Power transmission line safety risk assessment method and system based on deformation information

Abstract

The application provides a power transmission line safety risk assessment method and system based on deformation information, and relates to the technical field of power system state monitoring and safety assessment. The method comprises the following steps: collecting monitoring data of each monitoring point of the power transmission line; analyzing and encoding the original deformation time series data set to obtain a compressed deformation feature vector; constructing a power transmission line topology graph structure in combination with the compressed deformation feature vector; processing the power transmission line topology graph structure based on a symplectic geometry structure-preserving deformation dynamics numerical integration algorithm to obtain conductor deformation evolution trajectory data; performing modal identification on the conductor deformation evolution trajectory data to determine a time-varying modal feature sequence, a dominant modal frequency drift amount and a modal energy transfer rate; and calling a safety risk assessment model to process the topology graph deformation data representation, the time-varying modal feature sequence, the dominant modal frequency drift amount and the modal energy transfer rate to determine a safety risk assessment result of the power transmission line.

Description

Technical Field

[0001] This invention relates to the field of power system condition monitoring and safety assessment technology, and in particular to a method and system for assessing the safety risks of transmission lines based on deformation information. Background Technology

[0002] Transmission lines, as critical infrastructure of the power system, are constantly exposed to a complex and ever-changing natural environment, enduring various external excitations such as wind loads, ice loads, temperature variations, and electromagnetic forces. Traditional methods for assessing the safety of transmission lines mainly rely on manual periodic inspections and threshold judgments based on single-point sensor monitoring data. This approach has several limitations. First, manual inspections are limited by inspection cycles and personnel experience, making it difficult to achieve continuous real-time monitoring of the structural condition and easily missing early signs of damage. Second, threshold judgment methods based on single-point monitoring data ignore the spatial topological characteristics of transmission lines and cannot capture the propagation patterns of deformation along conductors and towers, resulting in a lack of effective inference capabilities for risk states where no sensors are installed. Third, when performing long-term deformation evolution calculations using traditional numerical simulation methods, the inherent numerical dissipation and spurious damping effects of explicit integration methods cause the fidelity of simulation results to degrade rapidly over time, making it impossible to reliably extrapolate long-term damage evolution trends from short-term monitoring data. Fourth, conventional frequency domain modal analysis methods are based on the assumptions of system linearity and time invariance. For the non-stationary deformation process of transmission lines under continuously changing external excitation conditions, they cannot accurately track the time evolution trajectory of the system's natural frequency and damping ratio, thus making it difficult to identify the structural stiffness degradation caused by fatigue damage in a timely manner. Summary of the Invention

[0003] To address the aforementioned technical problems, embodiments of the present invention provide a method for assessing the safety risks of transmission lines based on deformation information, comprising: Monitoring data from various monitoring points along the transmission line are collected. The monitoring data includes conductor displacement data, strain data, and temperature data. The monitoring data forms the original deformation time series dataset of the transmission line. The original deformation time series dataset is subjected to principal component analysis and regularized sparse coding to obtain compressed deformation feature vectors; By combining the compressed deformation feature vector, a transmission line topology graph structure corresponding to the actual topology of the transmission line is constructed. The tower nodes in the transmission line are mapped to graph nodes in the transmission line topology graph structure, and the conductor spans in the transmission line are mapped to graph edges in the transmission line topology graph structure. The topology of the transmission line is processed using a numerical integral algorithm based on symplectic geometric preserved structure deformation dynamics to obtain conductor deformation evolution trajectory data. The adaptive dynamic mode decomposition algorithm is used to identify the mode of the conductor deformation evolution trajectory data, and the time-varying mode feature sequence, dominant mode frequency drift and mode energy transfer rate are determined based on the identification results. The pre-built safety risk assessment model is invoked to process the topological graph variable data representation, time-varying modal feature sequence, dominant mode frequency drift, and mode energy transfer rate to determine the safety risk assessment result of the transmission line.

[0004] In one embodiment, obtaining a compressed deformation feature vector from the original deformation time series dataset through principal component analysis and regularized sparse coding includes: Principal component analysis was used to extract principal component components whose cumulative contribution rate met a preset threshold from the original deformation time series dataset through dimensionality reduction processing. The principal component components are subjected to regularized sparse coding to obtain the compressed deformable feature vector.

[0005] In one embodiment, the step of extracting principal component components whose cumulative contribution rate meets a preset threshold through principal component analysis and dimensionality reduction of the original deformation time series dataset includes: The original deformation time series dataset is standardized, and the covariance matrix is ​​calculated based on the standardized original deformation time series dataset. The covariance matrix is ​​decomposed into eigenvalues ​​and then sorted in descending order according to the decomposed eigenvalues. Based on the arrangement results, eigenvalues ​​with a cumulative contribution rate of 95% are selected to construct the projection matrix; The principal component components are obtained by multiplying the original deformation time series dataset with the projection matrix.

[0006] In one embodiment, the step of performing regularized sparse coding on the principal component components to obtain the compressed deformable feature vector includes: The principal component components are subjected to feature sparsification processing using the L1 regularized sparse coding algorithm to obtain the compressed deformable feature vector.

[0007] In one embodiment, the numerical integration algorithm based on symplectic geometric preserves structural deformation dynamics is used to process the transmission line topology to obtain conductor deformation evolution trajectory data, including: Based on the topology of the transmission line, a finite element discrete model of the conductor is established to divide the conductor into multiple units, and the node position and velocity of each unit are respectively used as generalized coordinates and generalized momentum. The system of equations of motion for the conductor is constructed based on Hamiltonian canonical equations, and the system of equations of motion has symplectic structure properties. The fourth-order symplectic Runge-Kutta method is used to discretize the equations of motion in time, and the generalized coordinates and generalized momentum in each time step are calculated. The Hamiltonian function value is calculated based on the generalized coordinates and generalized momentum of each time step, and the energy conservation error is judged based on the Hamiltonian function value to determine whether it meets the preset requirements. If the requirements are met, the calculation of the next time step continues until the calculation of the entire time domain is completed, and the conductor deformation evolution trajectory data is obtained.

[0008] In one embodiment, the method further includes: If the energy conservation error does not meet the preset requirements based on the Hamiltonian function value, the generalized coordinates and generalized momentum of the current time step are corrected based on the constrained projection method of the Lagrange multipliers. Determine the local truncation error estimates for the generalized coordinates and generalized momentum of the current time step; Adjust the step size of the next time step based on the estimated local truncation error.

[0009] In one embodiment, the modal identification of the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm includes: The deformation evolution trajectory data of the conductor is captured by a sliding window, and the multi-channel deformation data in the sliding window is arranged in chronological order into a state matrix. The column vectors of the state matrix represent the deformation state of the conductor at different times. The state matrix is ​​subjected to singular value decomposition, and the dominant mode and secondary mode are determined based on the proportion of singular value energy after decomposition. Based on the dominant mode, the time-shifted pairs of the state matrix are constructed, and the linear mapping matrix between the time-shifted pairs is solved by the least squares method as the dynamic operator matrix. The eigenvalues ​​and eigenvectors of the dynamic operator matrix are determined, and the instantaneous natural frequency parameters and damping ratio parameters are determined based on the eigenvalues. The mode shape parameters in physical space are determined based on the eigenvectors.

[0010] In one embodiment, determining the time-varying modal feature sequence, dominant mode frequency shift, and modal energy transfer rate based on the identification results includes: The modal frequency difference between the two windows is calculated based on the dynamic operator matrix of the current window and the previous window, and the modal frequency difference is defined as the dominant modal frequency drift. The change in the energy percentage of each mode is determined based on the state matrix of the current window and the previous window, and the change is defined as the mode energy transfer rate. The instantaneous modal parameter evolution curve is determined based on all window data, and the instantaneous modal parameter evolution curve is defined as the time-varying modal feature sequence.

[0011] In one embodiment, constructing the security risk assessment model includes: The initial evaluation model is constructed, including an input layer, a graph convolutional layer, a multi-scale graph pooling layer, a graph attention layer, a temporal convolutional layer, a fully connected fusion layer, and an output layer. Training data is formed based on the topological graphical representation of transmission lines under different environments and operating conditions during historical periods, time-varying modal feature sequences, dominant mode frequency drift, mode energy transfer rate, and obtained safety risk level labels. The initial evaluation model is trained based on the training data to obtain the security risk evaluation model.

[0012] Another embodiment of the present invention also provides a transmission line safety risk assessment system based on deformation information, comprising: The acquisition module is used to collect monitoring data from various monitoring points of the transmission line. The monitoring data includes conductor displacement data, strain data, and temperature data. The monitoring data forms the original deformation time series dataset of the transmission line. The first processing module is used to obtain a compressed deformation feature vector from the original deformation time series dataset through principal component analysis and regularized sparse coding. The construction module is used to construct a transmission line topology graph structure corresponding to the actual topology of the transmission line by combining the compressed deformation feature vector. The tower nodes in the transmission line are mapped to graph nodes in the transmission line topology graph structure, and the conductor spans in the transmission line are mapped to graph edges in the transmission line topology graph structure. The second processing module is used to process the topology of the transmission line based on the numerical integration algorithm of the deformation dynamics of the symplectic geometric preserved structure to obtain conductor deformation evolution trajectory data. The first determining module is used to perform mode identification on the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm, and determine the time-varying mode feature sequence, dominant mode frequency drift and mode energy transfer rate based on the identification results; The third processing module is used to call the pre-built safety risk assessment model to process the topological graph variable data representation, time-varying modal feature sequence, dominant mode frequency drift, and mode energy transfer rate to determine the safety risk assessment result of the transmission line.

[0013] Other features and advantages of this application will be set forth in the following description. The objectives and other advantages of this application can be realized and obtained through the structures particularly pointed out in the written description and drawings.

[0014] The technical solution of this application will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the power transmission line safety risk assessment method based on deformation information in an embodiment of the present invention.

[0017] Figure 2 This is a flowchart illustrating a transmission line safety risk assessment method based on deformation information according to another embodiment of the present invention.

[0018] Figure 3 This is a flowchart illustrating a transmission line safety risk assessment method based on deformation information in another embodiment of the present invention.

