of the processes in the method of Embodiment 1 described above can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and the implementation scheme can be combined arbitrarily.

[0068] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0069] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0070] The above description is merely an exemplary embodiment of the present invention and should not be construed as limiting the scope of the invention. Any equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of embodiments of the invention upon considering the specification and practicing the disclosure herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of the invention are defined by the claims.

Claims

1. A method for analyzing the icing status of transmission lines based on point cloud modeling and deep learning, characterized in that, Including the following steps: S1. Divide the time series point cloud of the target transmission line into multiple time subsequences; perform voxelization on the point cloud of each time subsequence to generate a voxel point set and perform persistent cohomology analysis to generate a persistent graph characterizing its topology; calculate the dynamic topological distance between different time subsequences based on the persistent graph. S2. Based on the dynamic topological distance and persistence graph, extract and construct the topological feature vector of each subsequence to form the original topological feature matrix; Similarity matrices are constructed under three views: topological distance, feature space, and feature distribution. The similarity matrices under the three views are used as multi-view inputs. A consensus subspace is obtained through multi-view joint learning. All time subsequences are clustered in the consensus subspace to obtain cluster label vectors. S3. Based on the clustering label vector and its corresponding physical context information, assign a physical pattern label to each cluster according to predefined rules, generate an icing status classification report and display it visually.

2. The method for analyzing the icing state of transmission lines based on point cloud modeling and deep learning according to claim 1, characterized in that, Step S1 generates a persistent graph representing its topological structure, specifically as follows: The point clouds of each time subsequence are fused and downsampled to generate a voxel point set; Construct a multi-scale Vietoris-Rips complex filtering sequence based on the voxel point set; Persistent cohomology computation is performed on the Vietoris-Rips complex filtering sequence to generate 0-dimensional persistent graphs and 1-dimensional persistent graphs.

3. The method for analyzing the icing state of transmission lines based on point cloud modeling and deep learning according to claim 1, characterized in that, In step S1, the dynamic topological distance between different time subsequences is calculated by calculating the multi-feature weighted temporal Wasserstein distance, and the calculation formula is as follows: ， In the formula, i and j Indices for the number of two distinct subsequences. and There are two subsequences. For indexes of time points, The time length of the subsequence. For subsequence and Multi-feature weighted temporal Wasserstein distance For time series weights based on Gaussian distribution, For Wasserstein distance, and Representing subsequences respectively and In the A 0-dimensional continuous plot at each time point and Representing subsequences respectively and In the A 1D continuous plot at each time point.

4. The method for analyzing the icing state of transmission lines based on point cloud modeling and deep learning according to claim 1, characterized in that, Step S2 extracts and constructs the topological feature vector of each subsequence to form the original topological feature matrix, specifically as follows: Primary topological feature statistics are extracted from the 0-dimensional and 1-dimensional persistence plots of each time subsequence and concatenated to form the topological feature vector of that subsequence. The original topological feature matrix is ​​constructed by treating the topological feature vectors of all subsequences as rows.

5. The method for analyzing the icing state of transmission lines based on point cloud modeling and deep learning according to claim 4, characterized in that, In step S2, the consensus subspace is obtained through multi-view joint learning, specifically as follows: The similarity matrices under the three views are used as adjacency matrices to smoothly propagate the original topological feature matrix; A multi-view joint objective function is constructed and optimized, which includes a cooperative regularization term to constrain the consistency of projection of different views into the consensus subspace, so as to obtain the consensus subspace basis matrix.

6. The method for analyzing the icing state of transmission lines based on point cloud modeling and deep learning according to claim 5, characterized in that, In step S2, clustering is performed on all time subsequences in this space, specifically as follows: Project the original topological features onto the consensus subspace; Spectral embedding is performed on the projected features to obtain a low-dimensional embedding clustering matrix; K-means clustering is performed on the embedded clustering matrix to obtain the cluster label vector.

7. The method for analyzing the icing state of transmission lines based on point cloud modeling and deep learning according to claim 1, characterized in that: The physical context information includes average icing thickness, average ambient temperature, and average relative humidity. The predefined rules are matching rules set based on the icing physical mechanism, which combine topological features and physical context information with a threshold for comparison.

8. A transmission line icing status analysis system based on point cloud modeling and deep learning, characterized in that, The method for analyzing the icing status of transmission lines as described in any one of claims 1 to 7 includes: The data acquisition and topology feature extraction module is used to divide the time series point cloud of the target transmission line into multiple time subsequences; to voxelize the point cloud of each time subsequence to generate a voxel point set and perform persistent cohomology analysis to generate a persistent graph characterizing its topology; and to calculate the dynamic topological distance between different time subsequences based on the persistent graph. The multi-view fusion clustering module is used to extract and construct the topological feature vector of each subsequence based on the dynamic topological distance and the persistent graph, forming the original topological feature matrix; construct similarity matrices under the three views of topological distance, feature space and feature distribution respectively; use the similarity matrices under the three views as multi-view input, obtain the consensus subspace through multi-view joint learning, and cluster all time subsequences in the consensus subspace to obtain cluster label vectors; The state mapping and visualization module is used to assign physical pattern labels to each cluster according to the cluster label vector and its corresponding physical context information, based on predefined rules, to generate an icing state classification report and perform visualization.

9. A computer device, the computer device comprising a memory, a processor, and a computer program, characterized in that, When the computer program is executed by the processor, it implements the transmission line icing status analysis method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the transmission line icing status analysis method as described in any one of claims 1 to 7.

Images