[0019] Figure 4 This is a structural block diagram of a power transmission line safety risk assessment system based on deformation information according to an embodiment of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but these are not intended to limit the scope of the invention.

[0021] It should be understood that various modifications can be made to the embodiments disclosed herein. Therefore, the following description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope of this disclosure will be apparent to those skilled in the art.

[0022] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the present disclosure and, together with the general description of the disclosure given above and the detailed description of the embodiments given below, serve to explain the principles of the disclosure.

[0023] These and other features of the invention will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.

[0024] It should also be understood that although the invention has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of the invention.

[0025] The above and other aspects, features and advantages of this disclosure will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.

[0026] Specific embodiments of the present disclosure are described thereafter with reference to the accompanying drawings; however, it should be understood that the disclosed embodiments are merely examples of the present disclosure and can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that could obscure the present disclosure. Therefore, the specific structural and functional details disclosed herein are not intended to be limiting, but are merely representative of the present disclosure and are intended to teach those skilled in the art to use the present disclosure in a variety of substantially any suitable detailed structures.

[0027] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in still another embodiment,” all of which may refer to one or more of the same or different embodiments according to this disclosure.

[0028] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] like Figure 1 As shown, this embodiment of the invention provides a method for assessing the safety risks of transmission lines based on deformation information, including: S1: Collect monitoring data from various monitoring points along the transmission line. The monitoring data includes conductor displacement data, strain data, and temperature data. The monitoring data forms the original deformation time series dataset of the transmission line. S2: The original deformation time series dataset is processed by principal component analysis and regularized sparse coding to obtain a compressed deformation feature vector; S3: Construct a transmission line topology graph structure corresponding to the actual topology of the transmission line by combining the compressed deformation feature vector. The tower nodes in the transmission line are mapped to graph nodes in the transmission line topology graph structure, and the conductor spans in the transmission line are mapped to graph edges in the transmission line topology graph structure. S4: The topology of the transmission line is processed based on the numerical integration algorithm of the deformation dynamics of the symplectic geometric preserved structure to obtain conductor deformation evolution trajectory data; S5: Based on the adaptive dynamic mode decomposition algorithm, the conductor deformation evolution trajectory data is subjected to mode identification, and the time-varying mode feature sequence, dominant mode frequency drift and mode energy transfer rate are determined based on the identification results; S6: Call the pre-built safety risk assessment model to process the topological graph variable data representation, time-varying modal feature sequence, dominant mode frequency drift, and mode energy transfer rate to determine the safety risk assessment result of the transmission line.

[0030] In this embodiment, a deformation monitoring sensor network can be deployed along the transmission line to collect displacement, strain, and temperature data at each monitoring point, forming an original deformation time series dataset. The monitoring points cover tower node locations and conductor span midpoint locations. Next, principal component analysis is performed on the original deformation time series dataset for dimensionality reduction, and regularized sparse coding is used to sparsify the features of the dimensionality reduction result, obtaining a compressed deformation feature vector. This feature vector is then used to construct a transmission line topology graph, mapping tower node locations to graph nodes and conductor spans to graph edges. The compressed deformation feature vector is used as the feature input for the graph nodes, thereby establishing a topological graph deformation data representation. Subsequently, a symplectic geometrically preserved structural deformation dynamics numerical integration algorithm is used to perform long-scale evolution simulations on the topological graph deformation data representation, generating deformation evolution trajectory data under different operating conditions. An adaptive dynamic modal decomposition algorithm is used to identify the modal characteristics of the deformation evolution trajectory data, extracting instantaneous natural frequency parameters, damping ratio parameters, and mode shape parameters to construct a time-varying modal feature sequence. The dominant mode frequency drift and modal energy transfer rate in the time-varying modal feature sequence are then calculated. After obtaining the above data, the system can call upon the topological graphical representation, time-varying modal feature sequence, dominant mode frequency drift, and modal energy transfer rate obtained from the pre-constructed safety risk assessment model to determine the current safety risk assessment result of the transmission line.

[0031] The following describes the various steps of the method of this application in detail with reference to different embodiments: In one embodiment, such as Figure 2 As shown, the process of obtaining compressed deformation feature vectors from the original deformation time series dataset through principal component analysis and regularized sparse coding includes: S201: Principal component analysis is used to extract principal component components whose cumulative contribution rate meets a preset threshold by dimensionality reduction processing of the original deformation time series dataset. S202: Perform regularized sparse coding on the principal component components to obtain the compressed deformation feature vector.

[0032] Specifically, the extraction of principal component components whose cumulative contribution rate meets a preset threshold through principal component analysis and dimensionality reduction of the original deformation time series dataset includes: S203: Standardize the original deformation time series dataset and calculate the covariance matrix based on the standardized original deformation time series dataset; S204: Perform eigenvalue decomposition on the covariance matrix and sort the decomposed eigenvalues ​​in descending order; S205: Based on the permutation results, select eigenvalues ​​with a cumulative contribution rate of 95% to construct the projection matrix; S206: Multiply the original deformation time series dataset with the projection matrix to obtain the principal component components.

[0033] The step of performing regularized sparse coding on the principal component components to obtain the compressed deformable feature vector includes: S207: The principal component components are subjected to feature sparsification processing using the L1 regularized sparse coding algorithm to obtain the compressed deformable feature vector.

[0034] The principal component analysis (PCA) dimensionality reduction steps in this embodiment specifically include standardizing the original deformation time series dataset and calculating the covariance matrix. Then, eigenvalue decomposition is performed on the covariance matrix, and the eigenvectors are arranged in descending order of eigenvalues. The top few eigenvectors that achieve a cumulative contribution rate of 95% are selected to form a projection matrix. The standardized original deformation time series dataset is then multiplied by the projection matrix to obtain the dimensionality-reduced principal component components. The L1 regularized sparse coding step specifically includes constructing a sparse coding optimization objective function, which includes a reconstruction error term and an L1 regularization penalty term. The sparse coding optimization objective function is iteratively solved using a coordinate descent algorithm to obtain a sparse coefficient matrix. Elements in the sparse coefficient matrix with absolute values ​​less than a threshold are set to zero, and elements with larger absolute values ​​are retained to form a compressed deformation feature vector. The dimension of the compressed deformation feature vector is 10% to 20% of the feature dimension of the original deformation time series dataset.

[0035] In one embodiment, the numerical integration algorithm based on symplectic geometric preserves structural deformation dynamics is used to process the transmission line topology to obtain conductor deformation evolution trajectory data, including: S401: Based on the topology of the transmission line, establish a finite element discrete model of the conductor to divide the conductor into multiple units, and use the node position and velocity of each unit as generalized coordinates and generalized momentum, respectively. S402: Construct a set of motion equations for the conductor based on Hamiltonian canonical equations, wherein the set of motion equations has symplectic structure characteristics; S403: The fourth-order symplectic Runge-Kutta method is used to discretize the equations of motion in time, and the generalized coordinates and generalized momentum in each time step are calculated. S404: Calculate the Hamiltonian function value based on the generalized coordinates and generalized momentum of each time step, and determine whether the energy conservation error meets the preset requirements based on the Hamiltonian function value; S405: If the requirements are met, continue to execute the calculation of the next time step until the calculation of the entire time domain is completed, and obtain the conductor deformation evolution trajectory data.

[0036] Furthermore, such as Figure 3As shown, the method further includes: S406: If the energy conservation error does not meet the preset requirements based on the Hamiltonian function value, then the generalized coordinates and generalized momentum of the current time step are corrected based on the constrained projection method of the Lagrange multipliers. S407: Determine the local truncation error estimates for the generalized coordinates and generalized momentum of the current time step; S408: Adjust the step size of the next time step based on the estimated local truncation error.

[0037] For example, the principle of the symplectic geometry-preserving deformation dynamics numerical integration algorithm is to represent the conductor deformation system as a Hamiltonian mechanics system, that is, to construct a conductor deformation model based on the Hamiltonian mechanics system, define generalized coordinates as the position vectors of each point on the conductor, define generalized momentum as the corresponding momentum vector, and then construct the Hamiltonian function of the system based on the generalized coordinates and generalized momentum. The Hamiltonian function includes kinetic energy terms and potential energy terms, and the potential energy term includes gravitational potential energy components, elastic strain energy components, and work done by external load components. A set of equations of motion for the deformation system is established based on the Hamiltonian canonical equations, and the set of equations of motion has symplectic structural characteristics. The symplectic Runge-Kutta integration scheme is used to discretize the set of equations of motion, and the symplectic Runge-Kutta integration scheme strictly preserves the symplectic geometry of the Hamiltonian system at the discretization level. In addition, this embodiment also introduces an adaptive time step control strategy in the numerical integration process. The time step is dynamically adjusted according to the local truncation error estimate. It can also perform multi-symplectic correction on the symplectic structure deviation accumulated during long-term integration. The numerical solution is projected back to the symplectic manifold through the projection algorithm to eliminate numerical dissipation effect and spurious damping effect.

[0038] When implementing the above steps using the numerical integration algorithm for structural deformation dynamics based on symplectic geometry, the specific steps include: establishing a finite element discrete model of the conductor, dividing the conductor into several elements, with the node position and velocity of each element serving as generalized coordinates and generalized momentum, respectively; then, constructing the Hamiltonian function expression of the system based on the stress-strain constitutive relation of the conductor material and the catenary geometric equations; subsequently, using the fourth-order symplectic Runge-Kutta method to discretize the Hamiltonian canonical equations, i.e., the equations of motion, over time, updating the generalized coordinates and generalized momentum through multi-level calculations within each time step; calculating the Hamiltonian function value after each time step to determine whether the energy conservation error exceeds a set threshold; when the energy conservation error exceeds the set threshold, correcting the generalized coordinates and generalized momentum using a constrained projection method based on Lagrange multipliers to bring the system back to a symplectic manifold. In addition, at the end of each time step, the step size of the next time step is adjusted by using a PI controller algorithm based on the estimated local truncation error of the current time step. The above steps are repeated until the numerical integration of the entire time domain is completed, and the position evolution trajectory, velocity evolution trajectory and stress evolution trajectory of each point of the conductor are output. The position evolution trajectory, the velocity evolution trajectory and the stress evolution trajectory constitute deformation evolution trajectory data.

[0039] By employing a symplectic geometry-preserving deformation dynamics numerical integration algorithm, the algorithm rigorously maintains the symplectic geometry of the Hamiltonian system at the numerical level, ensuring energy conservation and phase space volume invariance during deformation evolution. This avoids energy decay and phase trajectory contraction caused by numerical dissipation in traditional explicit integration methods, preventing spurious damping effects and stability deviations in long-term simulations. Furthermore, an adaptive time step control strategy dynamically adjusts the computational accuracy based on local truncation error estimates. The step size is automatically reduced during periods of rapid deformation change to capture details, and increased during stable periods to improve efficiency, achieving optimal allocation of computational resources. The multi-symplectic correction method compensates for geometric deviations caused by finite-precision calculations by periodically projecting the numerical solution back to the symplectic manifold. This ensures high fidelity in fatigue damage accumulation analysis spanning several years or even decades, providing a reliable numerical tool for extrapolating long-term damage evolution trends from short-term monitoring data and solving the problem of rapid accuracy degradation in long-term integration using traditional methods.

[0040] In one embodiment, the modal identification of the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm includes: S501: Use a sliding window to capture the deformation evolution trajectory data of the conductor, and arrange the multi-channel deformation data in the sliding window in chronological order into a state matrix. The column vectors of the state matrix represent the deformation state of the conductor at different times. S502: Perform singular value decomposition on the state matrix, and determine the dominant mode and secondary mode based on the energy ratio of the singular values ​​after decomposition; S503: Construct time-shifted pairs of the state matrix based on the dominant mode, and use the least squares method to solve for the linear mapping matrix between the time-shifted pairs as the dynamic operator matrix; S504: Determine the eigenvalues ​​and eigenvectors of the dynamic operator matrix, and determine the instantaneous natural frequency parameters and damping ratio parameters based on the eigenvalues, and determine the mode shape parameters in physical space based on the eigenvectors.

[0041] Furthermore, based on the identification results, the time-varying modal feature sequence, dominant mode frequency shift, and modal energy transfer rate are determined, including: S505: Calculate the modal frequency difference between the two windows based on the dynamic operator matrix of the current window and the previous window, and define the modal frequency difference as the dominant modal frequency drift. S506: Determine the change in the energy percentage of each mode based on the state matrix of the current window and the previous window, and define the change as the mode energy transfer rate; S507: Determine the instantaneous modal parameter evolution curve based on all window data, and define the instantaneous modal parameter evolution curve as the time-varying modal feature sequence.

[0042] In this embodiment, the principle of the adaptive dynamic mode decomposition algorithm is to treat the transmission line deformation system as a dynamic system and reconstruct the state-space representation of the system through time series data; to use a sliding window strategy to extract deformation data within a continuous time period and arrange the data in each window into a state matrix; to perform singular value decomposition on the state matrix and extract the singular vectors corresponding to the dominant singular values ​​as spatial modes, i.e., dominant modes; to construct a linear approximation of the dynamic operator through the time translation relationship of the state matrix, and to solve for the eigenvalues ​​and eigenvectors of the dynamic operator, where the modulus of the eigenvalue corresponds to the mode amplitude attenuation rate and the argument of the eigenvalue corresponds to the mode oscillation frequency; to use a recursive least squares algorithm to update the dynamic operator matrix in each new time window to achieve online tracking of mode parameters; and to identify the time-varying characteristics and non-stationary behavior of the system by monitoring the changes in mode frequency and mode energy distribution in adjacent time windows.

[0043] The implementation steps of the adaptive dynamic mode decomposition algorithm, as described above, specifically include: setting the sliding window length and window movement step size; extracting data from the first time window from the deformation evolution trajectory data; arranging the multi-channel deformation data within the time window into a state matrix in chronological order, where the column vectors of the state matrix represent the system states at different times; performing singular value decomposition on the state matrix and selecting the top few dominant modes based on the singular value energy ratio; constructing time-shifted pairs of the state matrix and using the least squares method to solve for the linear mapping matrix between them as an approximation of the dynamic operator; calculating the eigenvalues ​​and eigenvectors of the dynamic operator matrix and converting the eigenvalues ​​into continuous vectors. The modal frequencies and damping ratios of the time system are used to convert the eigenvectors into mode shapes in physical space. The modal frequencies, damping ratios, and mode shapes constitute instantaneous natural frequency parameters, damping ratio parameters, and mode shape parameters. The sliding window moves forward by one step to acquire new observation data, and the dynamic operator matrix is ​​updated using a recursive least squares algorithm. The difference in modal frequencies between the current window and the previous window is calculated as the dominant modal frequency drift, and the change in the energy proportion of each mode is calculated as the modal energy transfer rate. The above steps are repeated until the entire deformation evolution trajectory data is processed, and the instantaneous modal parameter evolution curves at each time point are output. The instantaneous modal parameter evolution curves constitute a time-varying modal feature sequence.

[0044] The adaptive dynamic modal decomposition algorithm achieves instantaneous modal tracking of time-varying systems through a sliding window and recursive update mechanism. It does not require assumptions about the system's linearity or time invariance and is suitable for non-stationary deformation analysis of transmission lines under continuously changing external excitations such as wind speed variations, snow accumulation, and temperature cycles. The algorithm can capture the temporal evolution trajectory of the system's natural frequencies and damping ratios. When conductor stiffness degrades due to fatigue damage, the natural frequencies of the dominant modes drift towards lower frequencies; when dampers fail, the damping ratio decreases significantly. These abnormal changes in modal parameters provide sensitive indicators for early warning of structural health. By monitoring the transfer of modal energy between different orders of modes, the system can identify abrupt changes in the conductor's dynamic characteristics. For example, a sudden increase in the energy of a higher-order mode may indicate a change in vibration modes caused by local damage to the conductor. The application of the recursive least squares algorithm allows modal parameter updates to utilize only newly arrived data without reprocessing all historical data, significantly reducing computational complexity, meeting the needs of real-time online monitoring, and providing a feasible solution for lightweight deployment of edge computing nodes.

[0045] After the above data is calculated, a pre-built safety risk assessment model can be used to assess the safety risks of the transmission line. The construction process of the safety risk assessment model in this embodiment includes: S7: Construct the initial evaluation model, including the input layer, graph convolutional layer, multi-scale graph pooling layer, graph attention layer, temporal convolutional layer, fully connected fusion layer, and output layer; S8: Training data is formed based on the topological graphical representation of transmission lines under different environments and operating conditions during historical periods, time-varying modal feature sequences, dominant mode frequency drift, mode energy transfer rate, and obtained safety risk level labels. S9: Train the initial evaluation model based on the training data to obtain the security risk evaluation model.

[0046] For example, the structure of the transmission line safety risk assessment model includes an input layer, a graph convolutional layer, a multi-scale graph pooling layer, a graph attention layer, a temporal convolutional layer, a fully connected fusion layer, and an output layer. The input layer receives topological graph-variable data representations and time-varying modal feature sequences; the graph convolutional layer extracts features from the topological graph structure, aggregating the neighborhood information of each graph node through a message passing mechanism, which calculates the information propagation weights between graph nodes based on the connection relationships of graph edges; the multi-scale graph pooling layer performs hierarchical downsampling of the topological graph structure, extracting topological features at different granularities. The multi-scale graph pooling layer includes coarse-grained pooling layers and fine-grained pooling layers. The coarse-grained pooling layer aggregates adjacent tower node positions into regional nodes, and the fine-grained pooling layer preserves the local features of individual tower node positions; the graph attention layer... The attention weight coefficients between graph node pairs are dynamically adjusted based on the dominant mode frequency drift and mode energy transfer rate. When the dominant mode frequency drift exceeds 3% of the initial frequency or the mode energy transfer rate exceeds 20%, the attention weight coefficients of the corresponding graph node pairs increase. The temporal convolutional layer is used to perform temporal modeling on the time-varying modal feature sequence and extract the temporal dependency of deformation evolution. The fully connected fusion layer is used to concatenate and fuse graph features and temporal features. The output layer is used to output the security risk assessment results using a multi-classification structure. The security risk assessment results include four categories: normal, low risk, medium risk, and high risk.

[0047] The steps for establishing the training dataset for the safety risk assessment model of the transmission line include: collecting transmission line deformation monitoring data under different meteorological conditions, load levels, and line operating years; using finite element simulation software to simulate deformation response data under extreme conditions to supplement dangerous scenarios not covered by measured data; processing each set of deformation monitoring data and deformation response data through steps S2 to S5 to generate corresponding topological deformation data representations, time-varying modal characteristic sequences, dominant mode frequency drift, and modal energy transfer rates. Then, determining whether the conductor stress exceeds the material yield strength, whether the tower tilt angle exceeds the safety limit, and whether the deformation rate exceeds the warning threshold, and labeling the data with safety risk levels, and labeling the above data with safety risk levels based on the determined results. The labeled topological deformation data representations, time-varying modal characteristic sequences, dominant mode frequency drift, modal energy transfer rates, and safety risk level labels are divided into training and validation sets in an 8:2 ratio (or other ratio). Next, data augmentation can be performed on the training set by adding Gaussian noise to the deformation data to simulate sensor measurement errors, and by randomly pruning and splicing time series data to increase sample diversity. The processed training set and validation set form the training dataset for the transmission line safety risk assessment model, which is used to train the model.

[0048] The training process of the transmission line safety risk assessment model includes: initializing the weight parameters of the graph convolutional layer, graph attention layer, temporal convolutional layer, and fully connected fusion layer; inputting the training set data from the transmission line safety risk assessment model training dataset into the model in batches, and calculating the probability distribution of the predicted safety risk level through forward propagation; constructing a loss function, which includes a cross-entropy classification loss term and a physical constraint regularization term. The physical constraint regularization term includes the residual of the conductor catenary equation and the stress-strain constitutive relation deviation. The residual of the conductor catenary equation is calculated by comparing the difference between the conductor shape predicted by the model and the theoretical catenary curve. The deviation of the stress-strain constitutive relationship is calculated by comparing the difference between the stress-strain relationship predicted by the model and the material constitutive model; the gradient of the loss function with respect to the weight parameters of each layer is calculated using the backpropagation algorithm; the weight parameters are updated using the Adam optimizer, with the learning rate set to 0.001 and the momentum parameter set to 0.9; the model performance is evaluated on the validation set every 10 training batches, and the classification accuracy and loss function value are recorded; when the loss function value on the validation set does not decrease for 20 consecutive evaluation cycles, the early stopping mechanism is triggered to terminate the training; the weight parameters with the best performance on the validation set are saved as the completed transmission line safety risk assessment model.

[0049] In this embodiment, a graph neural network and a deformation propagation algorithm for transmission line topology perception are employed to explicitly encode the physical topological relationships of the transmission line into a topological graph structure. This enables the transmission line safety risk assessment model to perform global inference of the deformation field using the connectivity and mechanical transmission paths of the line. The message passing mechanism of the graph convolutional layer simulates the propagation process of deformation along conductors and towers. When abnormal deformation occurs at a monitoring point along the transmission line, the transmission line safety risk assessment model can track the diffusion trend of deformation to adjacent graph nodes through graph edge relationships, thereby predicting the deformation state at locations where no sensors are installed. Multi-scale graph pooling layers extract topological features at different levels of abstraction. Coarse-grained features reflect regional deformation distribution patterns, while fine-grained features capture local anomalies of individual towers. The combination of these two types of features allows the transmission line safety risk assessment model to identify both large-scale coordinated deformation and isolated local damage. The attention weight coefficients of the graph attention layer dynamically adjust the information interaction weights between graph nodes based on the degree of anomaly in the modal parameters. Nodes with significant frequency drift or energy transfer are given higher attention, enabling the transmission line safety risk assessment model to focus on potential risk areas and improving its sensitivity to early damage signs. The combination of the temporal convolutional layer and the topological graph structure achieves spatiotemporal coupled modeling, considering both the propagation of deformation in the spatial topology and the evolutionary inertia in the time series, providing a foundation for predicting future deformation trends and the dynamic development of risks. The physical constraint regularization term embeds the conductor catenary equation and the material stress-strain constitutive relation into the loss function, guiding the transmission line safety risk assessment model to learn characteristic representations that conform to mechanical laws. This enhances the model's generalization ability in sparse training data regions and its predictive reliability under extreme conditions.

[0050] When the implementation of this embodiment is in the cloud, an edge-cloud collaborative computing architecture needs to be constructed. For example, the implementation can be as follows: a lightweight deformation feature extraction module and a preliminary risk judgment module are deployed on edge computing nodes along the transmission line; the deformation feature extraction module performs filtering, noise reduction, and feature dimensionality reduction processing on the real-time deformation monitoring data collected by the sensors to extract compressed deformation feature vectors; the preliminary risk judgment module uses threshold discrimination rules to trigger anomaly marking when the deformation amplitude, rate of change, or cumulative amount exceeds a preset threshold; the edge computing nodes only upload the real-time deformation monitoring data marked as abnormal and the corresponding compressed deformation feature vectors to the cloud server through a wireless communication network; the cloud server runs the trained transmission line safety risk assessment model, performs in-depth analysis on the uploaded abnormal real-time deformation monitoring data, and outputs safety risk assessment results and location information; the cloud server sends the safety risk assessment results to the edge computing nodes and maintenance personnel terminals to trigger corresponding early warning and response measures; for real-time deformation monitoring data that the edge computing nodes determine to be normal, the cloud server periodically samples and sends it back for verification.

[0051] The dominant mode frequency drift is defined as the ratio of the difference between the first-order mode frequency extracted in the current time window and the first-order mode frequency in the initial healthy state to the first-order mode frequency in the initial healthy state, where the initial healthy state refers to the state of the transmission line before any damage occurs during its initial commissioning. The modal energy transfer rate is the sum of the absolute values ​​of the differences between the energy proportions of each mode in the current time window and the corresponding energy proportions in the initial healthy state.

[0052] The physical constraint regularization term is constructed as follows: the residual of the conductor catenary equation is constructed by comparing the difference between the vertical displacement of each point of the conductor predicted by the transmission line safety risk assessment model and the calculated value of the catenary equation; the deviation of the stress-strain constitutive relation is constructed by comparing the difference between the stress calculated by substituting the strain value predicted by the transmission line safety risk assessment model into the stress-strain constitutive relation and the measured stress; the weight coefficient of the physical constraint regularization term in the loss function is dynamically adjusted according to the relative magnitude of the data fitting error and the physical consistency error during the training process. When the data fitting error is large, the weight of the physical constraint is reduced to accelerate convergence; when the physical consistency error is large, the weight of the physical constraint is increased to ensure that the transmission line safety risk assessment model conforms to the laws of mechanics.

[0053] The calculation steps for the remaining life prediction include: based on real-time deformation monitoring data, using the rainflow counting method to statistically analyze the amplitude and number distribution of stress cycles; based on the fatigue SN curve of the conductor material, calculating the allowable number of cycles corresponding to each stress cycle, where the fatigue SN curve describes the relationship between stress amplitude and fatigue life cycle number; according to the Miner linear cumulative damage criterion, summing the ratio of the actual number of stress cycles to the allowable number of cycles to obtain the cumulative damage degree; using the Monte Carlo method to randomly generate external load scenarios for future time periods, where the external load scenarios include random fluctuations in wind speed parameters, temperature parameters, and icing parameters; running the symplectic geometric preserver structural deformation dynamics numerical integration algorithm for each external load scenario to calculate the corresponding stress cycle distribution and accumulate the cumulative damage degree; repeating the above random simulation process several times, statistically analyzing the time distribution when the cumulative damage degree reaches a critical value, where the critical value is set to 1.0, and outputting the expected value and confidence interval of the remaining life prediction.

[0054] The topological graph data representation is as follows: the tower node positions and conductor spans of the transmission line are mapped into a topological graph structure, wherein the tower node positions are used as graph nodes, the conductor spans are used as graph edges, and the compressed deformation feature vectors are used as graph node features, forming a data structure for representing the correlation between the transmission line topology and deformation information.

[0055] The damage sensitivity index parameters are parameters used to quantify changes in the structural health status of transmission lines, including the dominant modal frequency drift and modal energy transfer rate. The dominant modal frequency drift reflects the degree of structural stiffness degradation, and the modal energy transfer rate reflects the degree of vibration mode change.

[0056] The solution of this application will be described below with reference to specific application embodiments: First, a sensor network was deployed at key monitoring locations along the transmission line. Triaxial accelerometers and strain gauges were installed at tower nodes, and displacement and temperature sensors were installed at the midpoint of the conductor span. Data communication between the monitoring points was achieved through a wireless ad hoc network protocol. The sensor sampling frequency was set to 100Hz to capture the vibration response of the conductor. The data acquisition system packaged and stored the displacement, strain, and temperature data of each channel at fixed time intervals to form a multidimensional time series dataset. This dataset contains the monitoring point number in the spatial dimension and the sampling time in the temporal dimension, providing the original data foundation for subsequent feature extraction and modal analysis.

[0057] The original deformation time series dataset was zero-mean and variance normalized to eliminate dimensional and numerical scale differences between different sensor channels. Next, the covariance matrix of the standardized data was calculated, reflecting the correlation structure between monitoring channels. Then, eigenvalue decomposition was performed on the covariance matrix, and principal component directions were obtained by sorting the eigenvalues ​​from largest to smallest. The top principal components with a cumulative contribution rate of 95% were selected to form the projection matrix. Multiplying the original data with the projection matrix completed a dimensionality reduction transformation, resulting in principal component components with significantly reduced dimensionality but retained key information. Principal component analysis (PCA) utilizes the second-order statistical properties of the data to extract the direction of maximum variance, effectively removing redundant information and measurement noise. Based on this, a sparse coding optimization objective function was constructed. This objective function consists of a reconstruction error term and an L1 regularization penalty term. The reconstruction error term ensures that the sparse representation can accurately restore the original principal component components, while the L1 regularization penalty term imposes sparsity constraints by summing the absolute values ​​of the sparse coefficients, causing most coefficients to tend towards zero. A coordinate descent algorithm is used to iteratively optimize the sparse coefficient matrix dimension by dimension. In each iteration, the coefficients of other dimensions are fixed and only the coefficients of the current dimension are updated. L1 regularized near-end mapping is achieved through a soft threshold shrinkage operator. After convergence, elements in the sparse coefficient matrix with an absolute value less than a set threshold of 0.01 are set to zero, while key coefficients with larger absolute values ​​are retained to form a compressed deformation feature vector. The dimension of this vector is 10% to 20% of the original feature dimension. The sparse coding method reduces storage and transmission overhead by discovering the sparse representation of data, while highlighting the feature components that contribute most significantly to structural deformation, thus improving the computational efficiency and generalization performance of subsequent modeling. After feature compression, the topology of the transmission line is constructed. The tower node positions are mapped to graph nodes, and the conductor spans are mapped to graph edges. The compressed deformation feature vector is assigned to the corresponding graph nodes as node features, establishing a topological graph deformation data representation that integrates spatial topological relationships and deformation information. This representation provides a structured data format for the input of the graph neural network.

[0058] The original deformation time series dataset is standardized, and the standardization formula is as follows: ; In the formula, For the standardized first The first sample The value of each feature, The first in the original deformation time series dataset The first sample The value of each feature, For the first The mean of each feature, For the first The standard deviation of each feature The range is from 1 to the total number of samples. The range is from 1 to the total number of features. Calculate the covariance matrix. The formula is expressed as follows: ; In the formula, The covariance matrix has a dimension equal to the product of the total number of features. The total number of samples, The standardized data matrix has a dimension equal to the total number of samples multiplied by the total number of features. for The transpose of the covariance matrix. Eigenvalue decomposition of the covariance matrix yields the eigenvalues. and the corresponding feature vector ,in The total number of features, with the feature values ​​arranged in descending order, satisfies the following conditions: Calculate the cumulative contribution rate. The formula is expressed as follows: ; In the formula, For the front The cumulative contribution rate of each principal component For the first 1 eigenvalue, The number of principal components to be selected. The selection process... The smallest Values, forming the projection matrix The projection matrix is ​​from the previous It consists of 1 feature vector, with a dimension equal to the total number of features multiplied by 1. Multiplying the standardized data matrix by the projection matrix yields the dimensionality-reduced principal component components. The formula is expressed as follows: ; In the formula, This is the principal component matrix after dimensionality reduction, with the dimension being the total number of samples multiplied by . The principal component components are optimized using sparse encoding. The objective function for optimization is expressed as follows: ; In the formula, It is a sparse coefficient matrix, with dimensions equal to the number of dictionary atoms multiplied by the total number of samples. It is a dictionary matrix with dimension 1. Multiply by the number of atoms in the dictionary, where the empirical value of the number of atoms ranges from 100 to 500. This is the regularization coefficient, with an empirical value ranging from 0.01 to 0.1. Represents the square of the L2 norm. Let L1 represent the norm. An iterative solution is obtained using a coordinate descent algorithm, where the absolute values ​​of the sparse coefficient matrix are less than a threshold. Set the elements to zero, threshold The empirical values ​​range from 0.001 to 0.01. Elements with larger absolute values ​​are retained to form a compressed deformation feature vector. The dimension of the compressed deformation feature vector is 10% to 20% of the original feature dimension. The tower node positions are mapped to graph nodes, the conductor spans are mapped to graph edges, and the compressed deformation feature vector is used as the feature input for the graph nodes to establish a topological graph deformation data representation.

[0059] Furthermore, combining the Hamiltonian mechanics system, the position vectors of each discrete node of the conductor are defined as generalized coordinates, and the corresponding momentum vectors are generalized momentum. A Hamiltonian function for the system is constructed, which includes nodal kinetic energy and system potential energy terms. The potential energy term consists of gravitational potential energy components, elastic strain energy components, and work done by external loads. Based on the Hamiltonian canonical equations, a set of equations of motion for the deformed system is established. This set of equations possesses a natural symplectic structure characteristic, meaning that the time evolution of canonical variables maintains the symplectic geometric invariance of the phase space. A fourth-order symplectic Runge-Kutta method is used to discretize the set of equations of motion in time. This method strictly preserves the symplectic geometric structure of the Hamiltonian system at the discrete level through multi-level computation. An adaptive time step control strategy is introduced during numerical integration. Based on the estimated local truncation error of the current step, a proportional-integral controller algorithm is used to dynamically adjust the step size of the next step. When the deformation changes drastically, the step size is automatically reduced to improve accuracy; when the deformation is stable, the step size is increased to improve efficiency. The Hamiltonian function value is calculated at regular intervals to determine if the energy conservation error exceeds a set threshold. When the energy conservation error exceeds the threshold, a constrained projection method based on Lagrange multipliers is used to correct the generalized coordinates and generalized momentum, projecting the numerical solution back onto the symplectic manifold to eliminate accumulated geometric deviations. This numerical integration process is repeated until the simulation of the entire time domain is completed, outputting the position evolution trajectory, velocity evolution trajectory, and stress evolution trajectory of each node on the conductor. These trajectory data constitute the deformation evolution trajectory data. In this embodiment, the symplectic geometry-preserving algorithm avoids the numerical dissipation and spurious damping of traditional explicit methods by maintaining the geometric invariance of the Hamiltonian system at the numerical level. This allows the algorithm to maintain high fidelity in long-term simulations spanning several years or even decades, providing a reliable numerical tool for extrapolating long-term damage evolution trends from short-term monitoring data. During the generation of deformation evolution trajectory data, the corresponding safety risk level label is marked according to whether the conductor stress exceeds the material yield strength. When the dominant mode frequency drift exceeds 3% of the initial frequency or the mode energy transfer rate exceeds 20%, the sampling density of that time period is increased to capture the rapid changes in the system dynamic characteristics and provide sufficient data resolution for subsequent mode identification.

[0060] The Hamiltonian function is expressed as follows: ; In the formula, For Hamiltonian functions, This is a generalized coordinate vector representing the position vectors of each node on the traverse. Its dimension is the total number of degrees of freedom at each node multiplied by 1, and its unit is meters. The total number of degrees of freedom at each node is the sum of the displacement degrees of freedom of all nodes; a single node typically has 3 degrees of freedom. This is a generalized momentum vector, with dimensions equal to the total number of degrees of freedom at each node multiplied by 1, and units of kilograms per second. This is the mass matrix, with dimensions equal to the product of the total number of degrees of freedom of each node, expressed in kilograms. It is the inverse of the mass matrix. This is the component of gravitational potential energy, measured in joules. This represents the elastic strain energy component, measured in joules. The work done by the external load is expressed in joules. For reference energy, the unit is joules, and the empirical value is... Gravitational potential energy components The formula is expressed as follows: ; In the formula, This represents the total number of conductor nodes. For the first The mass of each node, in kilograms. This is the acceleration due to gravity, measured in meters per second squared (m²), with an empirical value of 9.8. For the first The vertical coordinates of each node, in meters. For reference height, the unit is meters, and the empirical value is 100. Elastic strain energy components. The formula is expressed as follows: ; In the formula, This represents the total number of conductor units. This refers to the elastic modulus, measured in Pascals. For the first The cross-sectional area of ​​each unit, in square meters (m²). For the first The elongation of each unit, in meters. For reference length, the unit is meters, and the empirical value is 100. Component of work done by external load. The formula is expressed as follows: ; In the formula, For the action in the The external load at each node is expressed in Newtons. For the first The displacement of each node, in meters. The reference displacement is expressed in meters, with an empirical value of 1. The Hamiltonian canonical equation is as follows: ; In the formula, Time, in seconds. Let be the derivative of the generalized coordinate system with respect to time, expressed in meters per second. Let be the derivative of generalized momentum with respect to time, and its unit is Newton. Let be the partial derivative of the Hamiltonian function with respect to the generalized momentum. Let be the partial derivative of the Hamiltonian function with respect to the generalized coordinates. A fourth-order Singlonga-Kutta method is used for time discretization, and the energy conservation error is calculated after each time step. The formula is expressed as follows: ; In the formula, For the error of energy conservation, and For the first Generalized coordinates and generalized momentum at each time step and For the first Generalized coordinates and generalized momentum at each time step This represents the L2 norm of a vector. When the energy conservation error exceeds a set threshold... At that time, the Lagrange multiplier-constrained projection method is used for correction, and a threshold is set. Experience value to Based on the estimated value of local truncation error Adjusting the time step using a proportional-integral controller algorithm The formula is expressed as follows: ; In the formula, For the first The time step, in seconds. For the first The time step, in seconds. To allow for error, the empirical value is to , This represents the local truncation error of the current step. This is the local truncation error from the previous step. Let be the order of the method; for a fourth-order method, take . Repeat the above steps until the numerical integration of the entire time domain is completed, outputting the position evolution trajectory, velocity evolution trajectory, and stress evolution trajectory of each point on the conductor to form deformation evolution trajectory data. Then, determine whether the conductor stress exceeds the material's yield strength. Labeling of safety risk level, material yield strength The equation for the catenary conductor is determined based on the conductor material. ; In the formula, The vertical coordinates of the traverse are in meters. Horizontal tension, measured in Newtons. Weight per unit length, expressed in Newtons per meter. The coordinates are horizontal, and the unit is meters. For reference horizontal length, the unit is meters, and the empirical value is 100. (·) represents a hyperbolic cosine function. When the dominant mode frequency drift in the deformation evolution trajectory data exceeds 3% of the initial frequency or the mode energy transfer rate exceeds 20%, the sampling density of the deformation evolution trajectory data is increased.

[0061] Furthermore, but not limited to setting the sliding window length to 1024 sampling points and the window movement step size to 128 sampling points, the data for the first time window is extracted from the deformation evolution trajectory data. The multi-channel deformation data within the window are arranged in chronological order to form a state matrix, where each column of the state matrix represents the deformation state of all monitoring points in the system at a certain moment. Singular value decomposition is performed on the state matrix, and the top few dominant modes are selected based on the criterion that the cumulative singular value energy ratio reaches 99%. The time-shifted pairs of the state matrix are constructed by combining the dominant modes, and the linear mapping matrix between the two is solved using the least squares method as an approximation of the dynamic operator. Then, the eigenvalues ​​and eigenvectors of the dynamic operator matrix are calculated, converting the modulus of the eigenvalues ​​into modal amplitude attenuation rates, the argument of the eigenvalues ​​into modal oscillation frequencies, and the eigenvectors into mode shapes in physical space. These parameters constitute the instantaneous natural frequency parameters, damping ratio parameters, and mode shape parameters. Each time the sliding window moves forward by one step, new observation data is acquired, and the dynamic operator matrix is ​​updated using a recursive least squares algorithm. This algorithm updates parameter estimates using only newly arrived data increments, eliminating the need to reprocess all historical data and significantly reducing computational complexity. The ratio of the difference between the first-order modal frequencies of the current and previous windows to the initial frequency is calculated as the dominant modal frequency drift, and the sum of the absolute values ​​of the changes in the energy proportions of each mode is calculated as the modal energy transfer rate. These steps are repeated until the entire deformation evolution trajectory data is processed, outputting the instantaneous modal parameter evolution curves for each moment. These curves constitute a time-varying modal characteristic sequence. The adaptive dynamic modal decomposition method achieves instantaneous modal tracking of time-varying systems through a sliding window and recursive update mechanism, without assuming the linearity and time invariance of the system. It is suitable for non-stationary deformation analysis of transmission lines under continuously changing external excitation conditions. The algorithm described in this embodiment can capture the time evolution of natural frequencies and damping ratios. When the conductor's stiffness degrades due to fatigue damage, the natural frequency will drift towards lower frequencies. These abnormal changes in modal parameters provide sensitive indicators for early warning of structural health status.

[0062] In application, the length of the sliding window can be set. and window movement step Sliding window length The empirical value is 50 to 200 sampling points, with a window movement step size. The empirical value is 10 to 50 sampling points. Data from the first time window is extracted from the deformation evolution trajectory data, and the multi-channel deformation data within the time window are arranged in chronological order to form a state matrix. The dimension is the number of channels multiplied by the window length. Singular value decomposition is performed on the state matrix, as expressed in the following formula: ; In the formula, It is a left singular vector matrix, with dimensions equal to the number of channels multiplied by the number of channels. It is a singular value diagonal matrix, with dimensions equal to the number of channels multiplied by the window length. This is a right singular vector matrix, with dimensions equal to the window length multiplied by the window length. The top singular values ​​are selected based on their energy proportions. Each dominant mode, constructing the time-shifted state matrix and ,in Before the state matrix A matrix composed of columns After the state matrix A matrix composed of columns. The dynamic operator matrix is ​​solved using the least squares method. The formula is expressed as follows: ; In the formula, This is a dynamic operator matrix, with dimensions equal to the number of channels multiplied by the number of channels. for The pseudo-inverse matrix. Where the pseudo-inverse matrix is... The calculation formula is expressed as follows: ; In the formula, for The transpose of the matrix. Calculate the eigenvalues ​​of the dynamic operator matrix. and eigenvectors The eigenvalues ​​are converted into the modal frequencies of the continuous-time system. Damping ratio The formula is expressed as follows: ; In the formula, For the first First-order modal frequency, in Hertz. For the first 1 eigenvalue, It is the natural logarithm function. This is the sampling time interval, in seconds. To take the real part of a complex number, For the first Damping ratio, For the first Each eigenvector is converted into a mode shape in physical space. The modal frequencies, damping ratios, and mode shapes constitute the instantaneous natural frequency parameters, damping ratio parameters, and mode shape parameters. The sliding window moves forward by a step size. The dynamic operator matrix is ​​updated using a recursive least squares algorithm. The update formula for the recursive least squares algorithm is expressed as follows: ; ; ; In the formula, For the first The updated dynamic operator matrix, For the first The dynamic operator matrix of the step, The gain matrix has the same dimensions as... same, For the first The latter part of the time shift matrix of the step, For the first The first part of the time shift matrix of the step, for The transpose of the matrix, Let be the error covariance matrix. To measure the noise covariance matrix, the empirical value is... , It is the identity matrix. This is the updated error covariance matrix. The modal frequency difference between the current window and the previous window is calculated as the dominant modal frequency drift. The formula is expressed as follows: ; In the formula, The dominant modal frequency shift, The frequency of the first mode in the current time window, in Hertz. The first-order modal frequency is given in the initial healthy state, expressed in Hertz. The change in the energy percentage of each mode is calculated as the modal energy transfer rate. The formula is expressed as follows: ; In the formula, Mode energy transfer rate, For the current time window First-order mode energy, measured in joules. For the initial health state First-order mode energy, measured in joules. The number of dominant modes is determined. The above steps are repeated until the entire deformation evolution trajectory data is processed, and the instantaneous modal parameter evolution curves at each time point are output to form a time-varying modal feature sequence.

[0063] Furthermore, the process of constructing the safety assessment model includes the construction of training data and model training. The construction of training data involves combining topological graph data representation, time-varying modal feature sequences, dominant modal frequency drift, modal energy transfer rate, and safety risk level labels according to a data structure to form a training dataset for the transmission line safety risk assessment model. The dominant modal frequency drift and modal energy transfer rate are used as damage-sensitive indicators to quantify changes in structural health status. The dominant modal frequency drift reflects the degree of structural stiffness degradation. The modal energy transfer rate reflects the degree of vibration mode change. Deformation monitoring data of transmission lines under different meteorological conditions, load levels, and line operating years are collected. Finite element simulation software is used to simulate deformation response data under extreme conditions to supplement hazardous scenarios not covered by measured data. Feature extraction and modal identification were performed on each data set. Safety risk level labels were assigned based on whether the conductor stress exceeded the material yield strength, whether the tower tilt angle exceeded the safety limit of 5 degrees, and whether the deformation rate exceeded the warning threshold of 0.5 m / s. The labeled data were then divided into training and validation sets in an 8:2 ratio. Data augmentation was performed on the training set by adding Gaussian noise with a standard deviation of 5% of the measured value to simulate sensor measurement error in the deformation data, and by randomly pruning and splicing the time series data to increase sample diversity. This dataset provides rich sample and label information for model training.

[0064] During model training, a transmission line safety risk assessment model architecture integrating graph neural networks and a topology-aware deformation propagation algorithm is first constructed. This model architecture includes multiple functional layers. The input layer receives topology graph deformation data representations and time-varying modal feature sequences. Graph convolutional layers aggregate neighborhood information for each graph node through a message passing mechanism. This mechanism calculates the information propagation weights between graph nodes based on the edge connections, simulating the deformation propagation process along conductors and towers. Multi-scale graph pooling layers perform hierarchical downsampling of the topology graph structure. Coarse-grained pooling layers aggregate adjacent tower node positions into region nodes, while fine-grained pooling layers retain the local features of individual tower node positions, extracting topological features at different granularities. Graph attention layers calculate attention weight coefficients between graph node pairs, dynamically adjusting these coefficients based on the dominant modal frequency drift and modal energy transfer rate. When the dominant modal frequency drift exceeds 3% of the initial frequency or the modal energy transfer rate exceeds 20%, the attention weight coefficients of the corresponding graph node pairs increase, causing the model to focus on potential risk areas. Temporal convolutional layers perform temporal modeling of time-varying modal feature sequences, extracting the temporal dependencies of deformation evolution. Fully connected fusion layers concatenate and fuse graph features and temporal features. The output layer employs a multi-classification structure to output safety risk assessment results, including four categories: normal, low-risk, medium-risk, and high-risk. During training, after initializing the weight parameters of each layer, the training set data is input into the model in batches, and forward propagation is used to calculate the probability distribution of the predicted safety risk levels. A loss function is constructed, which includes a cross-entropy classification loss term and a physical constraint regularization term. The physical constraint regularization term includes the catenary equation residual and the stress-strain constitutive relation bias. The catenary equation residual is constructed by comparing the difference between the vertical displacement of the conductor predicted by the model and the calculated value of the catenary equation. The stress-strain constitutive relation bias is constructed by comparing the difference between the stress calculated by substituting the strain value predicted by the model into the constitutive relation and the measured stress. The weight coefficients of the physical constraint regularization term are dynamically adjusted according to the relative magnitudes of the data fitting error and the physical consistency error during training. When the data fitting error is large, the physical constraint weights are reduced to accelerate convergence; when the physical consistency error is large, the physical constraint weights are increased to ensure the model conforms to mechanical laws. The backpropagation algorithm is used to calculate the gradient of the loss function with respect to the weight parameters of each layer. The Adam optimizer is used to update the weight parameters, with a learning rate of 0.001 and a momentum parameter of 0.9. The model performance is evaluated on the validation set every 10 training batches. Classification accuracy and loss function values ​​are recorded. Training is terminated when the validation set loss function value does not decrease for 20 consecutive evaluation cycles. The weight parameters with the best performance on the validation set are saved as the completed transmission line safety risk assessment model.

[0065] This involves initializing the weight parameters for each layer and inputting the training set data into the model in batches for forward propagation. The cross-entropy classification loss term is also included. The formula is expressed as follows: ; In the formula, This refers to the batch sample size. This represents the total number of security risk categories, which defaults to 4. For the first The sample belongs to the first The actual label of the class, which takes the value 0 or 1. For the model to predict the first The sample belongs to the first The probability of the class. Constructing the loss function. The formula is expressed as follows: ; In the formula, For the total loss function, The residuals of the catenary equation for the conductor are... This represents the deviation from the stress-strain constitutive relation. and These are weighting coefficients, with empirical values ​​of 0.1 to 0.5 and 0.1 to 0.5, respectively. The dynamic adjustment formula for the weighting coefficients is expressed as follows: ; ; In the formula, and For the first Weight coefficients for each iteration and For the first Weight coefficients for each iteration For the first Cross-entropy classification loss value in rounds of iteration, For the first The residual values ​​of the catenary equation in the round iteration. For the first The deviation of the stress-strain constitutive relation in the cycle iteration, To prevent extremely small values ​​from being divided by zero, the empirical value is... The residual expression for the catenary equation is as follows: ; In the formula, This represents the total number of conductor nodes. The model predicts the first Vertical displacement of each node, in meters. The first calculation for the catenary equation Vertical displacement of each node, in meters. For reference height, the unit is meters, and the empirical value is 100. Stress-strain constitutive relation deviation. The statement is as follows: ; In the formula, The model predicts the first The stress is calculated by substituting the strain values ​​at each node into the stress-strain constitutive relation, and the unit is Pascal. For the first Measured stress at each node, in Pascals. The reference stress is expressed in Pascals, and its empirical value is [value missing]. The Adam optimizer is used to update the weight parameters, with a learning rate of 0.001 and a momentum parameter of 0.9. The model performance is evaluated on the validation set every 10 training batches. Training is terminated when the validation set loss function value does not decrease for 20 consecutive evaluation cycles. The weight parameters with the best performance on the validation set are saved as the completed transmission line safety risk assessment model.

[0066] Based on the above, the key technical approach of this application lies in utilizing a symplectic geometric-preserving structural deformation dynamics numerical integration algorithm. By strictly maintaining the symplectic geometric invariance of the Hamiltonian system, it avoids the numerical dissipation and spurious damping effects of traditional explicit integration methods, enabling the algorithm to maintain high fidelity even in long-term simulations spanning several years or even decades. This provides a reliable numerical tool for extrapolating long-term damage evolution trends from short-term monitoring data, solving the problem of inaccurate fatigue life prediction caused by the rapid degradation of accuracy in long-term integration using traditional methods. The proposed adaptive time step control strategy dynamically adjusts the calculation accuracy based on local truncation errors, automatically reducing the step size to capture details during periods of rapid deformation change and increasing the step size to improve efficiency during stable periods, thus achieving optimal allocation of computational resources. The adaptive dynamic mode decomposition algorithm achieves instantaneous mode tracking of time-varying systems through a sliding window and recursive update mechanism, without assuming the linearity and time invariance of the system, and is suitable for non-stationary deformation analysis of transmission lines under continuously changing external excitation conditions.

[0067] Based on this, the algorithm in this embodiment can capture the time evolution trajectory of the natural frequency and damping ratio. When the stiffness of the conductor degrades due to fatigue damage, the natural frequency will drift towards lower frequencies. These abnormal changes in modal parameters provide sensitive indicators for early warning of structural health status, overcoming the limitation of traditional frequency domain analysis methods that assume the system is time invariant and thus cannot identify progressive damage.

[0068] Graph neural networks and the deformation propagation algorithm for transmission line topology perception explicitly encode the physical topology of transmission lines into a topological graph structure, enabling the model to perform global inference of the deformation field using the connectivity and mechanical transmission paths of the lines. The message passing mechanism of the graph convolutional layer simulates the propagation process of deformation along conductors and towers. When abnormal deformation occurs at a monitoring point, the model can track the diffusion trend of deformation to adjacent nodes through graph edge relationships, thereby predicting the deformation state at locations without sensors. The dynamic weight adjustment mechanism of the graph attention layer assigns higher attention to potentially risky areas based on the degree of anomaly in modal parameters, improving sensitivity to early signs of damage. The physical constraint regularization term embeds the catenary equation of the conductor and the material stress-strain constitutive relation into the loss function, guiding the model to learn feature representations that conform to mechanical laws. This enhances the model's generalization ability in sparse training data regions and its prediction reliability under extreme conditions, overcoming the shortcomings of purely data-driven methods that are prone to overfitting and lack physical interpretability under small sample conditions.

[0069] The synergistic effect of these multiple features enables a complete chain from long-scale high-fidelity simulation to time-varying modal sensitivity identification and then to topology-aware intelligent reasoning. The high-quality deformation evolution data generated by the symplectic geometry algorithm provides reliable input for modal decomposition. The damage-sensitive index parameters extracted by modal decomposition provide an adaptive adjustment basis for the attention mechanism of the graph neural network. The graph neural network utilizes topological information and physical constraints to achieve accurate assessment and early warning of transmission line safety risks. Compared with traditional methods that rely solely on single-point monitoring data and empirical threshold judgments, the assessment model constructed in this invention has higher prediction accuracy, stronger generalization ability, and better interpretability.

[0070] In another embodiment, step S1 may also include the step of periodically monitoring the surface deformation of the transmission line corridor using spaceborne synthetic aperture radar differential interferometry (D-InSAR). This can acquire large-scale spatial distribution information of tower foundation settlement and ground deformation field with millimeter-level accuracy, making up for the inability of point sensors to cover the entire transmission line corridor. Specifically, multiple Sentinel-1 or ALOS-2 SAR images covering the transmission line corridor are selected, with the initial operation image as the master image and subsequent images as slave images. After accurate registration, removal of flat phase, phase unwrapping, and atmospheric phase correction, a cumulative settlement map of the tower foundation in each monitoring period is generated. The SAR differential interferometric deformation measurement results are fused with the displacement data of the contact sensor to jointly constitute the spatial deformation component of the original deformation time series dataset.

[0071] In another embodiment, step S3 may further include extracting the long-term deformation rate at each tower corner reflector using Permanent Scatterer Interferometry (PS-InSAR) technology based on temporal SAR imagery. This step can obtain the millimeter-level annual average settlement rate and cumulative deformation at each tower node location over several consecutive years, providing high-precision long-term deformation history information for the graph node features of the topology map. Specifically, a trihedral corner reflector is installed at each tower as an artificial SAR corner reflector. A permanent scatterer interferometry algorithm is used to identify highly coherent scatterers from the temporal SAR image stack, and the phase time series at each corner reflector location is extracted. After fitting a linear deformation rate model and processing the residual phase, the annual average settlement rate, acceleration component, and seasonal deformation amplitude at each tower node location are obtained. These parameters are then concatenated with the compressed deformation feature vector and used as the graph node feature input, enriching the temporal dimension information of the topology graph deformation data representation.

[0072] Furthermore, step S4 may also include a step of extracting the three-dimensional spatial distribution information of the conductor sag deformation using SAR tomography and cross-validating it with the adaptive dynamic mode decomposition results. This step can verify the mode identification results from an independent microwave remote sensing observation perspective, improving the reliability of the calculation of the dominant mode frequency drift and mode energy transfer rate. Specifically, a multi-baseline or multi-temporal SAR tomography processing method is used to perform tomographic focusing processing on multiple SAR images of the same span conductor sag position, reconstructing the three-dimensional deformation field of the conductor in the radar line of sight. The vertical displacement time series at the midpoint of the span extracted by SAR tomography is input into the adaptive dynamic mode decomposition algorithm to independently calculate the dominant mode frequency, which is then compared with the mode frequency calculated from the contact sensor data. When the deviation between the two exceeds 5%, a data quality warning is triggered; when the deviation is within the threshold, the weighted average is taken as the final mode frequency estimate, thereby reducing the impact of single sensor measurement errors on damage-sensitive index parameters.

[0073] Specifically, the amplitude and number distribution of stress cycles are statistically analyzed using the rainflow counting method, and the allowable number of cycles corresponding to each stress cycle is calculated based on the fatigue curve of the conductor material. The formula for the fatigue curve is as follows: ; In the formula, For the first The allowable number of stress cycles. Let be a material constant, with an empirical value of . to , For the first The stress amplitude of a stress cycle, in Pascals. The reference stress is expressed in Pascals, and its empirical value is [value missing]. , The material fatigue index is empirically defined as 3 to 5. The cumulative damage degree is calculated according to the Miner linear cumulative damage criterion. The formula is expressed as follows: ; In the formula, To accumulate damage, This represents the total number of stress cycle types. For the first The actual number of stress cycles, For the first The allowable number of stress cycles is determined. External load scenarios for future time periods are randomly generated using the Monte Carlo method. For each external load scenario, a symplectic geometric preserver structural deformation dynamics numerical integration algorithm is run to calculate the stress cycle distribution and accumulate the cumulative damage. This random simulation is repeated several times. The time distribution when the cumulative damage reaches the critical value of 1.0 is statistically analyzed, and the expected value and confidence interval of the remaining life prediction are output.

[0074] like Figure 4 As shown, another embodiment of the present invention also provides a transmission line safety risk assessment system based on deformation information, comprising: The acquisition module is used to collect monitoring data from various monitoring points of the transmission line. The monitoring data includes conductor displacement data, strain data, and temperature data. The monitoring data forms the original deformation time series dataset of the transmission line. The first processing module is used to obtain a compressed deformation feature vector from the original deformation time series dataset through principal component analysis and regularized sparse coding. The construction module is used to construct a transmission line topology graph structure corresponding to the actual topology of the transmission line by combining the compressed deformation feature vector. The tower nodes in the transmission line are mapped to graph nodes in the transmission line topology graph structure, and the conductor spans in the transmission line are mapped to graph edges in the transmission line topology graph structure. The second processing module is used to process the topology of the transmission line based on the numerical integration algorithm of the deformation dynamics of the symplectic geometric preserved structure to obtain conductor deformation evolution trajectory data. The first determining module is used to perform mode identification on the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm, and determine the time-varying mode feature sequence, dominant mode frequency drift and mode energy transfer rate based on the identification results; The third processing module is used to call the pre-built safety risk assessment model to process the topological graph variable data representation, time-varying modal feature sequence, dominant mode frequency drift, and mode energy transfer rate to determine the safety risk assessment result of the transmission line.

[0075] In one embodiment, obtaining a compressed deformation feature vector from the original deformation time series dataset through principal component analysis and regularized sparse coding includes: Principal component analysis was used to extract principal component components whose cumulative contribution rate met a preset threshold from the original deformation time series dataset through dimensionality reduction processing. The principal component components are subjected to regularized sparse coding to obtain the compressed deformable feature vector.

[0076] In one embodiment, the step of extracting principal component components whose cumulative contribution rate meets a preset threshold through principal component analysis and dimensionality reduction of the original deformation time series dataset includes: The original deformation time series dataset is standardized, and the covariance matrix is ​​calculated based on the standardized original deformation time series dataset. The covariance matrix is ​​decomposed into eigenvalues ​​and then sorted in descending order according to the decomposed eigenvalues. Based on the arrangement results, eigenvalues ​​with a cumulative contribution rate of 95% are selected to construct the projection matrix; The principal component components are obtained by multiplying the original deformation time series dataset with the projection matrix.

[0077] In one embodiment, the step of performing regularized sparse coding on the principal component components to obtain the compressed deformable feature vector includes: The principal component components are subjected to feature sparsification processing using the L1 regularized sparse coding algorithm to obtain the compressed deformable feature vector.

[0078] In one embodiment, the numerical integration algorithm based on symplectic geometric preserves structural deformation dynamics is used to process the transmission line topology to obtain conductor deformation evolution trajectory data, including: Based on the topology of the transmission line, a finite element discrete model of the conductor is established to divide the conductor into multiple units, and the node position and velocity of each unit are respectively used as generalized coordinates and generalized momentum. The system of equations of motion for the conductor is constructed based on Hamiltonian canonical equations, and the system of equations of motion has symplectic structure properties. The fourth-order symplectic Runge-Kutta method is used to discretize the equations of motion in time, and the generalized coordinates and generalized momentum in each time step are calculated. The Hamiltonian function value is calculated based on the generalized coordinates and generalized momentum of each time step, and the energy conservation error is judged based on the Hamiltonian function value to determine whether it meets the preset requirements. If the requirements are met, the calculation of the next time step continues until the calculation of the entire time domain is completed, and the conductor deformation evolution trajectory data is obtained.

[0079] In one embodiment, the device further includes: The correction module is used to correct the generalized coordinates and generalized momentum of the current time step based on the constrained projection method of Lagrange multipliers if the energy conservation error determined based on the Hamiltonian function value does not meet the preset requirements. The second determining module is used to determine the local truncation error estimate of the generalized coordinates and generalized momentum of the current time step; An adjustment module is used to adjust the step size of the next time step based on the estimated local truncation error.

[0080] In one embodiment, the modal identification of the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm includes: The deformation evolution trajectory data of the conductor is captured by a sliding window, and the multi-channel deformation data in the sliding window is arranged in chronological order into a state matrix. The column vectors of the state matrix represent the deformation state of the conductor at different times. The state matrix is ​​subjected to singular value decomposition, and the dominant mode and secondary mode are determined based on the proportion of singular value energy after decomposition. Based on the dominant mode, the time-shifted pairs of the state matrix are constructed, and the linear mapping matrix between the time-shifted pairs is solved by the least squares method as the dynamic operator matrix. The eigenvalues ​​and eigenvectors of the dynamic operator matrix are determined, and the instantaneous natural frequency parameters and damping ratio parameters are determined based on the eigenvalues. The mode shape parameters in physical space are determined based on the eigenvectors.

[0081] In one embodiment, determining the time-varying modal feature sequence, dominant mode frequency shift, and modal energy transfer rate based on the identification results includes: The modal frequency difference between the two windows is calculated based on the dynamic operator matrix of the current window and the previous window, and the modal frequency difference is defined as the dominant modal frequency drift. The change in the energy percentage of each mode is determined based on the state matrix of the current window and the previous window, and the change is defined as the mode energy transfer rate. The instantaneous modal parameter evolution curve is determined based on all window data, and the instantaneous modal parameter evolution curve is defined as the time-varying modal feature sequence.

[0082] In one embodiment, constructing the security risk assessment model includes: The initial evaluation model is constructed, including an input layer, a graph convolutional layer, a multi-scale graph pooling layer, a graph attention layer, a temporal convolutional layer, a fully connected fusion layer, and an output layer. Training data is formed based on the topological graphical representation of transmission lines under different environments and operating conditions during historical periods, time-varying modal feature sequences, dominant mode frequency drift, mode energy transfer rate, and obtained safety risk level labels. The initial evaluation model is trained based on the training data to obtain the security risk evaluation model.

[0083] Another embodiment of the present invention also provides an electronic device, comprising: One or more processors; Memory, configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the transmission line safety risk assessment method based on deformation information as described above.

[0084] Furthermore, one embodiment of the present invention also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the transmission line safety risk assessment method based on deformation information as described above. It should be understood that the various solutions in this embodiment have the corresponding technical effects in the above-described method embodiments, and will not be repeated here.

[0085] Furthermore, embodiments of the present invention also provide a computer program product, which is tangibly stored on a computer-readable medium and includes computer-readable instructions, which, when executed, cause at least one processor to perform a transmission line safety risk assessment method based on deformation information, such as the embodiment described above.

[0086] It should be noted that the computer storage medium of the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access storage medium (RAM), a read-only storage medium (ROM), an erasable programmable read-only storage medium (EPROM or flash memory), an optical fiber, a portable compact disk read-only storage medium (CD-ROM), an optical storage medium, a magnetic storage medium, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program configured for use by or in connection with an instruction execution system, system, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, antenna, optical fiber, RF, etc., or any suitable combination thereof.

[0087] Furthermore, those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0088] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; under the concept of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

Claims

1. A method for assessing the safety risks of transmission lines based on deformation information, characterized in that, include: Monitoring data from various monitoring points along the transmission line are collected. The monitoring data includes conductor displacement data, strain data, and temperature data. The monitoring data forms the original deformation time series dataset of the transmission line. The original deformation time series dataset is subjected to principal component analysis and regularized sparse coding to obtain compressed deformation feature vectors; By combining the compressed deformation feature vector, a transmission line topology graph structure corresponding to the actual topology of the transmission line is constructed. The tower nodes in the transmission line are mapped to graph nodes in the transmission line topology graph structure, and the conductor spans in the transmission line are mapped to graph edges in the transmission line topology graph structure. The topology of the transmission line is processed using a numerical integral algorithm based on symplectic geometric preserved structure deformation dynamics to obtain conductor deformation evolution trajectory data. The adaptive dynamic mode decomposition algorithm is used to identify the mode of the conductor deformation evolution trajectory data, and the time-varying mode feature sequence, dominant mode frequency drift and mode energy transfer rate are determined based on the identification results. The pre-built safety risk assessment model is invoked to process the topological graph variable data representation, time-varying modal feature sequence, dominant mode frequency drift, and mode energy transfer rate to determine the safety risk assessment result of the transmission line.

2. The method for assessing the safety risk of transmission lines based on deformation information according to claim 1, characterized in that, The process of obtaining compressed deformation feature vectors from the original deformation time series dataset through principal component analysis and regularized sparse coding includes: Principal component analysis was used to extract principal component components whose cumulative contribution rate met a preset threshold from the original deformation time series dataset through dimensionality reduction processing. The principal component components are subjected to regularized sparse coding to obtain the compressed deformable feature vector.

3. The method for assessing the safety risk of transmission lines based on deformation information according to claim 2, characterized in that, The process of extracting principal component components from the original deformation time series dataset through principal component analysis and dimensionality reduction to obtain principal component components whose cumulative contribution rate meets a preset threshold includes: The original deformation time series dataset is standardized, and the covariance matrix is ​​calculated based on the standardized original deformation time series dataset. The covariance matrix is ​​decomposed into eigenvalues ​​and then sorted in descending order according to the decomposed eigenvalues. Based on the arrangement results, eigenvalues ​​with a cumulative contribution rate of 95% are selected to construct the projection matrix; The principal component components are obtained by multiplying the original deformation time series dataset with the projection matrix.

4. The method for assessing the safety risk of transmission lines based on deformation information according to claim 2, characterized in that, The step of performing regularized sparse coding on the principal component components to obtain the compressed deformable feature vector includes: The principal component components are subjected to feature sparsification processing using the L1 regularized sparse coding algorithm to obtain the compressed deformable feature vector.

5. The method for assessing the safety risk of transmission lines based on deformation information according to claim 1, characterized in that, The numerical integration algorithm based on symplectic geometric preserved structure deformation dynamics is used to process the topology of the transmission line to obtain conductor deformation evolution trajectory data, including: Based on the topology of the transmission line, a finite element discrete model of the conductor is established to divide the conductor into multiple units, and the node position and velocity of each unit are respectively used as generalized coordinates and generalized momentum. The system of equations of motion for the conductor is constructed based on Hamiltonian canonical equations, and the system of equations of motion has symplectic structure properties. The fourth-order symplectic Runge-Kutta method is used to discretize the equations of motion in time, and the generalized coordinates and generalized momentum in each time step are calculated. The Hamiltonian function value is calculated based on the generalized coordinates and generalized momentum of each time step, and the energy conservation error is judged based on the Hamiltonian function value to determine whether it meets the preset requirements. If the requirements are met, the calculation of the next time step continues until the calculation of the entire time domain is completed, and the conductor deformation evolution trajectory data is obtained.

6. The method for assessing the safety risk of transmission lines based on deformation information according to claim 5, characterized in that, The method further includes: If the energy conservation error does not meet the preset requirements based on the Hamiltonian function value, the generalized coordinates and generalized momentum of the current time step are corrected based on the constrained projection method of the Lagrange multipliers. Determine the local truncation error estimates for the generalized coordinates and generalized momentum of the current time step; Adjust the step size of the next time step based on the estimated local truncation error.

7. The method for assessing the safety risk of transmission lines based on deformation information according to claim 1, characterized in that, The modal identification of the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm includes: The deformation evolution trajectory data of the conductor is captured by a sliding window, and the multi-channel deformation data in the sliding window is arranged in chronological order into a state matrix. The column vectors of the state matrix represent the deformation state of the conductor at different times. The state matrix is ​​subjected to singular value decomposition, and the dominant mode and secondary mode are determined based on the proportion of singular value energy after decomposition. Based on the dominant mode, the time-shifted pairs of the state matrix are constructed, and the linear mapping matrix between the time-shifted pairs is solved by the least squares method as the dynamic operator matrix. The eigenvalues ​​and eigenvectors of the dynamic operator matrix are determined, and the instantaneous natural frequency parameters and damping ratio parameters are determined based on the eigenvalues. The mode shape parameters in physical space are determined based on the eigenvectors.

8. The method for assessing the safety risk of transmission lines based on deformation information according to claim 7, characterized in that, Based on the identification results, the time-varying modal feature sequence, dominant mode frequency shift, and modal energy transfer rate are determined, including: The modal frequency difference between the two windows is calculated based on the dynamic operator matrix of the current window and the previous window, and the modal frequency difference is defined as the dominant modal frequency drift. The change in the energy percentage of each mode is determined based on the state matrix of the current window and the previous window, and the change is defined as the mode energy transfer rate. The instantaneous modal parameter evolution curve is determined based on all window data, and the instantaneous modal parameter evolution curve is defined as the time-varying modal feature sequence.

9. The method for assessing the safety risk of transmission lines based on deformation information according to claim 1, characterized in that, The security risk assessment model is constructed by including: The initial evaluation model is constructed, including an input layer, a graph convolutional layer, a multi-scale graph pooling layer, a graph attention layer, a temporal convolutional layer, a fully connected fusion layer, and an output layer. Training data is formed based on the topological graphical representation of transmission lines under different environments and operating conditions during historical periods, time-varying modal feature sequences, dominant mode frequency drift, mode energy transfer rate, and obtained safety risk level labels. The initial evaluation model is trained based on the training data to obtain the security risk evaluation model.

10. A transmission line safety risk assessment system based on deformation information, characterized in that, include: The acquisition module is used to collect monitoring data from various monitoring points of the transmission line. The monitoring data includes conductor displacement data, strain data, and temperature data. The monitoring data forms the original deformation time series dataset of the transmission line. The first processing module is used to obtain a compressed deformation feature vector from the original deformation time series dataset through principal component analysis and regularized sparse coding. The construction module is used to construct a transmission line topology graph structure corresponding to the actual topology of the transmission line by combining the compressed deformation feature vector. The tower nodes in the transmission line are mapped to graph nodes in the transmission line topology graph structure, and the conductor spans in the transmission line are mapped to graph edges in the transmission line topology graph structure. The second processing module is used to process the topology of the transmission line based on the numerical integration algorithm of the deformation dynamics of the symplectic geometric preserved structure to obtain conductor deformation evolution trajectory data. The first determining module is used to perform mode identification on the conductor deformation evolution trajectory data based on the adaptive dynamic mode decomposition algorithm, and determine the time-varying mode feature sequence, dominant mode frequency drift and mode energy transfer rate based on the identification results; The third processing module is used to call the pre-built safety risk assessment model to process the topological graph variable data representation, time-varying modal feature sequence, dominant mode frequency drift, and mode energy transfer rate to determine the safety risk assessment result of the transmission line.

Images